3.2.1 Time Horizons

In the planning and operation of power systems, Transmission System Operator (TSO) actions are performed in different processes and time horizons [1]. The main objective is to maintain a high level of reliability. Basically, three main processes can be distinguished: grid development, asset management and system operation. TSO actions are taken in time horizons ranging from long-term, mid-term and short-term to real-time. Table 3.1 shows some typical activities for each of the main TSO processes and for different time horizons. As can be seen, TSO actions range from long-term grid development (with a timescale of decades) to real-time system operation (with a timescale up to real-time).

3.2.2 Overlapping and Interaction

There is always some overlap between the three main TSO processes. For example, the installation of new assets can be a grid development activity or an asset management activity. In asset management, maintenance activities are scheduled, but these could be cancelled in system operation during critical situations. Traditionally, TSO activities were mostly performed sequentially, as shown in Figure 3.1. An example is the Dutch 380 kV-ring, which was designed to be n-2 redundant (n-1 redundant during maintenance). This gave enough room to plan maintenance activities in asset management while still providing n-1 redundancy during system operation. As can be seen in Figure 3.1, in the sequential approach there is only a small overlap between the TSO processes.

In the future, the developments as mentioned in the introduction of this chapter are expected to put more stress on the transmission network. The transmission network will become more heavily loaded, and the room to plan TSO activities will decrease. Consequently, the overlap between the processes will increase and the planning of TSO activities will be more interacted, as illustrated in Figure 3.1. A typical example is the development of offshore grids. Other studies showed that it is often not economical to apply full n-1 redundancy in offshore networks. However, if there is no redundancy, maintenance activities will have more consequences for the availability of the offshore grid and therefore maintenance planning can become challenging. Moreover, without offshore redundancy, failures of offshore networks can have a large impact on the onshore power system during system operation as it becomes more likely that a substantial amount of wind capacity is interrupted. Here, the interaction of grid development (n-1 redundancy), asset management (maintenance planning) and operational planning (remedial actions) can be clearly seen.

3.2.3 Remedial Actions

As there is always interaction and overlap between the three main TSO processes, it is important to model this in the reliability analysis as well. In the application example described in this chapter, the focus is on long-term grid development decisions and the consequences for (short-term and real-time) system operation. It will be discussed how the application of new underground cable connections in grid development is related to remedial actions like generation redispatch and load curtailment in system operation.

Remedial actions are operational interventions performed by the TSO to avoid/relieve overload of the network during critical situations. Typical remedial actions are: application of PSTs (Phase-Shifting Transformers), network reconfiguration, cancelling maintenance, local generation redispatch, cross-border-redispatch and load/wind curtailment [2]. With PSTs, the phase shift of the voltage can be varied such that the magnitude and direction of the power flow can be controlled. The network can be reconfigured by performing switching actions to reduce the loading of overloaded connections. This is mainly applied in the high voltage (HV) and lower voltage networks. If a connection is under maintenance in an overloaded part of the network, it can be decided to cancel this maintenance to relieve the loading of other connections. By performing local (or national) generator redispatch, the production of some generators is increased while it is decreased for other generators to reduce the loading of connections in overloaded parts of the network. If local generator redispatch is not sufficient, cross-border (or international) redispatch can be performed. Wind curtailment can also be used to relieve the overloading of the network. The last resort to relieve overloading is to disconnect load by load curtailment (or load shedding).

In system operation, a risk framework is often used in which the current risk status of the network is indicated. The reliability (measured as the level of redundancy) can be related to the risk categories of this risk framework [3]. Remedial actions can be related to these risk categories as well. Table 3.2 shows that remedial actions like PSTs and maintenance cancellation are often applied first, while load curtailment is the last resort. Although Table 3.2 shows how remedial actions are related to the risk categories and redundancy levels, the choice for a certain remedial action always depends on the situation.

