Nikolai Voropai; Dmitry Efimov; Irina Kolosok; Victor Kurbatsky; Anna Glazunova; Elena Korkina; Nikita Tomin; Daniil Panasetsky Energy Systems Institute, Irkutsk, Russia
The chapter presents the following contributions: (1) a general overview of an intelligent electric power system with an active and adaptive network (IESAAN) as the Russian vision of a smart grid, its technological platform, and control system; (2) state estimation (SE) techniques as informational support of the IESAAN control including SE with phasor measurements use, FACTS modeling in SE applications, dynamic SE, and cyber-physical security issues of SE; (3) intelligent operation and smart emergency protection in Russia including requirements for new protection systems; a novel system of monitoring, forecasting, and control of electric power system (EPS); artificial intelligence applications in EPS such as a forecast of state variables based on dynamic SE and hybrid data-driven approaches; a total transfer capability estimation method; an automatic decision tree-based system for online voltage security control; a multiagent coordination of emergency control devices; and an intelligent system for preventing large-scale emergencies; (4) a description of smart grid territorial clusters in the interconnected power systems of Russia.
Electric power system; Emergency control; Smart grid; Unified energy system of Russia; Intelligent electric power system; Active and adaptive network; Wide area measurement and wide area control systems; State estimation; Cyber-physical security; Multiagent automation system
Acronyms
AAN active and adaptive electric network
ANN artificial neural network
APEC Asia Pacific Economic Cooperation
ARIMA autoregressive integrated moving average
BDD bad data detection
BESS battery energy storage systems
CACS conventional automatic control system
CHPP combined heat and power plant
DC direct current
DSE dynamic state estimation
ECS emergency control system
EMD empirical mode decomposition
EMS energy management system
ENTSO-E European Network of Transmission System Operators for Electricity
EPS electric power system
ESS electricity supply system
FACTS flexible alternating current transmission systems
FGC UES JSC “Federal Grid Company of Unified Energy System”
GARCH generalized autoregressive conditional heteroscedasticity
GLONASS Global Navigation Satellite System in Russia
GPS global positioning system
HHT two-step algorithm, combining EMD and HT
HPP hydro power plant
HT Hilbert transform
HTSC high-temperature superconductivity
HVDC high-voltage direct current
ICOEUR intelligent coordination of operation and emergency control of European Union and Russian power grids, the international European/Russian FP7 project
IED intelligent electronic device
IEEE Institute of Electrical and Electronics Engineers
IESAAN intelligent electric power system with active and adaptive network
IMF intrinsic mode functions
IPS interconnected power system
MAAS multiagent automation system
MAE mean absolute error
MAPE mean absolute percent error
MLP multilayer perceptron
OLTC on load tap changer
OTL overhead transmission line
PDC phasor data concentrator
PDSRF algorithm of proximity-driven streaming random forest
PM pseudomeasurement
PMU phasor measurement unit
RMSE root mean square error
RTU remote terminal unit
SC series capacitor
SCADA supervisory control and data acquisition
SE state estimation
SO UES JSC “System Operator of Unified Energy System”
SS substation
STATCOM static synchronous compensator
STC static thyristor compensator
SVC static var. compensator
SVM support vector machine
TCR thyristor-controlled reactor
TCSC thyristor-controlled series capacitor
TE test equation
TPP thermal power plant
TSC thyristor switched capacitor
TTC total transfer capability
UES unified energy system of Russia
UNEG unified national electric grid
UVLS under voltage load shedding
VSC voltage source converter
WACS wide area control system
WAMS wide area measurement system
WAPS wide area protection system
The unified energy system (UES) of Russia is a power interconnection where seven interconnected power systems (IPSs) are combined by weak ties. Under emergency conditions, the Russian UES is able to disintegrate into autonomously operating self-balanced IPSs without grave consequences.
The trends in expansion of electric power systems (EPSs) and changes in the conditions of their operation have complicated EPS operations as well as increased its changeability and unpredictability that call for a more prompt and more adequate response of control systems.
A totally novel approach was suggested to make the transition to a new level of technology and control of the Russian UES. The main goal of this approach is the implementation of intelligent technologies in the Russian power industry to ensure an innovative breakthrough in the development of that industry and to increase the efficiency, reliability, and security of its operation.
The chapter presents the following contributions: intelligent energy system as a Russian vision of a smart grid; informational support of an active and adaptive network (IESAAN) control problems; intelligent operation and smart emergency protection; smart grid clusters in Russia.
In 2010–12, the concept of an intelligent EPS with IESAAN was developed in Russia. The concept stipulates that all subjects of the electricity market (generation, grid, and consumers) take an active part in the processes of electric power transmission and distribution.
The problem of efficient control of IESААN is among the most important scientific and technical problems. With the availability of an effective control system, IESААN can provide reliable interaction between consumer grid units of different functions and generators, using uniform principles and the common information-technological platform.
The section presents:
State estimation (SE) is the most important procedure that provides EPS control with reliable and quality information. The result of SE is a model of an EPS steady state based on measured parameters of the system and data on the state of network components. IESAAN involves new technologies and systems for measurement of state parameters which form new conditions for SE of EPS, and put forward new objectives to be accomplished on the way of modernization of the data computing system.
The section discusses the following issues:
This section discusses intelligent operation and smart emergency protection in Russia, including a description of an emergency control system (ECS), requirements to new operation and emergency protection, and a novel system of monitoring, forecasting, and control of EPS while also presenting several recent applications of artificial intelligence in EPS:
The process of IESAAN formation suggests the implementation of pilot projects and the creation of territorial smart grid clusters. These clusters are supposed to use information, technology, and control systems providing adaptive control of network parameters, remote control of switching devices, and real-time estimation of the technical state of the network. Currently, along with implementation of the pilot projects and creation of smart grid clusters, new equipment is being installed at the energy facilities of the unified national electric grid (UNEG) of Russia. This section describes several examples of smart grid clusters and pilot projects in Russia's EPSs:
A totally novel approach was suggested to make the transition to a qualitatively new level of technology and control of the UES of Russia. The key role in the approach belongs to the core of EPSs, that is, the electrical network with active-adaptive properties that provides reliable and efficient connection between generation and consumers. The main goal of this approach is the implementation of intelligent technology in the Russian power industry to ensure an innovative breakthrough in the development of the industry, and to increase the efficiency, reliability, and security of its operation.
In 2010–12, the concept of an intelligent EPS with an active and adaptive network (IESAAN) [1] was developed by the JSC R&D Centre for Power Engineering. The concept of IESААN stipulates that all subjects of the electricity market (generation, grid, consumers) take an active part in the processes of electric power transmission and distribution. Electric power consumers as a part of IESAAN should become its leading component. Transmission and distribution networks, which traditionally were passive, will turn into active components whose parameters and characteristics will become flexible according to the operation requirements of the entire system.
The decision on development of concepts of technological platforms for Russia was made in August 2010 by the Governmental Commission for High Technologies and Innovations [2]. A technological platform is a form of public-private partnership in the field of innovation, a way of uniting the efforts of all parties concerned (various departments, businesses, and the scientific community) to accomplish the final goals of certain strategic priority areas.
In 2010 the JSC Federal Grid Company of Unified Energy System (FGC UES) and Federal State Institution Russian Energy Agency supported the development of an application for creation of the technological platform, “Intelligent energy system of Russia,” and its submission to the Ministry of Economic Development of Russia (Fig. 1). The main mission of the technological platform is implementation of intelligent technologies in the Russian power industry to ensure an innovative breakthrough in development of the industry, and to increase the efficiency, reliability, and security of its operation [3].
An intelligent EPS (Fig. 2) is a customer-oriented EPS of a new generation that should provide high-quality, reliable, and efficient services for electricity supply through flexible interaction in all types of generation, electric networks, and consumers based on cutting-edge technologies and a common hierarchical system of control. An important role in the intelligent EPS is assigned to the active and adaptive electric network (AAN). Radically new properties imparted to the power system by ANN are [4,5]:
The UNEG carries out backbone functions and includes power segment grids, electricity supply systems (ESSs), and the bulk transmission and international grids.
The goals of an intelligent EPS [6,7] are presented in Table 1.
Table 1
The following technologies are used to accomplish the above goals [4,5,8,9]:
At the formation of IESААN, the UNEG components of different voltage levels should contain the devices changing the impedance of the grid elements and voltage (both magnitude and phase) at various points of the grid. Possibilities of combining alternating and direct (lines and back-to-back stations) currents as well as modern devices of short circuit current limitation in powerful switching equipment will be widely used.
