35
Novel AI-Based Soft Computing Applications in Motor Drives
Centre for Engineering Studies, Energy Research, University of Trinidad and Tobago UTT, Point Lisas Campus, Esperanza Road, Brechin Castle, Couva. P. O. Box 957
35.1 Introduction
35.2 Differences Between GA and PSO and Other Evolutionary Computation (EC) Techniques
35.3 Single Objective Genetic Optimization Search Algorithm (SOGA)
35.4 Single Objective Particle Swarm Optimization Search Algorithm (SOPSO)
35.5 Multi-Objective Optimization (MOO)
35.6 Multi-Objective Genetic Optimization Search Algorithm (MOGA)
35.7 Multi-Objective Particle Swarm Optimization Search Algorithm (MOPSO)
35.8 GA and PSO Applications in Speed Control of Motor Drives
35.8.1 Efficient Operation of Induction Motor Drives
35.8.2 A Novel Self-Regulating Hybrid (PV–FC–Diesel–Battery) Electric Vehicle-EV Drive System
35.8.3 Novel Self-Regulating Green Plug/Energy Management/Energy Economizer GP–EM–EE Schemes for Wind Driven Induction Generation
35.9 Conclusion
35.1 Introduction
The chapter introduces novel ideas in the efficient control of industrial motor drives. The use of Artificial Intelligence AI-Soft Computing techniques is gaining acceptance in the industry to ensure efficient energy utilization, robust and accurate speed reference tracking, minimum impact on the host electric grid system and extended life span of the motor drive system while avoiding magnetic saturation and inrush current conditions. The use of emerging soft computing random search techniques such as Particle Swarm Optimization (PSO) and Genetic Algorithm (GA) is also gaining acceptance as effective tools in dynamic on-line tuning, self-regulating, and adjustment of controller to ensure that specific objective functions are optimized. The use of dynamic online self-adjusting error driven control strategies is promising effective, robust, and efficient control strategies for DC and AC motor drives. The use of new topologies, converter architecture, and switching strategies is greatly facilitated by multi-loop regulation and dynamic absolute error minimization using PSO and GA search algorithms. In this chapter, a review of PSO and GA search methods is followed by single and multi-objective optimization search algorithms and how both soft computing AI-based techniques can be applied to optimal operation of induction motor drives, hybrid photovoltaic-Fuel Cell-Diesel-Battery Electric Vehicle drive system and finally the novel self-regulating Green Plug-Energy Management-Energy Economizer Schemes developed by the First Author as a number of an extended family of modulated power filters, switched capacitor compensators, and dynamic electric energy management devices and system.
Solving an optimization problem is one of the common scenarios that occur in most engineering applications. Classical optimization techniques such as Linear Programming LP and Non-Linaer Programming NLP are efficient approaches that can be used to solve special cases of optimization problem in power system and motor drive applications. As the complexities of the problem increase, especially with the introduction of parameter uncertainties, more complicated optimization techniques such as stochastic programming have to be used. However, these analytical methods are not easy to implement for most of the real-world problems. As a highly nonlinear, nonstationary motor drive system with nonlinear inertia, load, noise, and uncertainties, motor drives can have a large number of operating condition states and parameters. Implementing any of the classical analytical optimization search methods may not be feasible in most of the cases. On the other hand, GA and PSO search techniques can be attractive, alternative, and effective solutions. GA and PSO are stochastically based search techniques that have its roots in Artificial Life and Social Psychology, as well as in Engineering and Computer Science. In general, there are two possible optimization techniques based on GA and PSO. The two techniques are:
The main procedure of the SOO is based on deriving a single objective (SO) function for the optimization search problem. The SO function may be combined from several objective functions using weighting factors. The objective function is optimized (either minimized or maximized) using GA or PSO to obtain a single solution. On the other hand, the main objective of the Multi-Objective (MO) problem is finding the set of acceptable trade-off optimal solutions. This set of accepted solutions is called Pareto front. These acceptable solutions give more ability to the user to make an informed decision by seeing a wide range of solutions that are optimum from an “overall” standpoint. SO optimization may ignore this trade-off viewpoint. This chapter has described the basic concepts of GA and PSO and presents a review of some of the applications of GA and PSO in motor drive-based optimization problems to give the readers some insight of how GA and PSO can serve as acceptable near optimal solutions to some of the most complicated engineering optimization problems.
35.2 Differences Between GA and PSO and Other Evolutionary Computation (EC) Techniques
A comparison between Conventional Optimization Techniques and evolutionary algorithms (like GA and PSO) is presented in Table 35.1 [1].
• Unlike other random search algorithms, each potential solution (called a particle) is also assigned a randomized velocity and then flown through the problem hyperspace.
• The most striking difference between PSO and the other evolutionary soft computing algorithms is that PSO chooses the path of cooperation over competition. The other algorithms commonly use some form of decimation, survival of the fittest. In contrast, the PSO population is stable and individuals are not destroyed or created. Individuals are influenced by the best performance of their neighbors. Individuals eventually converge on optimal points in the problem domain.
• The PSO traditionally does not have any genetic operators like crossover between individuals and mutation, and other individuals never substitute particles during the run. Instead, the PSO refines its search by attracting the particles to positions with good solutions.
• Particles update themselves with the internal velocity.
• They also have memory, which is important to the algorithm.
• Compared with GAs, the information sharing mechanism in PSO is significantly different. In GAs, chromosomes share information with each other. So, the whole population moves like a one group toward an optimal area. In PSO, only gbest or pbest gives out the information to others. It is a one-way information sharing mechanism. The evolution only looks for the best solution.
• Compared to the GA, the advantages of PSO are that PSO is easy to implement and there are few parameters to adjust.
TABLE 35.1 Comparison between conventional optimization procedures and evolutionary algorithms
| Property | Evolutionary | Traditional |
| Search space | Population of potential solutions | Trajectory by a single point |
| Motivation | Natural selection and Social adaptation | Mathematical properties (gradient, Hessian) |
| Applicability | Domain independent, Applicable to variety of problems | Applicable to a specific problem domain |
| Point Transition | Probabilistic | Deterministic |
| Prerequisites | An objective function to be optimized | Auxiliary knowledge such as gradient vectors |
| Initial guess | Automatically generated by the algorithm | Provided by user |
| Flow of control | Mostly parallel | Mostly serial |
| CPU time | Large | Small |
| Results | Global optimum more probable | Local optimum, dependant of initial guess |
| Advantages | Global search, parallel, speed | Convergence proof |
| Drawbacks | No general formal convergence proof | Locality, computational cost |
35.3 Single Objective Genetic Optimization Search Algorithm (SOGA)
GAs are an evolutionary optimization approach which is an alternative to traditional optimization methods. GA is most appropriate for complex nonlinear models where the location of the global optimum is a difficult task. It may be possible to use GA techniques to consider problems which may not be modeled as accurately using other approaches.
Therefore, GA appears to be a potentially useful approach. GA is particularly applicable to problems which are large, nonlinear, and possibly discrete in nature, features that traditionally add to the degree of complexity of solution. Due to the probabilistic development of the solution, GA does not guarantee optimality even when it may be reached. However, they are likely to be close to the global optimum. This probabilistic nature of the solution is also the reason they are not contained by local optima. The GA procedure is based on the Darwinian principle of survival of the fittest. An initial population is created containing a predefined number of individuals (or solutions), each represented by a genetic string (incorporating the variable information). Each individual has an associated fitness measure, typically representing an objective value. The concept that fittest (or best) individuals in a population will produce fitter offspring is then implemented in order to reproduce the next population. Selected individuals are chosen for reproduction (or crossover) at each generation, with an appropriate mutation factor to randomly modify the genes of an individual, in order to develop the new population. The result is another set of individuals based on the original subjects leading to subsequent populations with better (min. or ma35.) individual fitness. Therefore, the algorithm identifies the individuals with the optimizing fitness values, and those with lower fitness will naturally get discarded from the population [2]. Figure 35.1 shows the general flow chart of the GA algorithm based on the total error iterative minimum search. The steps of the GA are depicted as follows:
1. [Start] Generate random population of n chromosomes (suitable solutions for the problem)
2. [Fitness] Evaluate the fitness f (x) of each chromosome x in the population
3. [New population] Create a new population by repeating the following steps until the new population is complete:
(a) [Selection] Select two parent chromosomes from a population according to their fitness (the better fitness, the bigger chance to be selected).
(b) [Crossover] With a crossover probability cross over the parents to form a new offspring (children). If no crossover was performed, the offspring is an exact copy of parents.
(c) [Mutation] With a mutation probability mutate new offspring at each locus (position in chromosome).
(d) [Accepting] Place the new offspring in a new population.
4. [Replace] Use new generated population for a further run of algorithm
5. [Test] If the end condition is satisfied, stop, and return the best solution in current population[Loop] Go to step 2
FIGURE 35.1 Flow chart for the GA minimizing search algorithm.
GAs can be applied to many scientific, engineering problems, once solutions of a given problem can be encoded to chromosomes in GA, and compare the relative performance (fitness) of solutions. An effective GA representation and meaningful fitness evaluation are the keys of the success in GA applications. The appeal of GAs comes from their simplicity and elegance as robust search algorithms as well as from their power to discover good solutions rapidly for difficult high-dimensional problems. The main advantage of GA is that models which cannot be developed using other solution methods without some form of approximation can be considered in an un-approximated form. Size of the model, i.e., number of probabilistic variables, has a significant effect on the speed of solution therefore model specification can be crucial. Unlike other solution methods, integer variables are easier to accommodate in GA than continuous variables. This is due to the resulting restricted search space. Further, variable bound values can be applied to achieve similar results. GAs can be used for problem-solving and for modeling when the search space is large, complex, or poorly understood, domain knowledge is scarce or expert knowledge is difficult to encode to narrow the search space, no mathematical analysis is available, and traditional search methods fail.
35.4 Single Objective Particle Swarm Optimization Search Algorithm (SOPSO)
PSO is an evolutionary computational technique (a search method based on a natural system), which was introduced by Kennedy and Eberhart in 1995 [3]. This optimization and search technique models the natural swarm behavior seen in many species of birds returning to roost, group of fish, and swarm of bees… etc. The PSO may be used to find optimal (or near optimal) solutions to numerical and qualitative problems [4–8]. PSOs methods are inspired by particles moving around in the defined search-space. The individuals in a PSO have a position and a velocity. The PSO method remembers the best position found by any particle. Additionally, each particle remembers its own previously best-found position. A particle moves through the specified solution space along a trajectory defined by its velocity, the draw to return to a previous promising search area, and an attraction toward the best location discovered by its close neighbors. PSO has been used for a wide range of search applications, as well as for specific optimization tasks. PSO can be easily implemented in most programming languages and has proven to be both effective and fast when applied to a diverse set of nonlinear optimization problems. PSO has been successfully applied in many areas:
• Artificial neural network training;
• Proportional and integral fuzzy system control; and
• Other near optimal search and optimization areas where GA can be applied.