3.3.1 Security-of-Supply Related Indicators

The main function of a power system is to supply the load. If the power system is not able to supply its load, it is considered as unreliable. Therefore, the reliability indicator security-of-supply is often regarded as the most important Key Performance Indicator (KPI) of power system reliability. Several reliability indicators exist that are directly related to security-of-supply. Some reliability indicators are specially defined to measure the reliability of a part of the power system, while others are developed for combined generation/ transmission/distribution systems.

Reliability indicators developed to measure the reliability of the generation system [4, 5]:

• LOLP (Loss of Load Probability): probability that the demanded power cannot be supplied (partially or completely) by the generation system. The LOLP is often determined based on a per-hour study for a studied time period (usually a year)
  3.1

3.3.2 Additional Indicators

Traditionally, power system reliability studies concentrate on these security-of-supply related indicators. Often, it is assumed that there is one TSO that can take any remedial action to secure the load supply. The liberalisation of the electricity market has, however, led to several new actors such as producers, consumers, service providers and system operators. All of these actors have their own vision on power system reliability. For example, consumers wish to have a reliable power supply, while producers wish to be connected to the grid and sell their electricity. For a TSO, the transmission network is reliable if the customers (i.e. producers and consumers) are able to trade in electricity, while the network should also be maintainable.

In this sense, the reliability of a power system is not only reflected by the security-of-supply, but also related to indicators like the probability of wind curtailment, the probability of generation redispatch and the maintenance possibilities. These aspects can be included in reliability analysis by calculating several reliability indicators instead of one. As remedial actions are related to the risk states (see Table 3.2), calculating the probability of these states can provide more insight as well. Possible additional reliability indicators are then [2]:

• Probability of Overload: probability that one or more connections in the network are overloaded during a given time period (usually on a yearly basis)
  3.9

3.4.1 Input Information

The following input information is needed to perform the reliability analysis:

• Load scenario: In the load scenario, the load is specified for every hour of a study year. The load is aggregated per substation in the studied network. The load scenario is based on the expected development of (domestic and industrial) load and assumptions about the development of RES and conventional generation.
• Wind scenario: In the wind scenario, the production of large-scale (offshore) wind parks is specified per hour of the study year. The wind scenario is based on a time series of the wind speed, taking parameters like the size and location of the wind farms into account.
• Generator parameters: For all generators in the studied power system, information such as the location, fuel type, costs and capacity is needed. This information also takes into account developments such as the installation or closure of units. National generators are included, as well as some generators in surrounding countries.
• Development scenario: In the development scenario, assumptions are made about the development of RES and conventional generation (e.g. [12]). Also, assumptions can be made about preferences for the generator units (e.g. preference for fuel type or location).
• Network topology & component parameters: This information includes the network topology, as well as all required information to perform a load flow calculation (e.g. component impedances and admittances, component capacities).
• Component repair times and failure frequencies: This information is needed to calculate the reliability of network components. Failure frequencies and repair times can be obtained from databases of historical failures. If the volume of available failure statistics is too limited (e.g. for new component types), estimations of failure frequency and repair time can be made.

3.4.2 Pre-calculations

The input information is now pre-processed in order to be used in the reliability analysis. The following calculations/analyses are performed:

• Unit commitment: In this analysis, the production of the generator units is scheduled such that all the load in the power system is supplied. The load scenario, wind scenario and generator parameters are the input information to this analysis. Network information and the assumptions of the development scenario can be used as restrictions for the unit commitment. The result of the unit commitment study is a generation scenario for the national generators and an import/export scenario, which is based on the generators in the surrounding countries.
• Component reliability: The reliability behaviour of network components is described based on the failure frequencies and repair times of the components. This information is then collected in contingency lists, which contain the contingencies to be considered in reliability analysis together with their parameters (such as failure frequency, repair time, unavailability [4, 5], component status). Independent as well as dependent failures, in which one cause leads to a simultaneous failure of multiple components, can be included in the contingency lists too.