The ESS structure can include sources of electric power, electric power transmission, distribution, transformation devices, and various auxiliary devices and constructions (power plants and feeding lines of regional power systems, transmission lines, substations, and distribution devices). One of the smart ESS's key functional characteristics is the motivation of active behavior of the end consumer. Active behavior implies consumer possibility to change independently the received electricity amount and functional properties (level of reliability, quality, etc.), using the information about prices, electricity supply volumes, reliability, quality, etc. The aim of the motivation is steering those changes according to the balance of the requirements and possibilities of ESS.
The problem of efficient control of IESААN, whose urgency and importance have been confirmed by recent large, human-caused failures, is among the most important scientific and technical problems. With the availability of an effective control system, IESААN can provide reliable interaction between consumer grid units of different functions and generators, using uniform principles and the common information-technological platform [6].
The IESААN control system should provide the solution for the following problems:
For the effective solution of IESААN control problems online in the conditions of incomplete information on parameters of EPS and disturbances, the use of uniform principles of control and qualitatively new kinds of techniques and technologies is necessary, including means and systems for:
Thus the structure of IESААN control (Fig. 3) should be formed on the basis of the following principles:
Large system blackouts that have occurred in different countries in the beginning of the new millennium prove the need for updating and improving of software for EPS monitoring and control (SCADA/EMS-applications). It supposes the use of
Creation of satellite communication systems provided a new generation of measurement equipment—PMUs [10]. Integrated into the WAMS, the PMU sensors give a real picture of the EPS state. WAMS technology is intended for the PMU-based calculation of angles between voltage and current vectors synchronized with accuracy up to 1 μs. Each PMU installed at the ith node provides direct measurements of voltage magnitude Ui and phase angle δi at buses, magnitude of current Iij along the outgoing line, and angle φij between current and voltage as well as the calculated flows of active Pij and reactive Qij power in a line. GPS-synchronized equipment is capable of measuring voltage magnitude with an accuracy of 0.1% and a phase angle with accuracy of 0.2°.
The updating of software for EPS monitoring and control (SCADA/EMS-applications) has become possible on a qualitatively new level owing to (WAMS) allowing the EPS state to be controlled synchronously and with high accuracy. The PMUs are the main measurement equipment in these systems. Since 2005, Russia has been creating a system for monitoring of transients (a Russian analog of WAMS). The main measurement equipment in the system is SMART-WAMS recorders of voltage and current phasors. One of the applications, which will be significantly affected by the introduction of PMU, is the SE.
Until recently, SE in EPS was mainly based on the SCADA measurements: magnitudes of bus voltages, generation of active and reactive powers at buses, power flows in lines, and, less often, currents in lines and at buses.
SCADA systems are intended for receiving and processing information once per second, and the remote control systems themselves allow delays in information delivery up to several tens of seconds. Another serious flaw of existing remote control systems is the absence of a highly accurate synchronization of measurements with respect to astronomical time. The lack of simultaneity in receiving measurements is particularly noticeable in the calculation of simultaneously operating subsystems that have their own devices for collection and transfer of information. The state variables calculated on the basis of SCADA measurements are behind the current state variables of EPS and are only their approximation, which can cause errors in control.
The advent of WAMS that contains PMUs as the main measurement equipment makes it possible to synchronously and accurately control the EPS state and essentially improve the results of SE [11]. PMUs are employed in EPS both to solve the local problems and to obtain a general picture of the EPS state, which is further used for solution of control problems. As compared to a standard set of measurements received from SCADA, PMU installed in the node can measure voltage phasor in this node and current phasors in some or all lines adjacent to this node with high accuracy. The use of PMU measurements improves the observability of the calculated network, enhances the efficiency of methods for BDD, increases the accuracy and reliability of the obtained estimates, and offers new possibilities in decomposition of the SE problem.
While SCADA measurements provide observability of the network, the critical measurements and critical sets exist, that is, the redundancy of measurements is quite low. To increase the redundancy of measurements and eliminate critical measurements and critical sets, PMU shall be added to the network buses.
WAMS is a set of recorders of synchronized PMUs, phasor data concentrators (PDCs), channels for data transfer among the recorders, data concentrators, and dispatching centers of the JSC “SO UES” as well as systems for processing the obtained information. WAMS measurements are synchronized by the systems of GPS/GLONASS. The hierarchical architecture of WAMS [12] is presented in Fig. 4.
PMUs measure the magnitudes and phases of nodal voltages and currents in lines, which are incidental to these nodes. The PDC collects, filters, processes, and retransmits the data. Besides, the PDC can register abrupt surges, distortions, parameters of switches, and parameters of loads and lines as well as identify generator parameters.
A super-PDC processes data and provides a dispatcher or an operator with a graphical interface and access to the data archive.
There are different approaches to the determination of sites for PMU placement in the system. In Russia's UES that consists of seven IPSs furnished to a different extent with SCADA measurements, the sites for priority placement of PMU for monitoring of transients were determined on the basis of practical engineering experience [13]. These are:
However, even the existing insignificant number of PMUs in combination with SCADA measurements make it possible to essentially improve the results obtained by solving the SE problem. Currently, there are algorithms for solving the problem of SE using only PMU or SCADA measurements, or a combination of both.
If there is a sufficient amount of PMUs to provide observability of the EPS scheme, SE can be performed using PMU data only.
Placement of PMU at a node of EPS provides a set of new measurements that are characterized by high accuracy. Inclusion of PMU measurements in the total set of measurements in the system increases the redundancy of measurements, which, in turn, helps in detecting bad data in the measurements and improves the quality of SE. The main types of measurements received from PMU are magnitudes Ui and phases δi of nodal voltages and currents Iij, ϕij in the outgoing lines.
Based on the vectors of nodal voltages and currents in the outgoing lines, it is possible to calculate the magnitudes and phases of nodal voltages at neighboring nodes, that is, to obtain the so-called calculated PMUs:
Ucalcj=√U2i−2UiIij(rijcosϕi+xijsinϕi)+I2ij(r2ij+x2ij)
δcalcj=δi−arctgIij(xijcosϕij−rijsinϕij)Ui−Iij(rijcosϕij+xijsinϕij)
where rij, xij is the resistance and reactance of line i − j.
Also, PMU measurements are used to calculate pseudomeasurements (PMs) of power flows. For example, PMs of power flows at the beginning of line PijPM can be calculated as follows:
PPMij=√3IijPMUUiPMUcosϕijPMU
QPMij=√3IijPMUUiPMUsinϕijPMU
Measurements from physical and calculated PMUs and flow PMs are the initial information for the SE problem.
Accuracy of flow PM owing to high accuracy of PMU measurements is considerably higher than accuracy of SCADA measurements. PMU measurements can be used to check the quality of SCADA measurements by the TE technique.
At the end of 2015, about 400 SMART-WAMS recorders were installed in the eastern synchronous zone of the UES of Russia. A priority set for creation of the smart grid in Russia is the intensive development of WAMS and WACS systems.
To obtain a general picture of the EPS state for further solving the control problems, the universal methods are necessary to place PMU to complement SCADA measurements. These methods should provide the best properties of the SE problem solution, such as observability of the calculated scheme, identifiability of bad data, and accuracy of obtained estimates. These methods should be based on the observability theory that was devised to place SCADA systems [14] and take into account the block character of PMU measurements. Different methods were offered to solve this problem. In the authors’ opinion, the most promising are topological approaches based on different strategies of random search [15–17].
SE is the most important procedure that provides EPS control with reliable and quality information. The result of SE is a model of the EPS steady state. The model is based on measured parameters of the system and data on the state of network components.
The SE problem consists of calculating the EPS steady-state conditions by measurements. The measurements applied in the EPS SE are mostly the measurements obtained from SCADA system and PMUs: Ui—magnitudes of nodal voltages; Pgi, Qgi—generation of active and reactive powers at nodes; Pni, Qni—loads of active and reactive powers at nodes; Pij, Qij—power flows in transformers and lines; δi—voltage phases at the nodes of the scheme in which PMUs are placed; Iij—magnitudes of currents in the lines incident to these nodes; and ϕij—an angle between current and voltage.
The vector of measurements looks as follows:
ˉy=(Pi,Qi,Pij,Qij,Ui,δi,Iij,ϕij)
where Pi = Pni + Pgi, Qi = Qni + Qgi.
The measurement model is as follows.
ˉyi=yss(i)+ξy(i)
yss(i)—true value, ξy(i)—normally distributed random error ξi(y) ∈ N(0, σi(y)2), where σi(y)2—a variance of the ith measurement.