35.4.1 Structure of the PSO Particle
The basic structure of any particle in a selected population consists of five components:
• 
is a vector containing the current location in the solution space. The size of is defined by the number of variables used by the problem that is being solved.
• Fitness is the quality of the solution represented by the vector , as computed by a problem-specific evaluation function.
• 
is a vector containing the velocity for each dimension of . The velocity of a dimension is the step size that the corresponding value will change into at the next iteration. Changing the values changes the direction the particle will move through in the search space, causing the particle to make a turn. The velocity vector is used to control the range and resolution of the search.
• Pbest is the fitness value of the best solution yet found by a particular particle.
• 
is the copy of the for the location that generated the particle's Pbest Jointly, Pbest and comprise the particle's memory, which is used to control the particle to go back toward a definite search region.
• Each particle is also aware of the current best fitness in the neighborhood for any given iteration. A neighborhood may consist of some small group of particles, in which case the neighborhoods overlap and every particle is in multiple neighborhoods. Particles in a swarm are related socially; that is, each particle is a member of one or more neighborhoods. Each individual tries to emulate the behavior of the best of its neighbors. Each individual can be thought of as moving through the feature space with a velocity vector that is influenced by its neighbors.
35.4.2 Basic Search Method
The position of each particle is represented by XY axis position; and also the velocity is expressed by Vx (the velocity of X axis) and Vy (the velocity of Y axis). Modification of the particle position is realized by the position and velocity information. Each particle knows its best value so far (Pbest) and its XY position. This information represents the personal experiences of each particle. Moreover, each particle knows the best value so far in the group (gbest) among Pbests. This information represents the knowledge of how the other particles around have performed. Namely, each particle tries to modify its position using the following information:
• The current positions (x, y);
• The current velocities (Vx, Vy);
This modification can be represented by the concept of velocity. Velocity of each particle can be modified by the following equation:
(35.1)
where:
• Vid is the value of dimension d in the velocity vector 
for particle i;
• C1 is the cognitive learning selected rate;
• C2 is the social learning selected rate;
• rand1 and rand2 are random values on the range [0.1];
• Xid is the current position of particle i along dimension d;
• W is the selected weighting factor;
• Pid is the location along dimension d at which the particle previously had the best fitness measure; and
• Pgd is the current location along dimension d of the neighborhood particle with the best fitness.
The basic concept of the PSO technique lies in accelerating each particle toward its Pbest and gbest locations, with a random weighted acceleration at each step and this is illustrated in Fig. 35.2,
FIGURE 35.2 Concept of modification of a searching point by PSO.
where
Using the above concept, a certain velocity, which gradually gets close to Pbest and gbest can he calculated. The current position (searching point in the solution space) can be modified by the following equation:
(35.2)
35.4.3 PSO Search Algorithm
Figure 35.3 shows the general flow chart of the PSO algorithm. The main steps in the PSO process are described as follows:
1. System initialized with a population of random potential solutions. Each potential solution is assigned a random “velocity” and is called a particle (It has position in the space; i.e., it is a point in the solution space and it has velocity). These particles are then “flown” through the search space of potential solutions.
2. Evaluate the fitness of each particle in the swarm.
3. For every iteration, compare each particle's fitness with its previous best fitness (Pbest) obtained. If the current value is better than Pbest, then set Pbest equal to the current value and the Pbest location equal to the current location in the d-dimensional space.
4. Compare Pbest of particles with each other and update the swarm global best location with the greatest fitness (gbest).
5. The velocity of each particle is changed (accelerated) toward its Pbest and gbest. This acceleration is weighted by a random term. A new position in the solution space is calculated for each particle by adding the new velocity value to each component of the particle's position vector.
6. Repeat steps (2)–(5) until convergence is reached based on some desired single or multiple objective criteria.
FIGURE 35.3 The general flow chart of PSO search algorithm.
35.5 Multi-Objective Optimization (MOO)
In many real-life applications, multiple and often conflicting objectives need to be satisfied. Satisfying these conflicting objective functions is called Multi-Objective Optimization (MOO). For example, to place more functional blocks on a chip while minimizing that chip's area and/or power dissipation are conflicting objectives that need performing a tradeoff analysis [9, 10]. The objective of MO optimization is to find a set of acceptable solutions and present them to the user, who will then choose from them. Generally, there are two general approaches to solve MOO. The first approach lies in combining the individual objective functions into a single composite function. Determination of a SO is possible with methods, such as the weighted sum method, but the problem lies in the correct selection of the weights. In practice, it can be very difficult to accurately select these weights, even for someone very familiar with the problem domain. In addition, optimizing a particular solution with respect to a SO can result in unacceptable results with respect to the other objectives [11, 12]. The second general approach is to obtain the optimal solution. There will be a set of optimal trade-offs between the conflicting objectives, but this optimal solution is called Pareto optimal solution set or Pareto front [13, 14]. A Pareto optimal set is a set of solutions that are nondominated with respect to one another. While moving from one Pareto solution to another, there is always a certain amount of importance in one objective to achieve a certain amount of gain in the other. Generating the Pareto set has several advantages. The Pareto set allows the user to make an informed decision by seeing a wide range of options. The Pareto set contains the solutions that are optimum from an “overall” standpoint. SO optimization may ignore this trade-off viewpoint. This feature is useful since it provides better understanding of this system in which all the consequences of a decision with respect to all the objectives can be explored [9].
The following definitions are used in the proposed MOO search algorithm:
Def. 1 The general MO problem requiring the optimization of N objectives maybe formulated as follows:Minimize
(35.3)
(35.4)
(35.5)
is the objective vector, the 
1 represent the constraints and * is a P-dimensional vector representing the decision variables within a parameter space Ω. The space spanned by the objective vectors is called the objective space. The subspace of the objective vectors satisfying the constraints is called the feasible space.
Def. 2 A decision vector 1 ε Ω is said to dominate the decision vector 2 ε Ω (denoted by 1 
1), if the decision vector 1 is not worse than 2 in all objectives and strictly better than 2 in at least one objective.
Def. 3 A decision vector 1 ε Ω is called Pareto-optimal, if there does not exist another 2 ε Ω that dominates it. An objective vector is called Pareto-optimal, if the corresponding decision vector is Pareto-optimal.
Def. 4 The non-dominated set of the entire feasible search space Ω is the Pareto-optimal set. The Pareto-optimal set in the objective space is called Pareto-optimal front.
35.6 Multi-Objective Genetic Optimization Search Algorithm (MOGA)
The Non-Dominated Sorting Genetic Algorithm (NSGA) is a MO genetic algorithm that was developed by Deb et al. [15]. This algorithm has been chosen over a conventional genetic algorithm for three principal reasons: (a) no need to specify a sharing parameter, (b) a strong tendency to find a diverse set of solutions along the Pareto optimal front, and (c) the ability to specify multiple objectives without the need to combine them using a weighted sum. The basic idea behind NSGA is the ranking process executed before the selection operation, as shown in Fig. 35.4. This process identifies non-dominated solutions in the population, at each generation, to form non-dominated fronts [16], after this, the selection, crossover, and mutation usual operators are performed. In the ranking procedure, the non-dominated individuals in the current population are first identified. Then, these individuals are assumed to constitute the first non-dominated front with a large dummy fitness value [16]. The same fitness value is assigned to all of them. In order to maintain diversity in the population, a sharing method is then applied. Afterward, the individuals of the first front are ignored temporarily and the rest of population is processed in the same way to identify individuals for the second nondominated front. A dummy fitness value that is kept smaller than the minimum shared dummy fitness of the previous front is assigned to all individuals belonging to the new front. This process continues until the whole population is classified into non-dominated fronts. Since the non-dominated fronts are defined, the population is then reproduced according to the dummy fitness values.
FIGURE 35.4 Flowchart of NSGA.
35.7 Multi-Objective Particle Swarm Optimization Search Algorithm (MOPSO)
In Multi-Objective Particle Swarm Optimization (MOPSO) [9–14], a set of particles are initialized in the decision space at random. For each particle i, a position xi; in the decision space and a velocity vi; are assigned. The particles change their positions and move toward the so far best-found solutions. The non-dominated solutions from the last generations are kept in the archive. The archive is an external population, in which the so far found non-dominated solutions are kept. Moving toward the optima (maximum or minimum) is done in the calculations of the velocities as follows:
(35.6)
(35.7)
where Pr,d, Pp,d are randomly chosen from a single global Pareto archive, ω is the inertia factor influencing the local and global abilities of the algorithm, Vi,d is the velocity of the particle i in the dth dimension, c1 and C2 are weights affecting the cognitive and social factors, respectively, r1 and r2 are two uniform random functions in the range [0, 1]. According to (35.7), each particle has to change its position Xi,d toward the position of the two guides Pr,d, Pp,d which must be selected from the updated set of non-dominated solutions stored in the archive. The particles change their positions during generations until a termination criterion is met. Finding a relatively large set of Pareto-optimal trade-off solutions is possible by running the MOPSO for many generations. Figure 35.5(a) shows the flow chart of the MOPSO. Also Fig. 35.5(b) explains the procedure of the MOPSO using pseudo code.
FIGURE 35.5 Procedure of the MOPSO (a) Flow chart of the algorithm; (b) Pseudo code of MOPSO.
Init Pop(); The particle swarm is initialized with random values corresponding to the ranges of the decision variables. These values are dependent on the test functions.
Init Velocity(); The velocities are initialized with zero values.
Check the feasibility of each particle. If the particle does not satisfy the constraints, then regenerate it.
Evaluate Pop(); The swarm is evaluated using the corresponding objective functions.
Update Fbest(); The fitness vectors are updated (Evaluate the multi-objective fitness value of each particle and save it in vector form). As we are dealing with MOO, these vectors store the values of each decision variable, whereby the particles obtain the best values in a Pareto sense.
At this stage of the algorithm, these vectors are filled with the results of the initial particle evaluations.
Update Pbest(); Analogously, these values are copied in the pbest vectors.
Insert nodom(); Calculate the multi-objective fitness values of each particle and check its Pareto optimality. Store the non-dominated particles in the Pareto archive. If the specific constraint does not exist for the archive, the size of the archive will be unlimited (all non-dominated particles are inserted in the grid; i.e., in the external file).
Gbestpos = rnd (0,nodomfileSize); The global gbest particle is randomly selected.
Update Velocity(); The velocity of each particle is updated, using (35.6). Two Pareto solutions are chosen randomly for Pr,d,Pi,d from the Pareto archive.
Update Particle(); The position of each particle is also updated using (35.7)
Keeping(); The keeping operation is carried out to maintain the particles into the allowable range values. Then, the particles are mutated. If the particle does not remain within the feasible solution region, it is discarded and mutated again.
Evaluate Pop(); Update Fbest(); Update Pbest(); The particles are evaluated, the fitness, and pbest vector are, if appropriate, updated.