3.4.3 Reliability Analysis

Based on the input information and the results of the pre-calculations (i.e. the load flow scenario, the network information and the contingency lists), the reliability of the network can now be analysed. There are two main approaches to probabilistic power system reliability analysis: state enumeration and Monte Carlo simulation [4, 5]. The first is a structured analytical study of possible contingencies, the second is a computer simulation. Both approaches are discussed in this section.

• State enumeration: In state enumeration, the reliability analysis is based on the possible contingency states. The general algorithm for state enumeration is shown in Figure 3.3. It can be seen that for every contingency state, the probability of this contingency state is calculated and a load flow is performed. If there is an overload in the network, remedial actions are applied. From the application of these remedial actions, information is gathered that will be used to calculate the reliability indicators. After applying remedial actions, the load flow is calculated again and if there still is an overload, more (or other) remedial actions are applied. The state enumeration stops when all contingency states to be considered have been studied.

  

• Monte Carlo simulation: In a Monte Carlo simulation, the reliability of the power system is studied by performing a computer simulation. The general algorithm of a Monte Carlo simulation is shown in Figure 3.4. The Monte Carlo simulation uses the same input information as the state enumeration, albeit in a somewhat different way. First, from the failure frequencies specified in the contingency lists, the first time to failure (TTF) is sampled assuming an exponential distribution [4, 5]. The component with the smallest TTF will fail first and using its repair time, the time at which this component will be repaired is calculated. Until the component is repaired, a contingency exists in the network and the load flow is calculated for this period using the generation/wind/export/load scenario. If there is an overload in the network, remedial actions are applied in a similar way to state enumeration. Also similarly to state enumeration, intermediate results are collected to calculate the reliability indicators. A somewhat more complicated situation exists when a second contingency occurs before the first contingency is repaired. In this case, there are multiple contingencies in the network, which can also be combinations of independent/dependent failures as specified in the contingency lists. This situation ends when one of the contingencies is repaired, after which the other contingency is still being repaired. In contrast to state enumeration, it is possible (but very unlikely) that higher-order failure combinations occur during the Monte Carlo simulation, because component failures are randomly sampled and multiple components can be sampled as ‘failed’. The simulation stops when the given simulation time is reached.

  

3.4.4 Output: Reliability Indicators

The results of the reliability analysis are various reliability indicators. As described in Section 3.3, these reliability indicators can be directly related to security-of-supply, but can also be additional reliability indicators. Together, these reliability indicators give a more complete understanding of the reliability of the studied transmission network.

3.5.1 Input Parameters

The application example examines how the installation of more EHV underground cables (UGC) in the Dutch transmission grid affects the overall reliability level. Although the cable failure frequency is very close to that of overhead lines (OHL), the additional components of UGC (joints and terminations) reduce the reliability of the whole system [13]. Therefore, the unavailability of a connection is higher when it is partially or fully cabled than when it is completely an overhead line. Moreover, underground cables have a lower characteristic impedance. This is why it is important to study how the increased unavailability and the reduced impedance of the connection affect the overall reliability level.

As already described in Section 3.4, the reliability analysis consists of four major parts: contingency definition, load flow analysis, application of remedial actions if necessary and calculation of KPIs. A state enumeration is performed for the contingency types: independent single circuit failures, dependent double circuit failures, dependent triple circuit failures. All these contingencies refer to OHL and UGC failures. In addition, combinations of two of these contingency types are considered. In this way, 2^nd-order contingencies can occur with two single circuit failures or with one double circuit failure and 6^th-order contingencies are (only) reached by the combination of two triple circuit failures. Moreover, in this example three corrective actions are used with the following priority: national generator re-dispatch, cross-border re-dispatch and load curtailment. Taking these remedial actions into account, the state enumeration algorithm as shown in Figure 3.3 now becomes as shown in Figure 3.6.