The traditional mathematical statement of the SE problem consists in minimization of the objective function:
J(y)=(ˉy−ˆy)TR−1y(ˉy−ˆy)
subject to constraints in the form of steady state equations:
w(y,z)=0
where Ry is a diagonal matrix, the elements of which are equal to measurement variances.
The state vector x is the (2n − 1)-dimensional vector (where n is the number of nodes in the calculated scheme). The vector includes the magnitudes U and phase angles δ of voltages х = (U, δ). Such a state vector determines all the other state parameters.
By using dependences y = y(x) as Eq. (6), the SE problem is reduced to a search for the values (estimates) of components of the state vector ˆx, such that the values y(ˆx) are maximum close to the measurements in a sense of the criterion:
J(x)=(ˉy−y(x))TR−1(ˉy−y(x))
This method is presented in detail in Refs. [18–20].
The SE method based on the use of TEs was developed in Refs. [18,21]. TEs are steady-state equations that contain only measured state variables y
wte(y)=0
which can be obtained from the system of steady-state equations (6) by elimination of unmeasured variables. By substituting the values of measurements in these equations, one can judge the presence of gross errors in measurements from the value of discrepancies obtained; therefore they are called TEs.
When TEs are used, the SE problem consists of the minimization of the criterion (5), that is, the calculation of estimates of measured variables ⌢y under the constraints represented by a system of TEs (8). After that, the estimates of basic measurements are used to calculate the state vector estimates, and these are used to calculate the estimates of unmeasured variables.
The TEs can be used to solve many problems related to the real-time SE: analysis of network topology, BDD, SE procedure, and identification of measurements variances.
The existing methods and approaches that are applied today in SE were developed in the conditions of insufficient telemetry data; low redundancy of measurements and incomplete visibility of the network; asynchronous measurements and considerable delays in their transfer; and low accuracy and reliability of analog measurements and remote signals. These conditions determined the main direction in the evolution of the theory and practice of EPS SE.
The creation of IESAAN suggests a wide use of
Change in the properties of an observed EPS itself due to saturation of the network with active components (FACTS, HVDC) forms new conditions for SE of EPS and puts forward new objectives to be accomplished in the modernization of the data computing system of the IESAAN.
The issues of SCADA and PMU data integration are algorithmically resolved as follows: measurements of magnitudes and phases of currents are considered as additional components of the measurement vector ˉy, and measurements of voltage magnitudes and phases can be emphasized especially because they are simultaneously components of both measurement vector and state vector x:
ˉxδi=xδi+ξδ,ˉxUi=xUi+ξU
where ξδ, ξU—errors in measurements δ, U, that have normal distribution with zero mean, Rδ, RU—diagonal covariance matrices with the measurement error variances of vector components x on their diagonal.
Problem statement: By using the given ˉy,Ry,ˉδ,Rδ,ˉU,RU find estimates ⌢δ,⌢U by minimizing the objective function
minδ,U[(ˉy−y(x))TR−1y(ˉy−y(x))+(ˉδ−δ)TR−1δ(ˉδ−δ)+(ˉU−U)TR−1U(ˉU−U)]
Expressions for the calculation of estimates are obtained when derivatives with respect to δ and U equal zero:
ˆδ=(∂y∂δ)TR−1y(y(δ)−ˉy)+R−1δ(ˉδ−δ)andˆU=(∂y∂u)TR−1y(y(U)−ˉy)+R−1u(ˉU−U)
Irrespective of the statement and method of solving the SE problem, the accuracy of the obtained estimates is determined by covariance matrices of erroneous estimates of the state vector ⌢δ,ˆU:
Pδ=[(∂y∂δ)TR−1y∂y∂δ]−1
Pu=[(∂y∂U)TR−1y∂y∂U]−1
When adding the measurements of magnitudes and phases of bus voltages obtained from PMU, the expressions (12), (13) are transformed into the form:
Pδ=[(∂y∂δ)TR−1y∂y∂δ+R−1δ]−1
Pu=[(∂y∂U)TR−1y∂y∂U+R−1u]−1
Criterion
maxdet[Pδ,Pu]
where Pδ, PU are determined by Eqs. (14), (15), can be used as a criterion for choosing the set of PMU placement from the viewpoint of maximum accuracy of estimates [17].
Results of the studies on the combined use of SCADA and PMU measurements, including those obtained by [11], indicate that in this case the main flaws characteristic of traditional SE still exist.
If the amount of PMUs is sufficient to provide the required observability of the EPS network, then SE can be made only on the basis of PMU data.
The TEs technique is used for a priori validation [23]. With substitution of measurements in the TE, the error in the measurements affects the value of the TE discrepancy. The presence of error can be detected by checking the condition
∣wk∣<d
The value of threshold for the jth TE dj is determined as
dj=γα√∑i∈ωjaijσ2i
where ωj a set of measurements entering the jth TE, σi2—variances of measurements, γα—inverse distribution N(0, 1), determined by the given probability of bad data penetration in the SE problem α; aij—coefficients of TE. If this (17) is not met, all the measurements in this TE should be analyzed. In this case, if all measurements except one are valid, the measurement with a gross error is replaced by a PM. If one TE contains several erroneous measurements, the additional information is required.
The algorithms of SE on the basis of the TE technique are less labor intensive and are very fast because the order of the TE set is, as a rule, considerably lower than the order of the initial set of steady-state equations that is used to obtain TEs.
The TE method was developed to detect bad data in SCADA measurements [23], then adapted to analyze the validity of the PMU measurements [18].
There are some reasons for the failures in PMU operation and the appearance of bad data in their measurements: low accuracy of instrument transformers at PMU inlets; difference in the error ranges in PMUs of different manufacturers, which manifests itself under large disturbances; failures in the system of data acquisition and transfer from PMU to the upper level; as well as the human factor (e.g., erroneous phase connection). Thus, the necessity arises to verify synchronized vector measurements coming from PMUs.
As is shown in Ref. [24], if calculations are made in rectangular coordinates of the state vector and the magnitudes and phase angles of currents are used as measurements, the measurement model becomes linear and the state vector estimates ˆx can be obtained from the equation
ˆx=[HTR−1yH]−1HTR−1yˉy
at the constant Jacobian matrix H=∂y∂x. State vector of the system x is defined by x={˙Ui}, where ˙Ui=Uai+jUri—complex numbers of nodal voltages, vector of measurements ˉy is also represented by the complex number vector ˉy={˙Ui,˙Iij,˙Ii}.
The TEs are based on the exclusion of components of state vector x from the equations describing the relationships between measured variables y and state vector x in rectangular coordinates:
y−H(x)=0
Each measurement of the electrical current complex number in a line will correspond to two equations from (20):
ˉIija−(Uai−Uaj)yaij+(Uri−Urj)yrij−Uaiyaiij+Uriyriij=0
ˉIijr−(Uai−Uaj)yrij−(Uri−Urj)yaij−Uaiyriij−Uriyaiij=0
where yaij, yrij—series conductance and susceptance, and yaiij, yriij—real and imaginary part of shunt admittance of line i − j.
Each voltage phasor measurement at node i will correspond to two equations:
ˉUai−Uia=0
ˉUri−Uri=0
The problems of PMU-based linear SE and BDD are addressed in a great number of studies. An in-depth explanation of the linear SE procedure is presented in Ref. [25], where the authors suggest a three-phase SE. To speed up the computation process, the author divides all the complex numbers into real and imaginary components and transforms the matrices describing the state of the system. In Ref. [26] the authors show that before the procedure of linear SE starts, it is necessary to make sure that the considered object of estimation (substation, region, intersystem tie line) can be reliably monitored by PMU and the PMU data can be verified. Also, the authors of Ref. [26] note that (a) no traditional measurements including zero injections should be used for linear SE, and (b) no PMU can be critical.
The problem of SE using the TEs is solved in two stages [27].
The first stage suggests a search for the estimates (20) of measured state variables entering the TE through the minimization of criterion (9) under the constraints represented by the system of TEs (12).
ˆy=ˉy−R(∂wk∂y)T[(∂wk∂y)R(∂wk∂y)T]−1wk(ˉy)
Because the basic set of measurements yb that enables us to determine x = f(yb) has already been chosen automatically by the software in the process of constructing TEs, it is only necessary at the second stage to determine the estimates of state vector ˆx on the basis of the estimates of basic measurements ˆyb that were obtained from the equation:
Hbˆx=−ˆyb
and calculate the unmeasured variables.