Insert nodom(); As the particles move in the search space because they have changed positions, the dominance of each particle is verified and, if appropriate, they are inserted in the grid.
(a) Check the Pareto optimality of each particle. If the fitness value of the particle is non-dominated when compared to the Pareto optimal set in the archive, save it into the Pareto archive.
(b) In the Pareto archive, if a particle is dominated by a new one, then discard it.
Gbestpos = rnd(0,nodomfileSize); Then, the new gbest is randomly selected. Two Pareto solutions are chosen randomly for pp;d and pr;d from the Pareto archive.
Repeat the cycle until the number of generations reaches a given n.
Print Statistics(); Generate Outfile(); Print the statistics and generate an output file, which contains the non-dominated particles.
35.8 GA and PSO Applications in Speed Control of Motor Drives
Many areas in power systems require solving one or more nonlinear optimization problems. While analytical methods might suffer from slow convergence and the problem of dimensionality, GA and PSO are known to effectively solve large-scale nonlinear optimization problems [17]. This section provides some motor drive applications that have benefited from the powerful nature of GA and PSO as optimization techniques. For each application, technical details that are required for applying GA and PSO and the most efficient fitness functions are also discussed. In this section, on-line energy optimization controllers are proposed for induction motor drives and PMDC motor Drives using GA and PSO. In many applications, efficiency optimization of induction motors and PMDC motors represents an important factor of control especially for autonomous electrical traction. This section will present some of the applications of GA and PSO in motor drives-based optimization problems to give some insight of how GA and PSO can serve as a solution to some of the most complicated engineering optimization problems [18–43].
35.8.1 Efficient Operation of Induction Motor Drives [18]
Induction motor is a high efficiency electrical machine when working close to its rated torque and speed. However, at light loads, no balance in between copper and iron losses, results considerable reduction in the efficiency. The part load efficiency and power factor can be improved by making the motor excitation adjustment in accordance with load and speed. To implement the above goal, the induction motor should either be fed through an inverter or redesigned with optimization algorithms [19]. This application presents an optimization technique for an efficient controller for the three-phase induction motors. MOPSO technique is implemented to tackle all the conflicting goals that define the search for the optimality problem. The PSO search deals with two main conflicting objective functions. These conflicting functions are: Maximizing the operating efficiency of the drive system for a given mechanical load and maximizing the equivalent power factor of the induction motor for start up and steady-state operation. In addition, the optimization ensures that maximum allowable stator current constraints are not exceeded. The proposed techniques are based on the principle that the flux level in a machine can be adjusted to give the required trade-off solution of maximum efficiency (ME) and maximum power factor (MPF) for a given value of speed and load torque. The optimum flux levels are functions of the machine load and speed requirements. Simulation results show that considerable efficiency and power factor improvements are achieved using MOPSO when compared with the Field Oriented Control (FOC) and Constant Voltage to Frequency Ratio based Control (CVFRC).
For optimal operation, three key objective functions that can affect the motor operation had been chosen. These three objective functions are based on motor efficiency, motor power factor, and motor stator current:
In addition, the optimization search ensures that all maximum stator current and inrush constraints are not exceeded. Figure 35.6 is a plot of input power versus stator current. As illustrated in this figure, the stator current and the input power are minimized almost simultaneously. Therefore, in practice the ME and the Minimum Stator Current (MSC) are not conflicting objective functions. So, when the efficiency is maximized, the stator current is minimized and the torque per ampere is maximized. So, the ME and power factor will be considered as the main conflicting objective functions. However, a good operation should represent the right compromise among different objectives but the problem consists in searching this “compromise.” The best tool to optimally solve this problem is represented by the multi-objective approach. This approach allows us to investigate how each single-objective and multi-objective problem affects the results in terms of performance and independent variables and, above all, allows us to have a wide range of alternative solutions among which the operator can choose a better solution. The MOGA and MOPSO algorithms have been applied to optimize the operation of a three-phase, 380 V, 1-HP, 50 Hz, 4-pole, squirrel cage induction motor. The block diagram of the optimization process based on PSO is shown in Fig. 35.7. In the proposed controller, the PSO algorithm receives the rotor speed and load torque, and then the PSO controller determines the slip frequency at which the optimal fitness function occurs at that rotor speed and load torque. As stated before, this part of simulation will consider the ME and the MPF problem to be solved using MOPSO. Figure 35.8 shows the solution to this problem and shows the Pareto fronts of the problem for different levels of rotor speed ωr = 0.2,0.4,0.8,1 PU and different levels of load torque TL = 0.2, and 1 PU. Tables 35.2 and 35.3 show the solution limits of the efficiency and the power factor for each operating point using MOGA and MOPSO. This range of solutions which is called the Pareto front enables the operator to choose the best compromise solution. Figure 35.10 shows the comparison between the CVFRS, the FOCS, and the available solutions from MOPSO at different levels of load torque and rotor speed ωr = 0.2, and 1 PU, respectively. The solution that has ME is selected from the Pareto front to achieve this comparison. Keep in mind that, the operator can choose the compromised solution from the Pareto front depending on the desired efficiency and power factor. It is obvious from Fig. 35.6 that the stator current is minimized using MOPSO. In addition, there is a great improvement in efficiency and power factor using MOPSO when compared with other strategies especially at light loads. It is noted that there is a very poor PF obtained using conventional methods (FOCS and CVFRS) especially at light loads. It is obvious from Fig. 35.9 that the efficiency improvement has a noticeable value especially at light loads and rotor speed ωr = 0.2 PU that can be as high as 80% using CVFRS. This difference decreases to 65% at rotor speed ωr = 1 PU. On the other hand, the power factor improvement reaches 55% at light load and rotor speed ωr = 0.2 PU Whereas this improvement reaches 160% at rotor speed ωr = 1 PU and light load. The performance difference between the FOC and the proposed control strategies MOPSO comes from the chosen value of the slip frequency and the air-gap flux for each strategy. For example, at loading condition of rotor speed equal to 0.2 PU and load torques varying from 0.2 to 1 PU, the values of slip frequency based on FOCS vary from 0.01 to 0.042 PU On the other hand, the values of slip frequency based on MOPSO vary from 0.0301 to 0.0524 PU This difference in slip frequency and air-gap flux causes the difference in performance of each control strategy. The difference in the flux level of the (FOCS) controller comes from the choice of the level of the magnetization current command. The command of magnetization current is set to a constant value, which produces rated torque at rated stator flu35. At light loads, the constant chosen value of magnetization current fails to choose the optimal flux level. On the other hand, this value yields the optimal flux level at rated loads. The proposed strategies overcome this problem and successfully choosing the optimal flux level especially at light loads.
FIGURE 35.6 Input power versus stator current at rated speed and various load powers.
FIGURE 35.7 The proposed drive system using PSO based on the loss model controller.
FIGURE 35.8 Pareto front of ME and MPF problem for different levels of rotor speed Ωr = 0.2,0.4,0.8,1 PU and different levels of load torque: (a) TL = 0.2 PU; (b) TL = 1 PU
TABLE 35.2 The limits pf the Pareto front of the two conflicting objective functions using MOGA
TABLE 35.3 The limits pf the Pareto front of the two conflicting objective functions using MOPSO
FIGURE 35.10 The proposed all-wheel electric vehicle drive system with PV-FC-Diesel-Battery AC/DC Generation.
FIGURE 35.9 The comparison between CVFRS, FOCS, and MOPSO for different levels of load torque TL and different levels of rotor speed: (a) Ωr = 0.2 PU;(b) ωr = 1PU.
35.8.2 A Novel Self-Regulating Hybrid (PV–FC–Diesel–Battery) Electric Vehicle-EV Drive System [20]
This application presents a number of novel self-regulating tri-loop error driven controllers for a hybrid PV–FC–Diesel–Battery powered all-wheel drive electric vehicle using Four Wheels Permanent Magnet DC (PMDC) motors, which are modeled to include existing nonlinearities in motor plus load inertia (J) and viscous friction (B). A Tri-Loop dynamic error driven scheme is proposed to regulate motor speed and current and avoid motor overloading/inrush conditions, in addition to motor speed dynamic reference tracking. The Proposed triloop dynamic error driven self-tuned control schemes utilized to ensure dynamic energy efficiency, control loop decoupling, grid interface stability while maintaining reference speed tracking capability. The integrated motor drive scheme is fully stabilized using a novel FACTS-based green filter compensators that ensures stabilized DC bus voltage, minimal inrush current conditions, and damped load excursions. This application presents a novel comparison of the MOPSO and Genetic search Algorithms MOGA optimization and search techniques for online dynamic tuning of the different controllers under varying renewable source conditions and load excursions.
The pressing need to utilize all abundant renewable green energy sources (Wind, Solar, PV, Wave, Tidal, Fuel Cell, Biogas, Hybrid…) is currently motivated by economic and environmental concerns. The increasing reliance on costly fossil fuels with increasing rate of resource depletion is causing a shift to energy alternatives, clean fuel replacement, and energy displacement of conventional sources to new green renewable, environmentally safe, and friendly counterparts [21, 22]. Electric vehicle is one of the solutions for the reduction of the fossil fuel consumption and pollutant emissions of gas responsible for the green house effect. However, pure battery electric vehicles have shown their own range limitations, because of their size the vehicle habitability is reduced so has its range. Adding different kinds of power supply in the same vehicle allows taking advantages from their different characteristics [23]. Several types of electric motors may be used for EV propulsion purposes. Earlier traction motors were exclusively dc motors, either series-excited or separately excited. Recently, more advanced ac drive systems have found application in EV propulsion using induction motors, permanent magnet synchronous motors, and permanent magnet brushless dc motors [24, 25]. The EV-DC motor speed or position control has been realized including conventional PI, PID, fuzzy logic based, nonlinear, adaptive variable structure, model reference adaptive control, artificial neural networks (ANN), feed forward computed torque control strategies [26–28]. The need for an on-line gains adaptation or a “tunable” control mechanism is highly stressed in the control of any nonlinear systems with un-modeled dynamics. The PSO and GA based self-regulating algorithms are utilized to track any reference speed trajectory under varying parameter and load conditions. The control system comprises of four different controllers used to track speed reference trajectory depicting the motor with minimum over current, inrush, ripple conditions. The proposed novel control scheme has been validated for effective dynamical speed reference trajectory tracking and enhanced power utilization.
35.8.2.1 Sample Study AC–DC System
Figures 35.10 and 35.11 show the proposed the Four-Wheel electric vehicle drive system scheme with the PV, FC sources, the diesel generator, and the backup battery. The DC compensator scheme developed by the First Author is used to ensure stable, efficient, minimal inrush operation of the hybrid renewable energy scheme. The novel PSO and GA self-tuned multi-regulators and coordinated controller are used for the following purposes:
1. Diesel AC generator set with control regulator is based on excess generation and load dynamic matching as well as stabilization of the common DC collection bus using six pulse controlled rectifier.