In order to examine how further 380 kV cabling in the Dutch transmission network will impact the overall reliability level, simulations are performed by installing UGCs in three different connections (named Con1, Con2 and Con3). The PLC, the probability of overload and expected redispatch costs are calculated. The network topology and the required data are determined according to TenneT's scenario2020. It provides hourly data regarding generation (conventional/wind), load and export/import for the specific year. It is also important to describe the configuration of UGCs which is used in this study. As shown in Figure 3.7a, the UGCs consist of two circuits, and each circuit phase consists of two separate cables. In this configuration, a failure of a separate cable leads to a failure of one circuit, as operation with only one cable per circuit phase is undesirable in system operation.

In the case study, the cable length in the considered connections varies from 0% to 100% cabling with steps of 25%. The percentage refers to the transmission length of the connection and 0% means purely OHLs while 100% means a fully-cabled connection (as shown in Figure 3.7a). For example, in Figure 3.7b, 50% of the connection is cabled. The number of cable sections per connection can vary as well. For example, in Figure 3.7c there are two cable sections in the connection. Figure 3.7d shows the standard OHL connection configuration.

Due to the limited experience with EHV UGCs, there are no accurate values for their failure frequency and repair time. An earlier survey among European TSOs is used where a high and a low estimation of the failure frequencies are given, as shown in Tables 3.3 and 3.4 [7]. The repair time of UGCs is estimated as 730 h (1 month), which could be reduced to 336 h (2 weeks) if the repair process becomes more optimised in future [9]. For OHL, long experience has given more insight into failure statistics. The values are derived from actual failure statistics of the Dutch network [11].

3.5.2 Results of Analysis

The reliability analysis algorithm developed (Figure 3.6) is used to execute a number of simulation sets, and both categories of KPIs (directly linked to loss of load and not directly linked to loss of load) are calculated. Fig. 3.8 shows how the PLC changes when varying cable length is installed in the three considered connections. While one of these connections includes UGCs, the others are full OHLs. There are three curves, each devoted to a specific connection. The y-axis represents the PLC in h/y, while in the x-axis the cable length is presented as percentage of the connection length. The point 0% indicates that the three connections are OHLs, and this is the starting point for all curves.

The installation of UGC in Con1 has no impact on the PLC, while the other two curves illustrate an upward trend as cable length increases. However, while 50% cabling in Con2 causes a 5% increase in PLC, the same percentage of cabling in Con3 leads to a 70 times higher probability. In order to explain this phenomenon, the relative loadings of these connections were studied, and it was observed that the maximum loadings present similar behaviour with the amount of increase of the indicator, namely very small loading for Con1, larger for Con2 and even more considerable for Con3. It seems that the loading of the connection where UGCs are installed influences the impact on the reliability level.

Fig. 3.9 examines the probability of overload when UGCs are installed in each of the three connections. By comparing Fig. 3.8 with Fig. 3.9, it can be seen that the probability of overload in Con1 shows a very small upward trend as cable length increases (and is not constant like the PLC). The other two connections are characterised by an increasing trend as well. It is also clear that the values of the probability of overload are orders of magnitude larger than those of the PLC.

Fig. 3.10 illustrates the expected redispatch costs when UGCs are installed in each of the three connections. By considering a specific percentage of cabling, the three points of the connections are depicted next to each other. This occurs for each cabling percentage. The diagram is normalised (logarithmic), where the base 1 represents the starting point (0% cabling, three connections as OHL). The expected redispatch costs confirm the remarks made so far. They show a growing trend with rising cable length, and the installation of UGCs in Con3 seems the most expensive option (from redispatch costs point of view). Con3 is a heavily-loaded connection, and its failure appears several times in the contingencies which lead to generator redispatch. On the other hand, Con1 appears to be the most economical solution, since even at 100% cabling, the increase of the indicator is less than 10% from the starting point.

The results were verified using a Monte Carlo simulation as shown in Figure 3.11.

To sum up, the installation of UGC in specific connections might not influence the reliability indicators related to load curtailment, compared to the starting point. However, this does not mean that the level of reliability remains the same. It was shown in the results that in these cases, the reliability indicators which are not directly related to security-of-supply might demonstrate significant change leading to a lower reliability level. Therefore, it can be concluded that indicators not directly linked to load curtailment should be used as well.