The systematic errors caused by the errors of the instrument transformers that exceed the class of their accuracy are constantly present in the measurements and can be identified by considering some successive snapshots of measurements. The TE linearized at the point of a true measurement, taking into account random and systematic errors, can be written as:
wk(ˉy)=∑l∈ωk∂w∂yl(ξyl+cyl)=∑aklξyl+∑aklcyl
where ∑ aklξyl—mathematical expectation of random errors of the TE, equal to zero; ∑ aklcyl—mathematical expectation of systematic error of the TE, ωk—a set of measurements contained in the kth TE.
The author of Ref. [28] suggests an algorithm for the identification of a systematic component of the measurement error on the basis of the current discrepancy of the TE. The algorithm rests on the fact that systematic errors of measurements do not change through a long time interval. In this case, condition (17) will not be met during such an interval of time. Based on the snapshots that arrive at time instants 0, 1, 2, …, t − 1, t…, the sliding average method is used to calculate the mathematical expectation of the TE discrepancy:
Δwk(t)=(1−α)Δwk(t−1)+αwk(t)
where 0 ≤ α ≤ 1.
Fig. 5 shows the curve of the TE discrepancy (a thin dotted line) calculated by (26) for 100 snapshots of measurements that do not have systematic errors.
It virtually does not exceed the threshold dk = 0.014 (a light horizontal line). Above the threshold, there is a curve of the TE discrepancy (a bold dotted line) that contains a measurement with a systematic error and a curve of nonzero mathematical expectation Δwk(t) ∈ [0.026; 0.03] (a black-blue thick line). However, the nonzero value of the calculated mathematical expectation of the TE discrepancy can only testify to the presence of a systematic error in the PMU measurements contained in this TE, but cannot be used to locate it.
Currently the decomposition problem of a large-dimension EPS is very important. The effective method of solution is the distributed processing of the information on the basis of decomposition. This approach is offered in articles of the Russian [29,30] and foreign scientists [31–34], etc. Distributed SE includes the following main procedures:
As boundary conditions at decomposition with boundary nodes, the equality of voltage modules U and angle phases δ of boundary nodes should be observed:
Ui=Uj=⋯=Uk
δi=δj=⋯=δk
Also the relationships of boundary balance should be observed. For example, for a boundary node l, which is general for i, j, …, kth subsystems
Ploadl−Pgenl+∑s=i,j,…,k∑m∈ωsPlm(U,lδl,Um,δm)=0
Qloadl−Qgenl+∑s=i,j,…,k∑m∈ωsQlm(U,lδl,Um,δm)=0
where Ploadl, Qloadl, Pgenl, Qgenl—active and reactive powers of loads and generations in a node l, ωs—set of nodes of sth subsystem, incident to lth node.
As boundary conditions at decomposition by boundary lines, the relationships for active Pij and reactive Qij power flows from the ith subsystem to the jth subsystem should be observed:
Pij=−Pji+ΔPij
Qij=−Qji+ΔQij
For boundary line
U2m−(Ul−Pmlrml+QmlxmlUm)2−(Pmlxml−QmlrmlUm)2=0
δm−δl−arctgPmlxml−QmlrmlPmlrml+Qmlxml=0
Installation of PMU at boundary nodes allows registering boundary variables U and δ on the values metered with high accuracy. In this case:
For coordination of phase angles of the voltages received from local SE, only one PMU measurement of the phase angle is enough in each subsystem. Such a node is appointed for subsystems as a reference node. PMU measurements coordinate the results of the SE of separate subsystems. Therefore, it is necessary to install the device PMU in a reference node or to appoint the node with PMU as a reference. Thus SE of separate subsystems can be calculated in parallel, independently from each other; accomplishment of iterative calculations on subsystems is not required.
Successful (reliable, quality, and economical) operation of IESAAN requires a wide range of advanced technical tools and technologies that afford the opportunity to endow the network with active-adaptive qualities.
There is a persistent trend in the international and Russian practices toward adopting the power electronics controlled devices in the electrical networks of EPSs, that is, flexible alternating current transmission systems (FACTS). Today, FACTS is among the most promising electrical network technologies [1,35,36], which makes it possible for the electrical network to turn from a passive facility for electricity transportation into the facility that takes an active part in the control of the network operation. The FACTS devices can be used to increase the transfer capability, improve static and dynamic stability, and provide better power quality. This technology provides control of interrelated parameters, including impedances, currents, voltages, phase angles, oscillation damping at different frequencies, etc., and opens up new opportunities to control EPSs.
Until recently, the models for the present-day FACTS devices based on power electronics were not involved in the design diagram when solving the SE problem. Developing mathematical EPS models for SE accounting for new devices is a topical problem nowadays. Such investigations and developments have been actively conducted in recent years [37–39]. A team of researchers at the Energy Systems Institute has been investigating modeling FACTS devices when solving the SE problem. By now, the models for the main FACTS devices and the algorithms to involve these models in the SE problem have been developed [40–43].
Because the parameters of equivalent circuits of FACTS devices change depending on EPS operating conditions, the problem of determining the parameters of equivalent circuits of these devices on the basis of PMU data is also topical. Therefore, developing the FACTS device models, identifying the parameters of these models, and involving them in the algorithms of the present-day EPS SE is a topical problem when creating the system of SG control.
Such an approach implementation when modeling TCSC and SVC at the EPS SE is considered below, together with a brief description of the models for these devices.
Thyristor-controlled series capacitors (TCSCs). Series capacitors (SCs) are the capacitor batteries connected in a series with a power line to compensate a part of a series inductive reactance. In different countries, SCs are widely applied in the regions where energy sources are far from customers, such as in Sweden. One of the first-ever SCs, the Tyret’ SC, was installed on the lengthy 500 kV Bratsk-Irkutsk high-voltage line (~ 700 km) within the Irkutsk power system. The Tyret’ SC is a group of capacitor batteries/banks, or bridges. Controlling the state by reactive power occurs due to switching the bridges on and off. Switching the bridges on and off is quite problematic to perform in the process tempo. The alternative option is using TCSCs, in which a part of the capacitor bank is shunted by a thyristor controller that enables it to smoothly vary its equivalent capacitance depending on the line operation state.
TCSCs, in fact, is a standard SC but supplemented with a thyristor control block.
Fig. 6 presents the power transmission scheme with a TCSC.
The TCSC impedance comprises the impedance of the parallel-connected bank and reactor [40], and depends on the thyristor firing angle α:
xTCSC(α)=xTCR(α)⋅xCxTCR(α)+xC
where
xTCR(α)=ωL⋅ππ−2α−sin2α
At SE, TCSCs are set by the model with variable conductance, xTCSC(α), or by the model with a variable firing angle α. An algorithm implementing the former model is developed. In this algorithm, TCSC is modeled by the i − j line with a variable reactance (yr = 1/xTCSC conductance) and by the active-power fixed overflow equal to the SCADA measurements. The nods restricting the i − j line are transit (have zero injections). The set of equations, in the designations typical of SE [21] for the i − j line shown in Fig. 6, looks like:
Pi−j=UiUjsin(δi−δj)yr
Qi−j=U2iyr−UiUjcos(δi−δj)yr
The yr susceptance is set as a state vector component, and is determined directly during the SE problem solution. The Jacobian matrix column corresponding to this state vector component contains derivatives of nodal injections at the nodes i and j, calculated through the formulas:
∂Pi∂yr=∂Pi−j∂yr=UiUjsin(δi−δj)
∂Qi∂yr=U2i−UiUjcos(δi−δj)
Similarly, one can calculate the derivatives of Pj, Qj.
By using the Jacobian matrix formed in this way, the SE can be performed, during which the yr estimation, and later the xTCSC estimation are calculated. The xTCR is determined from the xTCSC using (35). The thyristor firing angle α is calculated iteratively from (36), represented in the form:
f(αi)=π−2αi−sin(2αi)−ωπLXТРГ, where i is the iteration number.
Static var. compensator (SVC) is a multipurpose static device providing constant voltage regulation and a smooth or stepwise variation in the consumed and (or) reactive power produced by it at its connection buses. The SVC basis is accumulative devices (capacities, inductances), reactor-thyristor, and capacitor-thyristor blocks. In most cases, an SVC device comprises thyristor-switched capacitors (TSC) and a thyristor-controlled reactor (TCR). There are other possible combinations of devices, for example, a separate TSС or a separate TCR.
The smooth control of the reactive power in SVC devices is performed through a variation in the reactor's thyristor firing angle α. To maintain the specified voltage at a node, one should determine angle α.