2. AC/DC power converter regulator to regulate the DC voltage at the Diesel engine AC/DC interface and ensure limited inrush conditions as well as dynamic power matching to reduce current transients and improve utilization at the diesel engine interface AC–DC bus.
3. The DC side Green plug filter compensator GPFC–SPWM regulator for pulse width switching scheme to regulate the DC bus voltage and minimize inrush current transients and load excursions and/or PV and FC non linear Volt-Ampere characteristics. The GPFC device acts as a matching DC–DC interface device between the DC load dynamic characteristics and that of the hybrid main PV, FC and backup diesel generator set.
4. The PMDC motor drive with the speed regulator that ensure speed reference tracking with minimum inrush conditions and ensure reduced voltage transients and improved energy utilization.
FIGURE 35.11 Schematic diagram of a prototype all-wheel drive electric vehicle using four PMDC motors.
The unified DC–AC utilization scheme is fully validated using the Matlab/Simulink software environment under normal conditions, DC load excursion, PMDC motor torque changes and the PV, and FC source output variations due to the inherent Volt–Ampere nonlinear relationship. Other excursion conditions in the diesel engine generator set are also introduced to assess the control system robustness, effective energy utilization, and speed reference tracking.
A. Diesel Generator Set
From an electrical system point of view, a diesel driven AC generator can be represented as a prime mover and generator. Ideally, the prime mover has the capability to supply any power demand up to rated power at constant synchronous frequency. The synchronous generator connected to it must be able to keep the voltage constant at any load condition. The diesel engine kept the operating speed and frequency constant. When power demand fluctuates the diesel generator could vary its power output via fuel valve regulation and governor control. The synchronous generator must control its output voltage by controlling its excitation current. Thus, the diesel generating system, as an auxiliary source, must be able to control its frequency and its output voltage. The ability of the diesel generator to respond to any frequency changes is affected by the inertia of the diesel gen-set, the sensitivity of the governor, and the power capability of the diesel engine. The ability of the AC synchronous generator to control its terminal voltage can be affected by the field-winding time constant, the availability of DC excitation power to supply the field winding, and the time constant of the voltage control loop.
B. Photo Voltaic PV
The equivalent circuit shown in Fig. 35.12 is used to model the PV cells used in the proposed PV array [29]. This model consists of a current source, a resistor, and a reverse parallel connected diode. The PVA model developed and used in Matlab/Simulink environment is based upon the circuit given in Fig. 35.12, in which the current produced by the solar cell is equal to that produced by the current source, minus that which flows through the diode, minus that which flows through the shunt resistor:
(35.8)
where I = output current, IL = photo generated current, ID = diode current, ISH = shunt current. The current through these elements is governed by the voltage across them:
(35.9)
where Vj = voltage across both diode and resistor RSH (volts), V = voltage across the output terminals (volts), I = output current (amperes), RS = series resistance (Ω).
FIGURE 35.12 The equivalent circuit of a solar cell.
The current diverted through the diode is
(35.10)
where I0 = reverse saturation current (amperes), n = diode ideality factor (1 for an ideal diode), q = elementary charge, k = Boltzmann's constant, T = absolute temperature.
The characteristic equation of a solar cell, which relates solar cell parameters to the output current and voltage [30]:
(35.11)
where RSH = shunt resistance (Ω). The I –V curve of an illuminated PV cell has the shape shown in Fig. 35.13 as the voltage across the measuring load is swept from zero to VOC. The power produced by the cell in Watts can be easily calculated along the I –V sweep by the equation P = IV. At the ISC and VOC points, the power will be zero and the maximum value for power will occur between the two.
FIGURE 35.13 The behavior of a solar cell at particular intensities of solar radiation.
C. Fuel Cell Energy System Model
Fuel cell stacks were connected in series/parallel combination to achieve the rating desired. Figure 35.14 shows a simplified diagram of the PEMFC system [31, 32]. The FC model here is for a type of PEM, which uses the following electrochemical reaction:
(35.12)
FIGURE 35.14 Simplified diagram of the Fuel Cell PEMFC system.
Figure 35.15 shows a simulated V –I (voltage versus current) polarization curve of a fuel cell [31, 32]. As the cell current begins to increase from zero, a sudden drop of the output voltage of the fuel cell is seen. This drop of the cell voltage is due to activation voltage loss. Then, almost a linear decrease of the cell voltage is seen as the cell current increases beyond certain values, as shown in Fig. 35.16, which is a result of the ohmic loss. Finally, the cell voltage drops sharply to zero as the load current approaches the maximum current density that can be generated of the fuel cell. The sharp voltage drop is the effect of the concentration loss in the fuel cell. The fuel cell can be commonly modeled by simple equivalent first order circuit shown in Fig. 35.10. The open circuit voltage is modified as follows:
(35.13)
where
(35.14)
where R = 8.3145 J/(mol K), F = 96485 A s/mol, z = Number of moving electrons, En = Nernst voltage, which is the thermodynamics voltage of the cells and depends on the temperatures and partial pressures of reactants and products inside the stack, i0 = Exchange current, which is the current resulting from the continual backward and forward flow of electrons from and to the electrolyte at no load. It also depends on the temperatures and partial pressures of reactants inside the stack, α = Charge transfer coefficient, which depends on the type of electrodes and catalysts used, T = Temperature of operation. The fuel cell voltage VFC is modeled as [31, 32]:
(35.15)
where
(35.16)
(35.17)
(35.18)
FIGURE 35.15 (V –I) polarization curve of an SOFC.
FIGURE 35.16 DC-Equivalent circuit of an electrochemical fuel cell.
Fuel cell stacks were connected in series/parallel combination to achieve the rating desired. The output of the fuel cell array was connected to a DC bus through a DC/DC converter. The DC bus voltage was kept constant via a DC bus voltage controller.
D. DC Side Green Plug Filter Compensator GPFC
The common concerns of power quality are the long duration voltage variations (overvoltage, under-voltage, and sustained interruptions), short duration voltage variations (interruption, sags, and swells), voltage imbalance (voltage unbalance), waveform distortion (DC offset, harmonics, inter-harmonics, notching and noise), voltage fluctuation (voltage flicker), and power frequency variations [33]. To prevent the undesirable states and to reduce the power consumption, a GPF scheme is used to stabilize the common DC bus.
E. Dynamic Error driven Control
The proposed control system developed by First Author comprises of four sub-regulators or controllers named as a Diesel DC generator set value control regulator, DC side Green Plug Filter Compensator GPFC–SPWM regulator, the PMDC motor drive speed controller, and the AC/DC power converter regulator. Figures 35.11–35.14 depict the proposed multi-loop dynamic self-regulating controllers based on MOO search and optimization technique based on soft computing PSO and GA. The global error is the summation of the three loop individual errors including voltage stability, current limiting, and synthesize dynamic power loops. Each multi-loop dynamic control scheme is used to reduce a global error based on a tri-loop dynamic error summation signal and to mainly track a given speed reference trajectory loop error in addition to other supplementary motor current limiting and dynamic power loops are used as auxiliary loops to generate a dynamic global total error signal that consists of not only the main loop speed error but also the current ripple, over current limit and dynamic over load power conditions.
The global error signal is input to the self-tuned controllers shown in Figs. 35.17–35.20. The (per-unit) three dimensional-error vector (evg, eIg, epg) of the diesel engine controller scheme is governed by the following equations:
(35.19)
(35.20)
(35.21)
FIGURE 35.17 Tri-loop error driven self-regulating VSC/SMC/B–B Controller for the common DC side –GPFC Scheme.
FIGURE 35.18 Tri-loop error driven self-regulating VSC/SMC/B–B Controller for the diesel engine generator set.
FIGURE 35.19 Tri-loop error driven self-regulating VSC/SMC/B–B Controller for dynamic speed control PMDC motor drive.
FIGURE 35.20 Tri-loop error driven self-regulating VSC/SMC/B–B Controller firing angle α- controller for Diesel AC/DC interface rectifier scheme.
The total or global error etg(k) at the AC side scheme at a time instant:
(35.22)
In the same manner, the (per-unit) three dimensional-error vector (evd, eId, epd) of the GPFC scheme is governed by the following equations:
(35.23)
(35.24)
(35.25)
And the total or global error etd (k) for the DC side green plug filter compensator GPFC scheme at a time instant:
(35.26)
In addition, the (per-unit) three dimensional-error vector (evR, eIR, epR) of the three-phase controlled rectifier scheme is governed by the following equations:
(35.27)
(35.28)
The total or global error etR (k) for the three-phase controlled converter rectifier scheme at a time instant:
(35.30)
Finally, the (per-unit) three dimensional-error vector (eωm, eIm, epm) of the PMDC motor scheme is governed by the following equations:
(35.32)
(35.33)
And the total or global error etm (k) for the MPFC scheme at a time instant:
(35.34)
A number of conflicting objective functions are selected to optimize using the PSO algorithm. These functions are defined by the following:
(35.35)
(35.36)
(35.37)
(35.38)
(35.39)
In general, to solve this multi-objective complex optimality search problem, there are two possible optimization techniques based on PSO: Single aggregate selected Objective Optimization (SOO), which is explained and MOO. The main procedure of the SOO is based on selecting a single aggregate objective function with weighted SO parameters scaled by a number of weighting factors. The objective function is optimized (either minimized or maximized) using either GA or PSO methods to obtain a single global or near optimal solution. On the other hand, the main objective of the MO problem is finding the set of acceptable (trade-off) Optimal Solutions. This set of accepted solutions is called Pareto front. These acceptable trade-off multi-level solutions give more ability to the user to make an informed decision by seeing a wide range of near optimal selected solutions that are feasible and acceptable from an “overall” standpoint. SO optimization may ignore this trade-off viewpoint, which is crucial. The main advantages of the proposed MOO method are: It doesn't require a priori knowledge of the relative importance of the objective functions and it provides a set of acceptable trade-off near optimal solutions. This set is called Pareto front or optimality trade-off surfaces. Both SOO and MOO searching algorithms are tested, validated, and compared. The dynamic error driven controller regulates the controllers’ gains using the PSO and GA to minimize the system total error, the settling time, the rising time, and the maximum overshoot. The proposed dynamic Tri-Loop Error Driven controller, developed by the First Author, is a novel advanced regulation concept that operates as an adaptive dynamic type multi-purpose controller capable of handling sudden parametric changes, load and/or DC source excursions. By using the Tri-Loop Error Driven controller, it is expected to have a smoother, less dynamic overshoot, fast, and more robust speed controller when compared to those of classical control schemes. The proposed general PMDC Motor Drive Model with the novel Tri-Loop Error Driven controller is fully validated in this application for effective reference speed trajectory tracking under different loading conditions and parametric variations such as temperature changes while driving a complex mechanical load with nonlinear parameters and/or torque-speed characteristics.
F. Self-tuned Variable Structure Sliding Mode Bang-Bang VSC/SMC/B–B Controller
In the variable structure sliding mode controller scheme, an optimally adaptive and self-tuned variable structure sliding mode controller for PMDC motor drive systems using PSO Technique and GA as shown in Fig. 35.21. The slope of the sliding surface is designed as
(35.40)
FIGURE 35.21 VSC/SMC/B–B Controller block diagram.
With adaptive adjustable error scaling gain
(35.41)
where
(35.42)
The system control voltage has the following form in the time domain [33]: the control is an on–off logic; that is: When σ > 0, Vc = > 1, and when σ < 0, Vc = –1
The PSO and GA search and optimizations are implemented for tuning the various gains β0, β1, and α to minimize the selected five conflicting objective functions (J1 –J5). This test is to verify that the proposed VSC/B–B with adaptive β can maintain the motor speed with nonlinear J (ωm) and B(ωm) even if the load changes; accordingly, the robustness can be confirmed.
G. Self-tuned Artificial Neural Network Controller ANN
The neural network used in this application is the simplest one that uses three layers, which is widely used in the control of the electrical machines. Each layer is composed of neurons. Each neuron is connected via weights to the previous layer. The first layer is connected to the input variables. The second one is connected via weights to all the neurons of the previous layer, and the last one is composed of one neuron given the output value. The weights and the biases of the ANN network's are updated to ensure that the global error of the system is minimized. The proposed ANN regulator is tuned on-line using the back-propagation algorithm. The on-line ANN rule-based algorithm is used to update the ANN network weights and biases to ensure continuous effective dynamic response while keeping the motor inrush current under specified tolerable limits. The input vector with 3-layer ANN as shown in Fig. 35.22 is
(35.43)
FIGURE 35.22 ANN incremental Controlled Controller block diagram.
H. Self-tuned Fuzzy Logic Controller (FLC)
As shown in Fig. 35.23, the FLC system consists of three subsystems which are the fuzzification, rule base, and defuzzification. Fuzzification subsystem converts the exact inputs to fuzzy values using five membership functions: Positive Big (PB), Positive Small (PS), Zero (ZZ), Negative Small (NS), and Negative Big (NB). The rule base unit processes these fuzzy values with fuzzy rules. The defuzzification unit converts the fuzzy results to exact values. The FLCs input values are the global error, et and change in global error, det. According to these variables, a rule table is produced in the FLCs rule base unit as shown in Table 35.4.
FIGURE 35.23 Main structure of the FLC-incremental scheme.
TABLE 35.4 Fuzzy rules decision table
35.8.2.2 Digital Simulation Results
The integrated micro grid for PMDC driven Electric Vehicle scheme using the Photo Voltaic PV, Fuel Cell FC, and backup Diesel generation with battery backup renewable generation system performance is compared for two cases, with fixed and self-tuned type controllers using either GA or PSO. The second case is to compare the performance with ANN controller and Fuzzy Logic Controller (FLC) with the self-tuned type controllers using either GA or PSO. The Tuned Variable structure sliding mode controller VSC/SMC/B–B has been applied to the speed tracking control of the same EV for performance comparison. There are three different speed references. In the first speed track, the speed increases linearly and reaches the 1 PU at the end of the first 5 s, and then the reference speed remains speed constant during 5 s. At the tenth second, the reference speed decreases with same slope as at the first 5 s. After 15s, the motor changes the direction and EV increases its speed through the reverse direction. At the twentieth second, the reference speed reaches the –1 PU and remains at a constant speed at the end of the twenty-fifth second and then the reference speed decreases and becomes zero at 30 s. The second reference speed waveform is sinusoidal and its magnitude is 1 PU and the period is 12 s. The third reference track is constant speed reference starting with an exponential track. In all references, the system responses have been observed. Matlab–Simulink Software was used to design, test, and validate the effectiveness of the integrated micro grid for PMDC driven Electric Vehicle scheme using Photo Voltaic PV, Fuel Cell FC, and backup Diesel generation with battery backup renewable generation system with the FACTS devices. The digital dynamic simulation model using Matlab/Simulink software environment allows for low cost assessment and prototyping, system parameters selection, and optimization of control settings. The use of PSO-search algorithm is used in online gain adjusting to minimize controller absolute value of total error. This is required before full scale prototyping which is both expensive and time consuming. The effectiveness of dynamic simulators brings on detailed sub-models selections and tested sub-models Matlab library of power system components already tested and validated. The common DC bus voltage reference is set at 1 PU.
Digital simulations are obtained with sampling interval Ts = 20μs. Dynamic responses obtained with GA are compared with ones resulting from the PSO for the seven proposed self-tuned controllers. The dynamic simulation conditions are identical for all tuned controllers. To compare the global performances of all controllers, the Normalized Mean Square Error (NMSE) deviations between output plant variables and desired values, and is defined as
(35.44)
(35.45)
The digital simulation results validated the effectiveness of both GA- and PSO-based tuned controllers in providing effective speed tracking minimal steady-state errors. Transients are also damped with minimal overshoot, settling time, and fall time. The GA- and PSO-based self-tuned controllers are more effective and dynamically advantageous in comparison with the ANN controller, the FLC, and fixed type controllers. The self-regulation is based on minimal value of absolute total/global error of each regulator shown in Figs. 35.17–35.20. The control system comprising the three dynamic multi-loop error driven regulator is coordinated to minimize the selected objective functions. SOO obtains a single global or near optimal solution based on a single weighted objective function. The weighted SO function combines several objective functions using specified or selected weighting factors as follows:
(35.46)
Where α1 = 0.20, α2 = 0.20, α3 = 0.20, α4 = 0.20, α5 = 0.20 are selected weighting factors. J1, J2, J3, J4, J5 are the selected objective functions. On the other hand, the MO finds the set of acceptable (trade-off) Optimal Solutions. This set of accepted solutions is called Pareto front. These acceptable trade-off multi-level solutions give more ability to the user to make an informed decision by seeing a wide range of near optimal selected solutions.
Table 35.5 shows the optimal solutions of the main objective functions versus the Tuned Variable structure sliding mode controller Gains based SOGA and MOGA control schemes. On the other hand, Table 35.6 shows the optimal solutions of the main objective functions versus the Tuned Variable structure sliding mode controller Gains based SOPSO and MOPSO control schemes. Table 35.7 shows the DC bus behavior comparison using the GA-based Tuned Variable structure sliding mode controller for the three selected reference tracks. In addition, Table 35.8 shows the system behavior using the PSO-based Tuned Variable structure sliding mode controller. Figures 35.24–35.29 show the effectiveness of MOPSO and MOGA search and optimized control gains in tracking the PMDC-EV motor three reference speed trajectories. Comparing the PMDC-EV dynamic response results of the two study cases, with GA and PSO tuning algorithms and traditional controllers with constant controller gains results shown in Table 35.9, ANN controller in Table 35.10 and Figs. 35.30–35.32, and FLC in Table 35.11 and Figs. 35.33–35.35, it is quite apparent that the GA and PSO tuning algorithms highly improved the PMDC-EV system dynamic performance from a general power quality point of view. The GA and PSO tuning algorithms had a great impact on the system efficiency improving it from 0.906631 (constant gains controller), 0.928253 (ANN controller), and 0.937334 (FLC) to around 0.948156 (GA-based tuned controller) and 0.930708 (PSO-based tuned controller) which is highly desired. Moreover, The Normalized Mean Square Error (NMSE-VDC–Bus) of the DC bus voltage is reduced from 0.08443 (constant gains controller), 0.04827 (ANN controller), and 0.03022 (FLC) to around 0.007304 (GA-based tuned controller) and 0.005854 (PSO-based tuned controller). In addition the Normalized Mean Square Error (NMSE-ωm) of the PMDC motor is reduced from 0.053548 (constant gains controller), 0.02627 (ANN controller), and 0.02016 (FLC) to around 0.0076308 (GA-based tuned controller) and 0.006309 (PSO-based tuned controller). Maximum Transient DC Voltage Over/Under Shoot (PU) is reduced from 0.054604 (constant gains controller), 0.04186 (ANN controller), and 0.03126 (FLC) to around 0.009302 (GA-based tuned controller) and 0.007259 (PSO-based tuned controller). Maximum Transient DC Current –Over/Under Shoot (PU) is reduced from 0.087336 (constant gains controller), 0.07355 (ANN controller), and 0.04383 (FLC) to around 0.00292 (GA-based tuned controller) and 0.005987 (PSO-based tuned controller). DC bus voltage (PU) is improved from 0.917020 (constant gains controller), 0.932736 (ANN controller), and 0.94745 (FLC) to around 0.97417 (GA-based tuned controller) and 0.974602 (PSO-based tuned controller).
TABLE 35.5 Selected objective functions versus the tuned variable structure sliding mode controller based SOGA and MOGA control schemes
TABLE 35.6 Selected Objective Functions versus the Tuned Variable structure sliding mode controller Gains based SOPSO and MOPSO control schemes
TABLE 35.7 DC bus behavior comparison using the GA-based Tuned Variable structure sliding mode controller VSC/SMC/B–B
TABLE 35.8 DC bus behavior comparison using the PSO based Tuned Variable structure sliding mode controller VSC/SMC/B–B
FIGURE 35.24 EV-PMDC Motor speed response for the first speed track using GA-based tuned Tri-loop variable structure sliding mode controller VSC/SMC/B–B.
FIGURE 35.25 EV-PMDC Motor speed response for the first speed track using PSO-based tuned Tri-loop variable structure sliding mode controller VSC/SMC/B–B.
FIGURE 35.26 EV-PMDC Motor speed response for the Second speed track using GA-based tuned Tri-loop variable structure sliding mode controller VSC/SMC/B–B.
FIGURE 35.27 EV-PMDC Motor speed response for the Second speed track using PSO-based tuned Tri-loop variable structure sliding mode controller VSC/SMC/B–B.
FIGURE 35.28 EV-PMDC Motor Speed response for the third speed track using GA-based tuned tri-loop variable structure sliding mode controller VSC/SMC/B–B.
FIGURE 35.29 EV-PMDC Motor speed response for the third speed track using PSO-based tuned Tri-loop Variable structure sliding mode controller VSC/SMC/B–B.
TABLE 35.9 DC bus behavior comparison using the constant parameters variable structure sliding mode controller VSC/SMC/B–B
TABLE 35.10 DC bus behavior comparison using ANN Controller
FIGURE 35.30 EV-PMDC Motor Speed response for the first speed track using ANN-based controller.
FIGURE 35.31 EV-PMDC Motor Speed response for the second speed track using ANN-based controller.
FIGURE 35.32 EV-PMDC Motor Speed response for the third speed track using ANN-based controller.
TABLE 35.11 DC bus behavior comparison using FLC Controller
FIGURE 35.33 EV-PMDC Motor Speed response for the first speed track using FLC-based controller.
FIGURE 35.34 EV-PMDC Motor Speed response for the second speed track using FLC-based controller.