When solving the SE problem, SVC is modeled by a susceptance variable at node i of the SVC installation, and either conductance bSVC, or angle α is involved in the state vector x instead of Ui that is fixed. In both cases, one should calculate derivatives by these variables from measuring the injection at this node, to calculate which nodal balance equation for a reactive power is used:
Qi=Qgi+Qli+∑j∈ωiQij+Qshi
where: Qgi, Qli is the reactive power generated and consumed in node i, ∑j∈ωiQij is the total of the reactive power overflows over the lines incident to node i, ωi is a set of nodes incident to the ith, Qish is a shunt reactive power at node i determined for SVC by formula:
QSVCi=U2ibSVC(α)
In (37), only the last summand depends on bSVC ⋅ (α). Calculating the derivative of QiSVC with respect bSVC ⋅ (α) does not cause difficulties, and the entire SE algorithm, in this case, practically coincides with the algorithm developed for TCSC. To calculate angle α after obtaining the bSVC(α) estimation, the expression presented in [44,45] is used:
bSVC(α)=1XCXL{XL−XCπ[2(π−α)+sin2α]}
where XC is the capacitive and XL is the inductive SVC impedances.
The derivative of QiSVC with respect to α is calculated as:
QSVCi=U2iXCXL{XL−XCπ[2(π−α)+sin2α]},
obtained upon substituting (39) into (38). The derivative of the injection measurement at the SVC installation node with respect to α is:
∂Qi∂α=2U2ibLπ(1−cos2α)
Both algorithms were developed to test the method. The calculation results and their comparison are presented below.
A 19-node scheme of the Irkutsk power system is used for calculations (Fig. 7). The scheme contains 19 nodes, 28 lines, and 94 measurements (SCADA measurements and PMs of zero injections at transit nodes).
Modeling TCSC. The calculations were performed for a scheme with a TCSC connection instead of a discretely controlled Tyret’ SC (line 3–4) for two states: when transmitting 1386 and 1852 МW over the Bratsk-Irkutsk 500 kV line. The Tyret’ device has the full capacitance XSC = − 26.3 Ω; this value was accepted as an initial approximation when calculating with TCSC. Table 2 shows how the xTCSC values varied during iterative calculations.
Table 2
Iteration no. | Transmitted active power | |
---|---|---|
1386 МW | 1852 МW | |
1 | − 26.30 | − 26.30 |
2 | − 18.51 | − 25.40 |
3 | − 18.46 | − 25.33 |
4 | − 18.39 | − 25.27 |
5 | − 18.36 | − 25.22 |
The calculations show that the series compensation rate varies with the change in the transmitted power value. For the 1386 МW transmission, it is necessary to compensate the Bratsk-Irkutsk transit reactance by − 18.36 Ω. If the transmitted power is higher (1852 МW), the compensation will be also higher, by − 25.22 Ω.
Modeling SVC. In the real scheme, to compensate the redundant reactive power at the 500 kV Irkutskaya substation, synchronous compensators (SCs) are installed at low-voltage nodes 17, 18, and 19. For calculations, an SVC installation (instead of SСs) was modeled at the same nodes. Table 3 presents the calculation results by the two algorithms.
Table 3
Node no. | kV | USE (kV) | δ (deg.) | Рmes (МW) | РSE (МW) | Qmes (МVAr) | QSE (МVAr) | b (Ω− 1) | α |
---|---|---|---|---|---|---|---|---|---|
Algorithm I | |||||||||
16 | 224 | 224 | − 8.1 | − 394 | − 393 | − 397 | − 397 | ||
17 | 10.5 | 10.5 | − 8.2 | 0 | 0.0 | – | − 15 | 0.136 | − 1.086 |
18 | 10.5 | 10.5 | − 8.2 | 0 | 0.00 | – | − 15 | 0.136 | − 1.086 |
19 | 10.5 | 10.5 | − 8.2 | 0 | 0.00 | – | − 15 | 0.136 | − 1.086 |
Algorithm II | |||||||||
16 | 224 | 224 | − 8.1 | − 394 | − 395 | − 397 | − 397 | ||
17 | 10.5 | 10.51 | − 8.2 | 0 | 0.00 | – | − 15.1 | 0.137 | − 1.085 |
18 | 10.5 | 10.51 | − 8.2 | 0 | 0.00 | – | − 15.1 | 0.137 | − 1.085 |
19 | 10.5 | 10.51 | − 8.2 | 0 | 0.00 | – | − 15.1 | 0.137 | − 1.085 |
As it follows from the table, practically conterminous values of SVC conductance and angles α are obtained in both calculations. These calculations show that SVCs operate in the inductive state, and compensate the redundant reactive power generated by the 500 kV high-voltage line of an extended Bratsk-Irkutsk transit, thereby maintaining the specified voltages at the 220 kV buses of the Irkutskaya substation (node 16).
The quality of the results of the SE problem solved online can be improved by using retrospective information on the state variables. The algorithms are developed for this purpose which capable of considering interrelations between the time-varying state variables unlike the SE algorithms which use the only snapshot. The SE problem, for which a set of snapshots is the initial information, is called DSE [46–50].
The main DSE advantages are the ability to deliver good results in the cases of bad and incomplete data and to forecast state variables. In this chapter the DSE is based on the extended Kalman filter.
Based on a priori knowledge of the system behavior, the state space representation for discrete time-variant systems looks as follows [46].
xk+1=Fkxk+ξF(k)
yk=Hkxk+ξy(k)
The first type of equations (42) is a dynamics model of the process x(t). The second type of equations (43) is the dependence of measurements ˉy on state vector x (ˉy(x)).
The unmeasured state variables can be calculated at the availability of a certain set of variables that allows the values of all state variables to be obtained by the noniterative technique. Such a set of variables is the 2n-dimensional voltage magnitude and phase x = (U, δ), where n—number of nodes in the calculated scheme and is called a state vector. In the SE problem, 2n − 1 components of vector x = (U, δ) are calculated via the measured state variables by minimization of some objective function. One state vector component (voltage phase at the slack node) is considered as constant.
The objective function in the DSE looks as follows:
J(x)=(ˉy−y(x))ΤR−1(ˉy−y(x))+(˜x−x)TM−1(˜x−x)
where ˜x—forecast of the measured state vector components, R—covariance matrix of the measurement error, and M—covariance matrix of the forecasting errors, which is calculated in the process of the Kalman filter formation.
Mk+1=FkPkFТk+WF(k)
WF(k)—covariance matrix of dynamics model noise, Pk—covariance matrix of the state vector component estimates. The matrix of transition F is assumed to be a unit matrix because the time frames considered are small enough and the system changes extremely slowly. The covariance matrix WF(k) and Pk are calculated by the recurrence relations.
In the objective function (44), the measured and forecasting values are represented by different terms. The addend considers the process history.
The objective function (44) can be written in the form
J(x)=(ˉyf−y(x))ΤR−1f(ˉyf−y(x))
where Rf=[R00M]−[(m+2n−1)×(m+2n−1)] matrix, ˉyf=[ˉy˜x]−[(m+2n−1)×1], m—number of measurements.
The criterion (46) is minimized by the state vector to calculate estimates of the state vector components and the DSE problem is reduced to solving a system of nonlinear equations:
HTfR−1f(ˉyf−y(x))=0
where Hf=[HHm], which is a [(m + 2n − 1) × (2n − 1)] matrix.
The system of equations (47) is not underdetermined, that is,
rank(∂yf∂x)≥2n−1
Due to the nonlinear dependence y(x), the problem is solved iteratively by Newton's method. The system is linearized at each iteration and a system of linear equations is solved where correction vector is calculated by the equation
Δxik=Pif(k)HТ(i)f(k)R−1fΔy
Pif=(HТ(i)fR−1fHif)−1
where Δy=ˉyf−y(xik), k—snapshot number, i—iteration number.
The estimate accuracy is determined according to the following equation:
ϕSE=m∑i=1(ˉyi−ˆyi)2σ2(y)i+m1∑j=1(ˉxj−˜xj)2σ2(M)j
where σ(y)i2—the variance of ith measurement (component of matrix R), σ(M)j2—the variance of forecast error of jth state vector component (component of matrix M).
The calculation model of the current EPS state is constructed using the SE method. The SE result is the calculation of the EPS steady state (the current state) based on the measurements of state variables and the data on the scheme topology state. The power system state is estimated by one snapshot of telemetry through the static SE algorithms. Availability of several snapshots allows the application of the dynamic SE algorithms [28,50,51].