FIGURE 35.35 EV-PMDC Motor Speed response for the third speed track using FLC-based controller.
DC bus current (PU) is reduced from 0.769594 (constant gains controller), 0.67464 (ANN controller), and 0.64712 (FLC) to around 0.614695 (GA-based tuned controller) and 0.607674 (PSO-based tuned controller). PMDCM total controller Error (eTM) is reduced from 0.095145 (constant gains controller), 0.04200 (ANN controller), and 0.02154 (FLC) to around 0.009167 (GA-based tuned controller) and 0.0048638 (PSO-based tuned controller). DC side GPFC Error (etd) is reduced from 0.70746 (constant gains controller), 0.03416 (ANN controller), and 0.02416 (FLC) to around 0.004618 (GA-based tuned controller) and 0.0074294 (PSO-based tuned controller). The diesel engine gen set total controller Error (etg) is reduced from 0.067513 (constant gains controller), 0.04507 (ANN controller), and 0.02964 (FLC) to around 0.005121 (GA-based tuned controller) and 0.007013 (PSO-based tuned controller). The diesel engine converter total controller Error (etR) is reduced from 0.086233 (constant gains controller), 0.03978 (ANN controller), and 0.0260 (FLC) to around 0.003265 (GA-based tuned controller) and 0.0053836 (PSO-based tuned controller).
35.8.3 Novel Self-Regulating Green Plug/Energy Management/Energy Economizer GP–EM–EE Schemes for Wind Driven Induction Generation
The application presents a novel self-adjusting wind energy utilization scheme using a modified single phase operation of the three-phase induction generator supplemented by a voltage stabilization switched filter compensation scheme developed by the First Author. The series-parallel switched capacitor filter scheme is controlled by a dynamic PSO and GA error driven self-adjusting controller to ensure voltage stabilization, minimum impact of the electric load excursions, and wind variations on terminal voltage. The sinusoidal pulse width modulation complementary switching scheme is dynamically controlled using on-line minimal global error search that continuously adjusts and modifies the controller gains. The application presents a family of novel switched smart filter compensated devices using Green Plug/Energy Management/Energy Economizer (GP–EM–EE) devices for small single phase induction motors (SPIM) used in residential/commercial motor drives used in water pumping, ventilation, air conditioning, compressors, refrigeration applications. The GP–EM–EE devices are equipped with a dynamic online error driven optimally tuned controller that ensures improved power factor, reduced feeder losses, stabilized voltage, minimal current ripples, and efficient energy utilization/conservation with minimal impact on the host electric grid security and reliability. The proposed schemes can enhance the power quality; extend induction motor life span by reducing overheating due to inrush currents and harmonics. They prevent overheating and possible motor damage. The scheme is suitable for small scale applications of wind energy utilization in the range from 5 to 25 KVA.
Wind Energy Conversion Systems (WECS) are increasing their number as an environmental friendly power generation system. Small sized wind turbine systems in the output power range of 5–25 KVA to be used for the residential houses or small sized business complex are extending their application fields. There are various advantages of the small size wind turbine system. The rotational speed of the smaller sized wind turbines can more rapidly follow the change of the wind speed than the large-scale wind turbines system due to their smaller inertia. This feature contributes higher power generation efficiency of the wind turbines in the case that the wind speed frequently changes [34]. Induction generators are being increasingly utilized in a WECS since they are relatively inexpensive, rigid, and require low maintenance [35].
Small scale induction motors drives consume over 50% of the total electrical energy generated in the developed countries [36]. The electric utility industry and consumers of electrical energy around the world are facing new challenges for cutting electric energy cost, improving energy utilization, enhancing energy-efficiency, demand-side management, improving supply waveform-power quality, reducing safety hazards to personnel, and protecting sensitive computer and automatic-data processing networks [37]. There is a mushrooming use of nonlinear electric loads especially in large motor drives, arc furnaces, and power electronic converter loads. All these nonlinear loads are byproduct of analog (saturation or Limiter type) or Digital (converter, solid state switching type) nonlinearities [38]. Nonlinear type loads cause severe waveform distortion, power quality problems interference, and extra feeder losses due to excessive inrush currents and severe voltage sags. The extended use of power electronic switching conveners and devices in motor drives, process-industries: Mining. Oil and Gas Industries and industrial DC and AC arc type furnaces have resulted in a polluted grid and unreliable radial distribution/utilization system with serious inherent voltage and power quality problems [39]. These nonlinear type electric loads are used with ventilation, air conditioning, water pumping, and low power factor industries such as sewing, printing, shear and press machinery, and food processing plants. These nonlinear loads also fall in the category of inrush or arc type motorized loads and combined with fluorescent lighting can cause waveform distortion, harmonic interference, and voltage flickering [40–42]. Generally, direct online motor starting is an economical method for starting induction motors. But direct starting will result in severe voltage sags and extra heating. When starting large induction motors, excessive voltage dips result in overheating and loss of motor life expectancy [43].
35.8.3.1 Sample Study Motorized System
Figure 35.36 depicts the block diagram of the utilization (single-phase induction motor SPIM) and the connection of switched smart filter compensated device using GP–EM–EE devices and the speed control drive system to the SPIM Load. Figures 35.37 and 35.38 show the proposed tri-loop dynamic tracking controller to ensure both objectives of (energy/power) saving as well as power quality enhancement of the supply system current and load bus voltage. The novel PSO and GA self-tuned multi-regulators and coordinated controller are used for the following purposes: (1) Green plug filter compensator GPFC–SPWM regulator for pulse width switching scheme to regulate the DC bus voltage and minimize inrush current transients and load excursions and (2) the SPIM drive with the speed regulator that ensure speed reference tracking with minimum inrush conditions and ensure reduced voltage transients and improved energy utilization. Figures 35.39–35.43 depict the proposed family of switched smart filter compensated devices using GP–EM–EE devices. All filters objectives can be either: (a) Harmonic reduction and power quality (PQ) enhancement or (b) Electric power/energy savings and dynamic reactive compensation for the SPIM loads. The proposed utilization scheme is fully validated using the Matlab/Simulink software environment under normal conditions, load excursion, SPIM motor torque changes to assess the control system robustness, effective energy utilization, and speed reference tracking.
FIGURE 35.36 The proposed switched smart filter compensated devices using GP–EM–EE devices for Single Phase Induction Motor (SPIM) drive system.
FIGURE 35.37 Tri-loop error driven self-regulating dynamic controller for control of SPIM drive.
FIGURE 35.38 Tri-loop error driven self-regulating dynamic controller for the GP–EM–EE scheme.
FIGURE 35.39 Low cost switched power filter compensator scheme-A.
FIGURE 35.40 Low cost switched power filter compensator scheme-B.
FIGURE 35.41 Low cost switched power filter compensator scheme-C.
FIGURE 35.42 Low cost switched power filter compensator scheme-D.
FIGURE 35.43 Low cost switched power filter compensator scheme-E.
The common concerns of power quality are the long duration voltage variations (overvoltage, under-voltage, and sustained interruptions), short duration voltage variations (interruption, sags, and swells), voltage imbalance (voltage unbalance), waveform distortion (DC offset, harmonics, inter-harmonics, notching and noise), voltage fluctuation (voltage flicker), and power frequency variations. To prevent the undesirable states and to reduce the power consumption, a GPF scheme is used to stabilize the system.
A. Dynamic Error driven Control
The proposed control system comprises two sub-regulators or controllers named as DC side Green Plug Filter Compensator GPFC-SPWM regulator and the SPIM drive speed controller. Figures 35.37 and 35.38 depict the proposed multi-loop dynamic self-regulating controllers based on MOO search and optimization technique based on soft computing PSO and GA. The global error is the summation of the three loop individual errors including voltage stability, current limiting, and synthesize dynamic power loops. Each multi-loop dynamic control scheme is used to reduce a global error based on a tri-loop dynamic error summation signal and to mainly track a given speed reference trajectory loop error in addition to other supplementary motor current limiting and dynamic power loops are used as auxiliary loops to generate a dynamic global total error signal that consists of not only the main loop speed error but also the current ripple, over current limit and dynamic over load power conditions.
The global error signal is input to the self-tuned controllers shown in Fig. 35.38. The (per-unit) three dimensional-error vector (evs, eIs, eps) of the diesel engine controller scheme is governed by the following equations:
(35.47)
(35.48)
(35.49)
The total or global error ets (k) for the GP–EM–EE side scheme at a time instant:
(35.50)
In the same manner, the (per-unit) three dimensional-error vector (eωm, eIm, epm) of the SPIM motor scheme is governed by the following equations:
(35.51)
(35.52)
(35.53)
And the total or global error etm (k) for the MPFC scheme at a time instant:
(35.54)
A number of conflicting objective functions are selected to optimize using the PSO algorithm. These functions are defined by the following:
(35.55)
(35.56)
(35.57)
(35.58)
(35.59)
In general, to solve this complex optimality search problem, there are two possible optimization techniques based on PSO: Single aggregate SOO, which is explained and MOO. The main procedure of the SOO is based on selecting a single aggregate objective function with weighted SO parameters scaled by a number of weighting factors. The objective function is optimized (either minimized or maximized) using either GA or PSO search algorithm (PSO) methods to obtain a single global or near optimal solution. On the other hand, the main objective of the MO problem is finding the set of acceptable (tradeoff) Optimal Solutions. This set of accepted solutions is called Pareto front. These acceptable trade-off multi-level solutions give more ability to the user to make an informed decision by seeing a wide range of near optimal selected solutions that are feasible and acceptable from an “overall” standpoint. SO optimization may ignore this trade-off viewpoint, which is crucial. The main advantages of the proposed MOO method are: it doesn't require a priori knowledge of the relative importance of the objective functions and it provides a set of acceptable tradeoff near optimal solutions. This set is called Pareto front or optimality trade-off surfaces. Both SOO and MOO searching algorithms are tested, validated, and compared.
The dynamic error driven controller regulates the controllers’ gains using the PSO and GA to minimize the system total error and the selected objective functions. The proposed dynamic Tri-Loop Error Driven controller, developed by the First Author, is a novel advanced regulation concept that operates as an adaptive dynamic type multi-purpose controller capable of handling sudden parametric changes, load and/or source excursions. By using the Tri-Loop Error Driven controller, it is expected to have a smoother, less dynamic overshoot, fast, and more robust controller when compared to those of classical control schemes.
B. Self-tuned conventional PID controller
Fundamentally, the conventional PID controller comprises three basic control actions. They are simple to implement and they provide good performance. The tuning process of the gains of PID controllers can be complex because is iterative: first, it is necessary to tune the “Proportional” mode, then the “Integral,” and then add the “Derivative” mode to stabilize the overshoot, then add more “Proportional,” and so on. The PID controller has the following form in the time domain as shown in Fig. 35.44:
(35.60)
FIGURE 35.44 Optimally tuned conventional PID controller block diagram.