The SE result is correct if there are no gross errors in the initial data. Such errors in initial information are the source of distortion of the calculated state obtained by SE, which can lead to wrong decisions at the EPS control and development of severe emergencies. Therefore the detection of gross errors (hereafter—BDD) in measurements and suppression of their effect on the estimates of the EPS state variables is a most topical issue in the SE problem solution [11]. In real time, the preference is given to the algorithms of a priori BDD [11,52], which allows the bad data to be detected and eliminated before calculating the estimates of the state variables.
At present many BDD methods have been devised based on the analysis of:
This work deals with the issues of applying the DSE methods to test the validity of measurements in the areas of low information redundancy, which contain critical measurements and critical sets [58].
The suggested method of bad data detection in measurements is based on the analysis of retrospective and forecasting information on state variables.
The measurements or estimates obtained at the previous snapshot are applied as retrospective information. The forecast value is calculated using DSE.
The retrospective information is tested in accordance with the conditions:
|ˉyi(k)−ˉyi(k−1)|≠0
|ˉyi(k)−ˉyi(k−1)|<di
|ˉyi(k)−ˆyi(k−1)|<ˆdi
Conditions (53), (54) are interchangeable. With the use of DSE, preference is given to (54). In the case of a change in the state, the results of the analysis of inequalities Eqs. (53), (54) will be interpreted as an error in the measurement. This disadvantage is eliminated by forming an additional condition consisting of innovation that takes into account forecasting information; it gives a correct answer in the case of the state change:
|ˉyi(k)−˜yi(k)|<˜di
In (52)–(55) ˉyi(k)—ith measurement at snapshot k; ˜yi(k)—forecast of the ith measurement at snapshot k; ˉyi(k−1)—ith measurement at snapshot k − 1; ˆyi(k−1)—estimate of the ith measurement at snapshot k − 1; di, ˜di, ˆdi—thresholds calculated by the formulas:
di=γ⋅√σ2i,˜di=γ⋅√σ2i+σNi2,ˆdi=γ⋅√σ2i+Yi
where γ—quantile of distribution N(0,1), is determined by the specified probability of error of the first kind α (at γ = 3 the measurement with the probability 0.997 is accepted as erroneous at the threshold violation); σN2—diagonal element of matrix ˜N, Yi—diagonal element of matrix Y, where Y = HPHТ, ˜N=HMHT.
Conditions (52)–(55) can be processed in parallel. As a result of the analysis, the three-digit error code XYZ is formed, where the first element (X) indicates satisfaction—1 or dissatisfaction—0 of inequality Eq. (52), Y and Z are Eq. (53) or (54), (55), respectively. The errors of the first and second kinds are possible by virtue of the erroneous results when using the BDD methods [59]:
The studies performed to minimize the errors of the first and second kinds are presented below.
The stability of the network toward physical and cyber intrusion is an attribute of the IESAAN [1]. While developing the conceptual smart grid models and projects, researchers nowadays pay great attention to the issue of cybersecurity.
Security is one of the most important properties of IESAAN, which ensures specified operation of the system in case of changes in conditions, failures of components, and unexpected disturbances. Malicious activities or cyberattacks should also be considered as such disturbances that may lead to a large-scale outage. Security of the IESAAN is necessary for the whole of its structure, which comprises:
The systems of data collection and processing—SCADA and WAMS, which belong to the subsystems of the smart grid—are most vulnerable to the physical failures and information attacks that are dangerous in terms of their consequences [60–62].
The SE tool is considered as a link between physical and information-communication subsystems. It acts as a barrier to the corruption of data on current operating conditions of the EPS in the control problem, including the data corruption caused by cyberattacks against data collection and processing systems of the EPS.
This section presents the analysis of possible consequences of cyberattacks against these systems and the accuracy of solutions to the SE problem. Also, consideration is given to the methods applied to detect bad data in measurements as well as identifying and mitigating their consequences.
The SCADA/EMS systems aimed at supporting the dispatching personnel actions in the operation and emergency control of EPS include RTUs installed at EPS substations to record signals on the state of switching equipment and measurements of state variables; communication channels; databases; systems of online representation of state variables; a software package (a set of programs of EMS applications) for processing of measurement results; and formation of control commands for the dispatching control facilities.
There are several vulnerable points in the architecture of SCADA systems, which can be used for cyberattacks. These can be direct damage of RTU communication channels between the RTU and control center and damage of software located on the servers of the EPS control center, including the SE software and databases in the control center.
Cyberattacks against SCADA systems can result in a long-term failure of the data collection system, which can cause the emergence of bad data in measurements. Therefore, it is appropriate to apply the effective bad data detection methods capable of identifying a group of the most severe cyberattacks that can, at present, be missed by technical and physical protection systems of SCADA in most cases of modern intelligent “attacks” [63].
The WAMS is a set of recorders of synchronized PMUs, PDCs, channels connecting recorders, and data concentrators. The existing SCADA measurements and PMU measurements duplicating them make some energy companies think that WAMS is not a strategically important system and the technical means of WAMS are not cybercritical components of the energy infrastructure. However, WAMS is vulnerable to cyberattacks. Analysis of potential cyberattacks [64–66] has shown that the greatest harm to WAMS can be done in the following way:
The specific feature of technological control in the modern EPS is imperative transmission of large data volumes to the upper levels of control. Normally, the data are transmitted through the corporate network, but recently points of interface with the Internet have appeared that make the network more accessible and thus vulnerable. In some cases, it is suggested to apply cloud technologies, particularly in WAMS, for collection and transmission of synchronized vector measurements from the PMU level to the PDC and super-PDC level. With a constantly growing number of cyberattacks against the information structure of energy facilities, the authors of [62] suggest enhancing cybersecurity, firstly, by a reduction in visible zones of potential attacks with the methods of traffic engineering, and, secondly, by the development of the next generation commercial anti-viruses and systems to detect the intrusion and block network attacks, and at the same time to detect the emergence of new network devices.
The PMU measurements coming at a high frequency make it possible to implement fast linear algorithms of SE for areas in the scheme of EPSs that are observed using PMUs.
The TE method makes it possible to perform a priori bad data detection (see Section 3.3.1.1).
The PMU devices are, first of all, placed at large power plants and high voltage substations. Historically, the same facilities were the first to be equipped with telemetry devices. Hence, they were monitored by SCADA measurements. And finally, the same facilities are currently the top priority targets for cyberattacks.
Let us see if the SE procedure copes with the detection of cyberattacks against WAMS system, using PMUs alone.
The modification of the Crout reduction method [27] for the linear estimation of such measurements gives both the estimates and the list of critical measurements at the same time, which is very important under the conditions of potential cyberattacks:
In a case where false information comes from individual recorders, the SE can independently surmount this problem by a priori detecting an erroneous measurement and replacing it with a PM. In the case where a large number of recorders are attacked, the number of erroneous measurements can lead to the situation that the computational process of SE will not converge. A simple check before the SE as to whether the current measurements lie within the technological limits can immediately show if an information failure has occurred in the system. According to the settings, the SE procedure is blocked in the case of a great amount of erroneous measurements to avoid startup of the SE, which can result in its failure.
A great number of rejected measurements can make the system unobservable. In this context a reminder is appropriate that, until now, the SCADA system and WAMS have been independent of one another. Hence, if a local region or an object is observable on the basis of PMU measurements, they are also observable by SCADA measurements. Therefore, it has been suggested, when needed, carrying out independent simultaneous (by one timestamp) bad data detection and SE by SCADA and WAMS measurements to find out if there is a malicious attack against one or another system.
According to the above considerations, the following methodology to identify cyberattacks against SCADA and WAMS is suggested:
To do calculations based on the suggested technique, a fragment of a real network is considered (Fig. 8). The fragment includes 750/330 kV lines. There are four PMUs (rectangles) at the ends of the lines 1–2 and 1–3 and a PDC installed at node 1 is common for them. Nodes 4–13 are observable by SCADA.
The calculations were done in the simulation experiment when the PMU measurements were simulated by distorting the parameters of the steady-state calculation.
The following types of cyberattacks were simulated:
Table 4 presents the SE results based on the data exposed to attacks, and measures undertaken by the SE methods to detect such data.
Table 4
Type of cyberattack | U (kV); Р (MW), Q (MVAr) | Reference y0 | Measurement ˉy | State estimation results | |||
---|---|---|---|---|---|---|---|
Value of measurement at attack yatt | Detection of attack by the state estimation methods | Corrected measurement ycorr | Estimate ˆy | ||||
Attack 1, against phasor data concentrator | U1 × 1.5 | 757.1 | 758 | 1136 | ˉU1 exceeded the limits | Excluded | 757.7 |
δ1,∘ | 2.7° | 2.7° | 2.7° | 2.7, true | 2.7° | 2.7° | |
P12 × 2.25 | 236 | 235.5 | 531 | ˉP12, rough error, is substituted: P12 = P21 − ΔP21 | 236 | 235 | |
Q12 × 2.25 | − 410 | − 411 | − 925 | ˉQ12—doubt measurement, its variance is increased σQ12 | − 925, is not corrected | − 402 | |
P13 × 2.25 | − 1051 | − 1052 | − 2370 | P13 = P31 − ΔP31(see ˉP12) | − 1049.4 | − 1056 | |
Q13 × 2.25 | − 126.5 | − 127 | − 295 | Q13 = Q31 − ΔQ31(see ˉP12) | − 81.2 | − 97 | |
Attack 2, denial-of-service | 757.1 | – | – | U1 = f1(U2, I21, δ2) | 750.0 | 757.1 | |
δ1,∘ | 2.7° | – | – | δ1 = f2(U2, I21, δ2) | – | 2.7° | |
P12 | 236 | 235.5 | – | P12 = f3(U1, I12, ϕ12) | – | 234 | |
Q12 | − 410 | − 412 | – | Q12 = f4(U1, I12, ϕ12) | – | − 412 | |
P13 | − 1051 | − 1052 | – | P13 = f3(U1, I13, ϕ13) | – | − 1050 | |
Q13 | − 126.5 | − 127 | – | Q13 = f4(U1, I13, ϕ13) | – | − 128 | |
Attack 3, de-synchronization | U3 | 761.8 | 761.9 | 761.9 | 761.9—true | 761.9 | 761.8 |
δ3,∘ | 6.4° | 6.4° | 15.4° | Error is detected by the TE (6) | 6.4° | 6.4° | |
P31 | 1058 | 1056 | 1056 | 1056 | 1056 | 1057 | |
Q31 | − 127 | − 126.5 | − 126.5 | − 126.5 | − 126.5 | − 127 |
Attack 1: Based on the PMU data alone, the SE procedure identified the measurements that arrived at PDC: 1 measurement goes beyond the limits, 3—contain bad data, 1—is doubtful. So many erroneous measurements at one PDC mean a complete failure in its operation and initiate the identification of reasons for such a failure. It is necessary to independently detect bad data by using SCADA measurements.
Attack 2: Accurate data of PMUs 1 and 4 provided accurate estimates at node 2.
Attack 3: As was said above, if an angle shift occurs in a certain PMU channel, the same shift will happen in all phase channels. Therefore, the same shift in the voltage angle δ3 and current angle ψ31 will not affect the angle between voltage and current vectors ϕ31 = δ3 − ψ31; hence, this error will not change the values of active and reactive power. In this case, if the neighboring PMU gives a valid angle measurement, then δ3 will be calculated correctly.
At present the SE procedure is becoming increasingly more important in the conditions of adjustment of the automated systems for acquisition of phasor measurements as well as in the conditions of various cyberthreats. The mathematical tools of SE allow straightening out a tangle of true and erroneous measurements and make certain conclusions on the operability of the devices for collection and primary processing of measurements. The results of bad data detection using two independent data collection and processing systems make it possible to conclude whether or not there is a malicious cyberattack against the considered energy facility.
Emergency control, including emergency operation dispatching and automatic emergency control that provides reliability and survivability of the EPS, takes an important part in controlling the operating conditions of the UES of Russia. Emergency control is performed by the technological (dispatching and automatic) control systems that include the automatic systems of voltage, power, and frequency regulation, basic automatic systems of EPS elements, relay protection and automatic line control, and the ECS [67,68].
Integration of EPSs, liberalization, and modernization of the electric power industry increase the changeability and unpredictability of EPS operation and generate the need to improve and develop principles as well as systems of operation and emergency control. New conditions call for the development of a comprehensive system for monitoring, forecasting, and controlling EPS. Artificial intelligence application is an advanced way to carry out smart emergency control in EPS [4,69,70].
Emergency control philosophy in the UES of Russia is a hierarchical approach and is realized by the coordinated operation of many control devices, which maintain EPS stability and interrupt the expansion of an emergency situation in the case of stability violation and a threat of undesirable cascade emergency development. Coordinated emergency control is realized by joint participation of generators, networks, and consumers.
The electric power object called the UES of Russia is a power interconnection where seven IPSs are combined by weak ties. Under emergency conditions, the UES of Russia as a power interconnection is able to disintegrate into autonomously operating self-balanced IPSs without grave consequences.
At the same time, disintegration of any of the seven IPSs is far from smooth. Because of large power flows in the ties, practically any disintegration creates parts with power lack and power surplus. As a result, load and generator disconnections are quite possible and can acquire a cascade character, that is, turn into a system crash fault. For this reason, at the IPS level the generator and load disconnections by special system emergency devices (in order to unload cutsets and to maintain admissible values of operation parameters) are of course more preferable than cascade disconnection by protection devices (after these parameters go beyond the fixed limits). Just this principle is basic for arranging emergency control in Russian EPSs [67,68,71,72].
The ECS in Russia's IPSs includes the following components (Table 5):
Table 5
The emergency control system of Russia's UES has just now the elements of smart grid ideology, including:
Modern electricity grids continue to be vulnerable to large-scale blackouts. During the past two decades, events in North America, Europe, and Russia have clearly demonstrated an increasing likelihood of large blackouts, which cause huge economic losses; disruptions in communication and transport, heating, and cooling; water supply; emergency services; and financial trading. Blackouts now typically spread across borders, escalating their impacts, and the trend toward national grid integration is likely to increase the risk of even larger power outages in the future.
In recent decades, power industries worldwide have experienced two major changes: liberalization of the electricity market and the expansion of renewable energy. Liberalization has resulted in the separation of power generation, transmission, and distribution businesses. This process has created additional boundaries that have adversely impacted communication and coordination activities between power system operators. The increasing penetration of renewable energy has led to an increasing dependency on the volatile nature of renewable energy sources. It is not only a shortage of power supply that can result from a blackout; oversupply can lead to frequency and voltage instability problems.
Because load centers are often remote from the sources of renewable generation, large amounts of energy often have to be transported over long distances or even through distribution grids that were not designed for this purpose. This leads to higher loading of the grid and the necessity for costly grid upgrades. However, due to several factors—including public resistance—planning and extending the electricity network can take many years. In the meantime, grids have to operate closer to their operational limits, and the probability of large-scale emergencies and blackouts is inevitably increasing.
The analysis showed that at present, the most relevant types of disturbances that lead to major system failures are the voltage collapse (violation of restrictions on voltage levels) as well as overloading equipment (violation of current restrictions). It has been found that currently the interconnections are well protected against these types of disturbances by means of stability control automation. In turn, the appearance and uncontrolled development of voltage instability and current overload in distribution power networks with a complex structure usually leads to serious consequences.
In the near future, the development of the networks in megalopolises and large industrial centers in Russia will result in the formation of systems with a complex multiloop structure. In these conditions, one should expect large-scale system emergencies that will occur according to the scenario in which the electrical current and voltage constraints become decisive in case of emergency operating conditions unfolding. The first such system blackout happened in the UES of Russia in the Moscow power system in May 2005.
It is necessary to note that the existing emergency control concept does not have full defense from voltage collapse because the running processes are much faster than the speed of control action implementation from the emergency control. The applied automatic devices intended for limitation of the voltage drop are extremely slow due to a time delay to consider short circuit duration. The existing automation for limiting the equipment overload does not regulate overload of active and reactive power sources.
Modern automatic stability control can theoretically cope with voltage collapse, but its principle of adjustment does not allow performing an adequate control in off-design conditions under nonstandard disturbances that occur at cascade emergencies in a complex interconnected network of megalopolises. The mentioned flaws of the existing automation as well as an increase in criticality of operating conditions and their control lead to the need to develop automatic intelligent systems for monitoring and control of security, which will operate online and be capable of harmoniously supplementing the modern automatic operation and ECSs without contradicting the concept of their operation. The complexity of the problem of the development of such systems lies in the fact that most of the dangerous preemergency states of EPSs that lead to large blackouts are unique and there is no single algorithm to effectively detect such conditions under the time shortage conditions. In addition, the problem is complicated by the fact that the security limits of EPS are permanently changing.
Large-scale blackouts that occurred in 1965–2014 in the power interconnections of different countries were analyzed in [74]. The analysis made it possible to identify the general regularities of their development, which are expressed in some typical phases (Fig. 9): preemergency state, initiating events, cascading development of emergency, final state, and restoration. It was established that the main types of emergency disturbances that occur in the quick phase of development were a voltage collapse and a considerable overload of equipment.