Where e(t) is the selected system error, u(t) the control variable, Kp the proportional gain, Ki the integral gain, and Kd is the derivative gain. Each coefficient of the PID controller adds some special characteristics to the output response of the system. Because of this, choosing the right parameters becomes a crucial decision. In this scheme, the Tri-loop Error Driven Controller is utilized with traditional PID controller. PID controller gains (KP, KI, KD) are dynamically self-tuned using the PSO and GA dynamic search and optimization criterion based on total error minimization, steady-state error, maximum overshoot, settling time, and rising time.
35.8.3.2 Digital Simulation Results
The family of (GP–EM–EE) devices system performance is compared for two cases; without (as shown in Table 35.12) and with the (GP–EM–EE) devices, the second case is studied with fixed (as show in Table 35.13) and self-tuned type controllers using either GA or PSO. In addition, the second case is studied to compare the performance with ANN controller (as show in Table 35.14) and FLC (as show in Table 35.15) with the self-tuned type controllers using either GA or PSO. The self-tuned type controllers based either GA or PSO is Tuned conventional PID controller, has been applied to the speed tracking control of the same system parameters for performance comparison. Matlab–Simulink Software was used to design, test, and validate the effectiveness of the (GP–EM–EE) devices for small motors used in household appliances, washers, dryers, fans, water pumps, ventilation systems, air-conditions, and other applications in dispersing machines, actuators, and small converters with induction motor size from 5 to 25 KVA. The digital dynamic simulation model using Matlab/Simulink software environment allows for low cost assessment and prototyping, system parameters selection, and optimization of control settings. The use of GA and PSO- search algorithms are used in online gain adjusting to minimize controller absolute value of total error. This is required before full scale prototyping which is both expensive and time consuming. The effectiveness of dynamic simulators brings on detailed sub-models selections and tested sub-models Matlab library of power system components already tested and validated. The dynamic simulation conditions are identical for all tuned controllers. To compare the global performances of Tuned conventional PID controller, the NMSE deviations between output plant variables and desired values, and is defined as
(35.61)
(35.62)
(35.63)
TABLE 35.12 System behavior without (GP–EM–EE) schemes
| Without (GP–EM–EE) | |
| RMS Motor voltage (PU) | 0.8578 |
| RMS Motor current (PU) | 0.8724 |
| Maximum Transient Voltage Over/Under Shoot (PU) | 0.1797 |
| Maximum Transient Current –Over/Under Shoot (PU) | 0.1875 |
| System Efficiency | 0.7853 |
| NMSE_V | 0.4672 |
| NMSE_ωm | 0.6476 |
| NMSE_I | 0.4238 |
| THDv_Bus L (%) | 19.647 |
| THDi_Bus L (%) | 20.346 |
| THDv_Bus M (%) | 21.675 |
| THDi_Bus M (%) | 19.564 |
| Motor Power Factor | 0.7167 |
TABLE 35.13 System dynamic behavior comparison using the constant parameters conventional PID controller
TABLE 35.14 System dynamic behavior comparison using ANN controller
TABLE 35.15 System dynamic behavior comparison using the FLC controller
The control system comprising the three dynamic multi-loop error driven regulator is coordinated to minimize the selected objective functions. SOO obtains a single global or near optimal solution based on a single weighted objective function. The weighted SO function combines several objective functions using specified or selected weighting factors as follows:
(35.64)
Where α1= 0.20, α2 = 0.20, α3 = 0.20, α4 = 0.20, α5 = 0.20 are selected weighting factors. J1, J2, J3, J4, J5 are the selected objective functions. On the other hand, the MO finds the set of acceptable (trade-off) Optimal Solutions. This set of accepted solutions is called Pareto front. These acceptable trade-off multi-level solutions give more ability to the user to make an informed decision by seeing a wide range of near optimal selected solutions. Table 35.16 shows the system behavior using traditional controllers with constant controller gains for the (GP–EM–EE) schemes. In addition, Table 35.10 shows System behavior comparison using the SOGA-based Tuned conventional PID controller and Table 35.17 shows the system behavior comparison using the MOGA-based Tuned conventional PID controller. Finally, Tables 35.18 and 35.19 show system behavior comparison using the SOPSO- and MOPSO-based tuned conventional PID controller, respectively. Comparing the system dynamic response results of the two study cases, with GA and PSO tuning algorithms and traditional controllers with constant controller gains results, ANN controller and FLC, it is quite apparent that the GA and PSO tuning algorithms highly improved the system dynamic performance from a general power quality point of view. The GA and PSO tuning algorithms had a great impact on Motor RMS voltage (PU) is improved from 0.8578 (without the (GP–EM–EE) device), 0.9762 (constant gains controller), 0.9790 (ANN controller), and 0.9598 (FLC) to around 0.9804 (SOGA-based tuned controller), 0.9865 (MOGA-based tuned controller), 0.9887 (SOPSO-based tuned controller), and 0.9852 (MOPSO-based tuned controller). Motor RMS current (PU) is reduced from 0.8724 (without the (GP–EM–EE) device), 0.7029 (constant gains controller), 0.6837 (ANN controller), and 0.6863 (FLC) to around 0.6247 (SOGA-based tuned controller), 0.6304 (MOGA-based tuned controller), 0.6358 (SOPSO-based tuned controller), and 0.6471 (MOPSO-based tuned controller). Maximum Transient Motor Voltage Over/Under Shoot (PU) is reduced from 0.1797 (without the (GP–EM–EE) device), 0.0905 (constant gains controller), 0.0769 (ANN controller), and 0.0788 (FLC) to around 0.0472 (SOGA-based tuned controller), 0.0394 (MOGA-based tuned controller), 0.0441 (SOPSO-based tuned controller), and 0.0365 (MOPSO-based tuned controller). Maximum Transient Motor Current –Over/Under Shoot (PU) is reduced from 0.1875 (without the (GP–EM–EE) device), 0.0701 (constant gains controller), 0.0871 (ANN controller), and 0.0828 (FLC) to around 0.0431 (SOGA-based tuned controller), 0.0533 (MOGA-based tuned controller), 0.0412 (SOPSO-based tuned controller), and 0.0563 (MOPSO-based tuned controller). Moreover, the Normalized Mean Square Error (NMSE-V) of the Motor voltage is reduced from 0.4672 (without the (GP–EM–EE) device), 0.0828 (constant gains controller), 0.0790 (ANN controller), and 0.0844 (FLC) to around 0.00396 (SOGA-based tuned controller), 0.00429 (MOGA-based tuned controller), 0.00489 (SOPSO-based tuned controller), and 0.00440 (MOPSO-based tuned controller). In addition the (NMSE-ωm) of the SPIM motor is reduced from 0.6476 (without the (GP–EM–EE) device), 0.0809 (constant gains controller), 0.0715 (ANN controller), and 0.0838 (FLC) to around 0.00458 (SOGA-based tuned controller), 0.00608 (MOGA-based tuned controller), 0.00608 (SOPSO-based tuned controller), and 0.00513 (MOPSO-based tuned controller). The (NMSE-I) of the Motor current is reduced from 0.4238 (without the (GP–EM–EE) device), 0.0722 (constant gains controller), 0.0837 (ANN controller), and 0.0860 (FLC) to around 0.00361 (SOGA-based tuned controller), 0.00499 (MOGA-based tuned controller), 0.00602 (SOPSO-based tuned controller), and 0.00465 (MOPSO-based tuned controller). Total Harmonic Distortion THD (%) of the supply voltage is reduced from 19.647 (without the (GP–EM–EE) device), 7.8228 (constant gains controller), 10.4023 (ANN controller), and 8.2175 (FLC) to around 6.2535 (SOGA-based tuned controller), 4.8408 (MOGA-based tuned controller), 4.2581 (SOPSO-based tuned controller), and 4.1081 (MOPSO-based tuned controller). THD (%) of the supply current is reduced from 20.346 (without the (GP–EM–EE) device), 7.3532 (constant gains controller), 9.3573 (ANN controller), and 7.5181 (FLC) to around 4.1530 (SOGA-based tuned controller), 4.0103 (MOGA-based tuned controller), 5.2433 (SOPSO-based tuned controller), and 5.1378 (MOPSO-based tuned controller). THD (%) of the motor voltage is reduced from 21.675 (without the (GP–EM–EE) device), 8.0566 (constant gains controller), 8.5408 (ANN controller), and 9.9806 (FLC) to around 5.7887 (SOGA-based tuned controller), 4.9313 (MOGA-based tuned controller), 4.5873 (SOPSO-based tuned controller), and 4.3629 (MOPSO-based tuned controller). THD (%) of the motor current is reduced from 21.675 (without the (GP–EM–EE) device), 10.8589 (constant gains controller), 9.3120 (ANN controller), and 11.0353 (FLC) to around 3.9572 (SOGA-based tuned controller), 3.9998 (MOGA-based tuned controller), 5.1673 (SOPSO-based tuned controller), and 5.1216 (MOPSO-based tuned controller). The system efficiency is improved from 0.7853 (without the (GP–EM–EE) device), 0.8577 (constant gains controller), 0.8943 (ANN controller), and 0.9195 (FLC) to around 95.4132 (SOGA-based tuned controller), 95.4882 (MOGA-based tuned controller), 95.4999 (SOPSO-based tuned controller), and 90.7467 (MOPSO-based tuned controller). Motor power factor is improved from 0.7167 (without the (GP–EM–EE) device), 0.9082 (constant gains controller), 0.8609 (ANN controller), and 0.9169 (FLC) to around 95.1040 (SOGA-based tuned controller), 90.8137 (MOGA-based tuned controller), 91.7450 (SOPSO-based tuned controller), and 91.4433 (MOPSO-based tuned controller). Reduction in KWh Consumption (%) is reduced from 0.000 (without the (GP–EM–EE) device), 14.4937 (constant gains controller), 12.3129 (ANN controller), and 11.4634 (FLC) to around 18.3913 (SOGA-based tuned controller), 16.5388 (MOGA-based tuned controller), 15.9691 (SOPSO-based tuned controller), and 16.5328 (MOPSO-based tuned controller).
TABLE 35.16 System dynamic behavior comparison using the SOGA-based Tuned conventional PID controller
TABLE 35.17 System dynamic behavior comparison using a selected solution from the MOGA Pareto Frontbased Tuned conventional PID controller
TABLE 35.18 System dynamic behavior comparison using the SOPSO-based Tuned conventional PID controller
TABLE 35.19 System dynamic behavior comparison using a selected solution from the MOPSO Pareto Frontbased Tuned conventional PID controller
The application validated the operation of a novel wind driven-squirrel cage induction generator fed from single phase supply. The scheme utilizes a simple but effect series-parallel switched capacitor-filter to ensure dynamic terminal voltage stabilization, effective energy utilization, and minimal impact of electric load switching and dynamic excursions and wind velocity changes. The novel error driven self-regulating multi-loop dynamic controller utilizes random search optimization algorithms using PSO and GA to continuously search for effective and optimized controllers gains that can cope with specified objective functions restrictions of minimum voltage deviations, limited generated current and power excursions due to terminal load changes and wind velocity excursions and gusting. The application presents a family of novel low cost green plug electricity saving devices/Energy Management/Energy Economizer (GP–EM–EE) devices equipped with a dynamic online error driven optimally tuned controller using a dynamic online error driven optimally tuned controllers. The GP–EM–EE are a small low cost energy conservation devices in the form add-on dynamic switched capacitor filter compensator schemes for low horse power motors used in household appliances, washers, dryers, fans, water pumps, ventilation systems, air-conditioners, and other cyclical motorized loads used in dispersing machines, actuators, and small converters with induction motor size from 5 to 25 KVA. The GP–EM–EE devices family can be used with all types of small horse power motors with billions of small motors used in multitude of applications that consumes over 50–60% of total electrical energy generated in the world. The GP–EM–EE device can save 10–20% of the electricity cost of operating the small motor over the estimated motor life span of 7–10 years. This translate into Millions of dollars in daily electricity savings, reduced electrical utility overloading conditions, blackouts, brownouts and enhanced secure and reliable operation with additional sizable power capacity release due to reduced electric demand and feeder losses.
35.9 Conclusion
The chapter covers novel techniques for the dynamic control of industrial motor drives. The use of AI-Soft Computing PSO and GA random search techniques is presented. The specific use of PSO and GA SO and multi-objective controller optimization and dynamic gains adjusting is illustrated using three key applications of optimal operation of induction motor drives, hybrid photovoltaic–Fuel Cell–Diesel–Battery Electric Vehicle drive system and finally the novel self-regulating Green Plug-Energy Management-Energy Economizer Schemes developed by the First Author as a number of an extended family of modulated power filters, switched capacitor compensators, and dynamic electric energy management devices and system. The control strategies are all based on multi-loop error driven regulation with on line dynamic gain adjusting based on PSO and GA search algorithm. The use of AI Self-regulating dynamic controllers based on AI Soft computing in both novel and effective. It allows meeting a multitude of conflicting objective functions of effective speed reference tracking, minimal losses, efficient utilization, limited inrush currents under parametric variations, supply sources excursions, inertia, and mechanical load nonlinear loads. The other added benefit is targeting use of power electronic devices and switched capacitor compensator/filters in ensuring efficient operation of motors and motors drives especially in future applications using renewable and green energy system.
REFERENCES
1. Gudise VG, Venayagamoorthy GK. Evolving Digital Circuits Using Particle Swarm. In: 20–24 July 2003: 468–472. Proc. International Joint Conf. on Neural Networks; vol. 1.
2. Davis L. Handbooks of Genetic Algorithm. Reinhold, NY: Van Nostrand; 1991.
3. Kennedy J, Eberhart R. Particle swarm optimization. In: 1995: 1942–1948. Proc. IEEE International Conf on Neural Networks; vol. 4.
4. Shi Y, Eberhart R. Empirical study of particle swarm optimization. In: 1999: 1945–1950. Proc. Congress on Evolutionary Computation; vol. 3.
5. Shi Y, Eberhart R. A modified particle swarm optimizer. In: IEEE International Conf. on Evolutionary Computation. 1998: 69–73.
6. Eberhart R, Shi Y. Particle swarm optimization: Developments, applications and resources. In: 2001: 81–86. Proc. Congress on Evolutionary Computation; vol. 1.
7. Shi Y, Eberhart R. Parameter selection in particle swarm optimization. In: Proc. Seventh Annual Conf. on Evolutionary Programming. 1998: 591–601.
8. Fukuyama Y et al. A particle swarm optimization for reactive power and voltage control considering voltage security assessment. IEEE Trans. Power Systems. November 2000; vol. 15.
9. Ngatchou P, Zarei A, El-Sharkawi A. Pareto multi objective optimization. In: Proc. of the 13th International Conf. Intelligent Systems Application to Power Systems. 6–10 November 2005: 84–91.
10. Berizzi A, Innorta M, Marannino P. Multiobjective optimization techniques applied to modern power systems. In: February, 2001. IEEE Power Engineering Society Winter Meeting; vol. 3.
11. Coello CA, Lechuga MS. A proposal for multiple objective particle swarm optimization. In: IEEE Proc. World Congress on Computational Intelligence. 2003: 1051–1056.
12. Fieldsend JE, Singh S. A multi-objective algorithm based upon particle swarm optimisation, an efficient data structure and turbulence. In: U.K. Workshop on Computational Intelligence. 2002: 34–44.
13. Hu X, Eberhart R, Shi Y. Particle swarm with extended memory for multiobjective optimization. In: IEEE Swarm Intelligence Symposium. 2003: 193–198.
14. Mostaghim S, Teich J. Strategies for finding good local guides in multi-objective particle swarm optimization. In: IEEE Swarm Intelligence Symposium. 2003: 26–33.
15. Deb K, Paratap A, Agarwal S, Meyarivan T. A fast and elitist multi-objective genetic algorithm: NSGA-II. IEEE Trans. Evolutionary Computation. 2002; vol. 6 [pp. 182–197].
16. Srinivas N, Deb K. Multiobjective optimization using nondominated sorting in genetic algorithms. Kanput, India: Technical Report, Dept. Mechanical Engineering; 1993.
17. del Valle Y, Venayagamoorthy GK, Mohagheghi S, Hernandez J, Harley RG. Particle swarm optimization: basic concepts, variants and applications in power systems. IEEE Trans. Evolutionary Computation. April 2008; vol. 12(no. 2):171–195.
18. Hamid RHA, Amin AMA, Ahmed RS, Elgammal AAA. Optimal operation of induction motors based on multi-objective particle swarm optimization (MOPSO). In: The 33rd Annual Conf. IEEE Industrial Electronics Society. 5–8 November 2007: 1079–1084.
19. Thanga Raj C, Srivastava SP, Agarwal P. Energy efficient control of three-phase induction motor - A review. International Journal of Computer and Electrical Engineering. April 2009; vol. 1(no. 1):61–70.
20. Sharaf AM, El-Gammal AAA. A novel self regulating hybrid pv-fc-diesel-battery electric vehicle drive system. International Journal of Electric and Hybrid Vehicles. 2010; vol. 3(no. 1):1–43.
21. Billinton R, Gao Y. Multistate wind energy conversion system models for adequacy assessment of generating systems incorporating wind energy. IEEE Trans Energy Conversion. March 2008; vol. 23(no. 1):163–170.
22. Hu W, Wang Y, Yao W, Wu J, Zhang H, Wang Z. An efficient experimental method for high power direct drive wind energy conversion systems. In: IEEE Power Electronics Specialists Conf.. 15–19 June 2008: 3955–3959.
23. Pera MC, Hissel D, Kauffmann JM. Fuel cell systems for electrical vehicles. In: 2002: 2097–2102. IEEE 55th Vehicular Technology Conf. VTC Spring; vol. 4.
24. N. Parspour, “Novel drive for use in electrical vehicles”, in IEEE, 1st Vehicular Technology Conf. VTC Spring 2005, vol. 5, pp. 2930–2933.
25. Sharaf AM, Ozkop E, Altas IH. A hybrid photovoltaic PV array-battery powered EV-PMDC drive scheme. In: IEEE Electrical Power Conf.. 2007: 37–43.
26. Zhu ZQ, Chan CC. Electrical machine topologies and technologies for electric, hybrid, and fuel cell vehicles. IEEE Vehicle Power and Propulsion Conf. 2008; 1–6.
27. Vas JV, Venugopal S, Nair VG. Control scheme for electrical drive of solar powered vehicles. In: 2008: 75–80. Annual IEEE India Conf.; vol. 1.
28. Zhu ZQ, Howe D. Electrical machines and drives for electric, hybrid, and fuel cell vehicles. In: 2007: 746–765. Proc IEEE; vol. 95 [no. 4].
29. Luque A, Hegedus S. Handbook of Photovoltaic Science and Engineering. New Jersey, United States: John Wiley and Sons; 2003.
30. Nelson J. The Physics of Solar Cells. London, UK: Imperial College Press; 2003.
31. Thounthong P, Rael S, Davat B. Analysis of supercapacitor as second source based on fuel cell power generation. IEEE Trans. Energy Conversion. March 2009; vol. 24(no. 1):247–255.
32. Gebregergis A, Pillay P, Bhattacharyya D, Rengaswemy R. Solid oxide fuel cell modeling. IEEE Trans. Industrial Electronics. January 2009; vol. 56(no. 1):139–148.
33. Altas IH, Oz kop E, Sharaf AM. A novel PV-powered standalone village electricity utilization fuzzy logic dynamic controller strategy. In: International Symposium on Innovations in Intelligent Systems and Applications. 29 June–1 July, 2009: 234–240.
34. Fujinami K, Takahashi K, Kondo K, Sato Y. A restarting method of an induction motor speed-sensorless vector control system for a small-sized wind turbine power generator system. In: International Conf. Electrical Machines and Systems. 2009: 1–5.
35. Chen W, Hsu Y. Controller design for an induction generator driven by a variable-speed wind turbine. IEEE Trans. Energy Conversion. 2006; vol. 21(no. 3):625–635.
36. Zahedi B, Vaez-Zadeh S. Efficiency optimization control of single-phase induction motor drives. IEEE Trans. Power Electronics. 2009; vol. 24(no. 4):1062–1070.
37. Mademlis C, Kioskeridis I, Theodoulidis T. Optimization of single-phase induction Motors-part I: Maximum energy efficiency control. IEEE Trans. Energy Conversion. 2005; vol. 20(no. 1):187–195.
38. de Rossiter Correa MB, Jacobina CB, da Silva ERC, Lima AMN. Vector control strategies for single-phase induction motor drive systems. IEEE Trans. Industrial Electronics. 2004; vol. 51(no. 5):1073–1080.
39. Sharaf AM, Ciaxia C, Huang H. A smart PWM-modulated power filter for single phase motorized loads. In: 1998: 778–781. IEEE Canadian Conf. Electrical and Computer Engineering; vol. 2.
40. Sharaf AM, Kreidi P. Dynamic compensation using switched/modulated power filters. In: 2002. IEEE Canadian Conf. Electrical and Computer Engineering; vol. 1 [230–235].
41. Sharaf AM, Chalet R. A low cost on-off modulated power filter for single phase motorized loads. In: 1998: 862–865. IEEE Canadian Conf. Electrical and Computer Engineering; vol. 2.
42. Sharaf AM, Aljankawey A. Voltage stabilization using a facts modulated power filter. In: 2006: 1937–1942. IEEE International Symposium on Industrial Electronics; vol. 3.
43. Yang L, Yang GY, Xu Z, Dong ZY, Wong KP, Ma X. Optimal controller design of a doubly-fed induction generator wind turbine system for small signal stability enhancement. Generation, Transmission and Distribution, IET. 2010; vol. 4(no. 5):579–597.