The analysis showed that in the phase of initiating events, the above-standard disturbances occur. The postemergency conditions that occur at the end of this phase are off-design for the existing emergency control devices and for the dispatching personnel. Therefore, the existing ECSs furnished with the up-to-date automation means and the actions of a transmission system operator may prove to be ineffective to prevent the subsequent catastrophic development of the emergency. Thus, the following drawbacks of the existing ECSs are:
The results of the studies testify to the necessity of the development of next-generation intelligent systems to complement modern ECSs, taking into account its weak points [75,76]. Based on the above, the specific requirements for such intelligent ECSs have been developed. The systems should:
Advanced smart devices and technologies such as PMU, artificial intelligence, and so on give new possibilities for solving the complex and comprehensive problems of emergency control using monitoring, forecasting, and control of EPS operating conditions. The time sequence of the individual stages of monitoring, forecasting, and control of operating conditions is shown in Fig. 10.
In essence, this combination of stages represents a comprehensive system providing stability, reliability, and controllability of modern EPSs. At the same time, from the viewpoint of the efficiency of this system, it is essential to increase the adaptability of control and improve coordination of the control stages, means, and systems. The blocks of monitoring and forecasting of the EPS normal, preemergency, and postemergency operating conditions solve the following problems:
Effective organization of the system of EPS operating condition monitoring and control is possible by an extensive involvement of new tools for the analysis and calculation of operating conditions and control actions, primarily technologies of artificial intelligence.
The balance between electricity consumption and generation must be kept in the electricity grid at any moment—otherwise disturbances in power quality or supply may occur. Forecasting EPSs parameters may be considered at different time scales, depending on the intended application. Forecasts for milliseconds up to a few minutes can be used, for instance, for turbine active control. This type of forecast is usually referred to as very short-term forecasts. Forecasts for the following 48–72 h are needed for power system management or energy trading.
The short-term forecasting of EPS parameters can be carried out both with the aid of classical approaches of dynamic estimation, statistical methods of analysis of time series, and regressive models, and with the aid of artificial intelligence. Many techniques have been employed for such purposes, including machine learning techniques—ANNs [77,78], support vector machines (SVMs) [79], random forest models [79,80], etc. Moreover, time series models, such as autoregressive integrated moving average (ARIMA), generalized autoregressive conditional heteroscedasticity (GARCH) models [81,82], and Kalman filter-based algorithms [83,84] have also been proven to be effective in EPS parameter forecasting. It is worth noting that EPSs parameters often feature sharply variable nonstationary behavior, which limits the efficiency of the stated technologies for the forecasting of time series. Studies have shown that hybrid approaches, that is a combination of different intelligent techniques, have great potential and are worth pursuing [85]. Previously, several hybrid forecasting approaches have been proposed, including ARIMA-ANN models [86], Fuzzy-ANN-based models [87], Wavelet-ANN-based models [88], and Fuzzy-expert system-based models [89], among others [90]. For a state-of-the-art review and bibliography of the methods for dynamical system identification and forecasting, readers may refer to the monograph [91].
It is well known that for effective forecasting of EPS parameters, a number of additional, correlated characteristics must be taken into account, which can be influenced by a predictable parameter. For instance, everything from salmon migration to forest fires can affect current and future electricity prices. But quite often, information about these characteristics is either unavailable or limited and contradictory. As a result, scientific researchers and power engineers have to work with only one retrospective predictable parameter time series without additional characteristics, which imposes limitations on forecasting performance. In such a case, forecasting performance can be improved by using a transition from a time series task to a regression task, when an initial time series is decomposed into some components. An investigation of properties of obtained components can help in improving the performance of an EPS parameter forecast.
Analysis of the mechanisms of the development of large-scale blackouts an emphasizes the voltage instability as a main reason for blackouts in EPSs [92,93]. Enhancement of the relay protection systems, voltage controllers, and SCs as well as an increase in generator rotor speed led to a rise in the power flow dynamic stability limits of EPSs, which allowed the transmission of large amounts of power at considerable distances. Against the background of the rise in the transfer capability, it becomes necessary to meet the requirements for a necessary level of system security by providing sufficient backup sources of reactive power, which is not always fully implemented. This, in turn, facilitates voltage instability, which can lead to voltage collapse in a large-scale EPS.
At present, system operators worldwide widely use a “manual” approach to reactive power control. The control includes online adjustment of voltage settings as well as reactive power loading of generators according to the schedule, connection, and disconnection of static capacitor banks and shunt reactors, resetting of FACTS device settings, and automatic control of the transformation ratio at on-load tap changers. Moreover, the online control of EPSs deals with the notion of “prediction of emergency situation” (probability of its occurrence), which represents a problem of monitoring and security assessment. The prediction functions can be applied in various software packages, for example, dispatcher advisors. However, in the end, any control action is implemented directly by the operator that for some reason cannot make fast and effective decisions capable of preventing further emergency development.
In a general case, to assess voltage stability in EPS in real time, the problem of finding the distance of an operating point from the voltage stability is raised. The measure obtained may be qualitative or quantitative. The qualitative measure normally requires less computation time compared to the quantitative. It does not provide an estimate of the exact power reserve in megawatts but determines some dimensionless number (that normally varies from 0 to 1), which can be interpreted as a stability indicator. By definition, such indicators of security are scalar magnitudes and are determined by monitoring of change in certain parameters of EPSs. These stability indices seem quite convenient for dispatchers, first of all, to generate respective preventive control actions.
In the traditional statement, the application of such stability indices implies a purely algorithmic approach where the specified equations or partial derivatives are used to calculate numerical values of these indices for each current state of the EPS. However, as the practice of operation shows, such an approach has a number of significant downsides: low robustness to erroneous inputs, computational complexity erroneous identification of states, and some others [94]. One of the effective solutions to this problem is the use of a combination of traditional approaches on the basis of voltage stability indices and machine learning algorithms. The main idea here lies in an intelligent model learning to independently determine the current value of an assumed indicator on the basis of input data, thus identifying the current state of the EPS. As the studies [95,96] show, such a modified approach makes it possible to neutralize the drawbacks of traditional algorithmic approaches, owing to the original properties of the machine learning technologies.
To monitor that the system is within its limits, determined either online or offline, the primary measurement tools are SCADA systems and postprocessing by a state estimator [97]. The ENTSO-E network code on operational security requires each transmission system operator to classify its system according to the system operating states [98]. Significant advances have been made in the field of online security assessment technologies in [99], where 19 tools for dynamic security assessments in use, under testing, and under development were reviewed. A review of 15 of these state-of-the-art tools shows a wide variety of implementations. These systems include assessment of voltage security, transient security, and small signal security. The range of assessment capabilities includes determination of critical contingencies, transfer limits, and determination of remedial measures necessary to ensure security. The computational methods used for each type of security assessment are varied, and depend on the specific requirements, system characteristics, and, in some cases, the techniques available in the state-of-the-art tools used. All the tools reviewed in [97] were deterministic tools; however, commercially developed online risk tools do exist. In recent studies [100,101], state-of-the-art methodologies on risk assessment of cascading outages are presented. The methodologies are divided into two main categories: detailed modeling and simulation methods and the bulk analysis methods [102].
Computational intelligence techniques such as decision trees and ANNs have been applied to achieve the essential online intelligent security monitoring and assessment performance. A research on computational intelligence applications to EPS security [103] demonstrates that intelligent systems can play a major role in preventing major emergencies and blackouts by providing information that suggests optimal remedial action.
Beyond better monitoring, new control approaches are required. Six new control technologies and control schemes have been reported in the recent international European/Russian FP7 project ICOEUR [104]. The focus is on coordinated power flow control, transmission technologies, HVDC/VSC-HVDC technologies, optimization of interconnections and interarea oscillations, and different interconnection concepts.
Multiagent systems have been proposed for the last decade as a major component of intelligent control grids [105]. The research project ICOEUR also introduced a multilayer architecture for system monitoring and control. In this architecture, the physical layer was represented by the IPS, which consisted of several subsystems. Control actions that affected the subsystem were performed only if boundary conditions remained within the specified limits. If the boundary conditions were violated, coordinated control actions were applied using a multiagent control system [104]. The research of multiagent voltage and reactive power control system development is presented in [106]. The prototype of the system has been developed by the R&D Center at FGC UES (Russia). The control system architecture is based on the innovative multiagent system theory application that leads to the achievement of several significant advantages (in comparison to traditional control system implementation) such as control system efficiency enhancement, control system survivability, and cybersecurity.
3.141.192.120