Contingency Analysis

Evaluation of contingency is an important exercise for knowing about emergency conditions in power systems. Without understanding the severity and effect of a specific contingency, the system operator at Energy Management Systems (EMS) can not commence preventative measures. Assessment of contingency is an signif- icant instrument for assessing system security. On the other side, prior forecast of critical contingencies (which may represent a future threat to system stability (voltage or rotor angle)) enables system operators to perform corrective actions and maintaining the power system in a secure state.

A lot of research work has been carried out [39, 43–68, 128–140] in this area in the past several decades. The various approaches for contingency analysis are broadly classified as: Conventional methods, ANN-based methods, other AI-based and hy- brid methods etc.

2.1.1.1 Conventional Methods

In order to perform the contingency analysis, the system quantities are calculated for all probable contingencies. For this several load flow methods such as the Gauss- Seidal (GS), the Newton-Raphson (NR) and the Fast Decoupled (FD) [128] were used. These methods are very useful in order to obtain the load flow solutions un- der the contingency scenario, which helps to compute and rank the system severity.

Most of the methods utilize a scalar quantity called the Performance Index (PI).

Severity ranking given to the system contingency is based on the value of PI, which is the used to measure the contingency severity or system stress expressed in forms of the system parameters like reactive power and voltage magnitudes etc. Contin- gency with severity is given higher rank as compared to a contingency which is less severe. The application of the AC load flow in order to solve the outage cases with respect to the reactive power and voltage magnitudes are discussed in [129–133].

In the literature [39, 134], methods with higher accuracy are proposed in order to calculate the distribution factors based on the decoupled and the Newton-Raphson load flow methods using network sensitivities. The obtained factors are used to

compute the post-outage reactive power flows and the voltage magnitudes following a transmission line or a generator outage.

The AC load flow equations need to be solved for each contingency case in order to evaluate the limit violations. However, it is not feasible to perform online, because of the computational barrier. In order to overcome the barrier various approximate methods have been developed. There exist two techniques namely the explicit and the implicit techniques. The explicit methods [135–140] are the ranking methods, where the contingencies are ranked on basis of the order of severity using any PI.

Performance Indices

Most of the methods presented in literature for contingency screening and ranking are based on analytical techniques. Out of them PIs based methods are widely accepted [43–68]. The following section discusses about various performance indices capable of predicting the severity of contingencies and the power system security status. On the basis of literature review it is judged that the contingency ranking is performed by the scalar PI that measure the system security in terms of violation of transmission line loading and bus voltage. System behavior after contingencies is dynamic in nature and relying upon the loads at various loads at various buses, so it’s observed that a critical contingency may be a non-critical one at some other loading condition. In order to evaluate the security by these approaches, it is necessary to compute the stress experienced by the system for a specific contingency which is termed as the severity. The severity is basically computed using the violations of the line flows, bus voltages etc. Over the years, the severity is referred by different names such as performance indices, the composite indices, the overall performance indices, the severity indices etc. Though the names appear different, they all compute the severity of the contingencies, but by considering different violations or by combining two types of severities. The indices which used to compute the severity are listed below.

1. Severity Index: This index computes the MVA line flow violations.

2. Active power performance index: This index computes the active power line flow violation.

3. Voltage performance index: This index computes the voltage violations at buses.

4. Composite performance index: This index computes the line flow and voltage violations as a single value.

5. Static Severity index: This index computes the MVA line flow violations.

The literature reveals the indices with the same name but a variation in the formulation in order to compute the severity based on the specific application. The most widely used indices are the active power and the voltage performance indices.

Contingency with highest severity is ranked as number one and going down the list with the least severity. Whereas, implicit methods use the network solutions in order to recognize the system violations and rank the severity for the various outages [40,141–146]. A partial system solution approach in [40,141,142] and an approximate approach in [143,146], were used to improve the computational speed. The method of concentric relaxation is introduced for the security monitoring. In this approach the system is treated as electrically rigid and gradually relaxed the results to compute the actual flexibility of transmission. These approaches consider only part of the system network in order to identify the branch flow violations. However, obtaining voltage violation is very complex. Thus, the authors in [147] have proposed complete bounding method to identify the line flow violations and the voltage violations. This method reduces the number of line flow computations and limits checking. The zero mismatch method [148] is proposed for quick power flow solutions by exploiting the difference in the convergence rate of individual nodes. Many PI-based method suffer from misclassification and false alarm. These methods are also highly complex and time consuming, hence not suitable for on-line applications.

2.1.1.2 ANN-based Methods

Conventionally, the method to contingency analysis and classification is carried out on the basis of the Performance Indices (PIs), which acquired by solving the load flow equations. The speed and accuracy of the security assessment method rely on the sort of methodology employed for the ranking approach. Thus, in past decade, the

literature has disclosed the implementation of the Artificial Neural Networks (ANN) to the static security assessment of the power system [43–75, 149–151]. The authors discuss the importance and applicability of the neural networks for the assessment and control of power system security [63, 84, 150, 151]. ANN’s computing speed and generalization capacity makes it feasible for the smart power systems for the on-line security monitoring [152]. Swarup et al. [43] proposed a 3-layer perceptron network with back propagation learning technique for line flow and voltage contin- gency screening. The ranking and screening modules are composed of Feed-Forward Neural Network (FFNN) [44–46, 153]. The authors in [44, 46] have investigated a CNN, in which the filter and the ranking module are incorporated with a forward network, for quick line flow contingency screening and ranking.

The authors in [47–51,153] have investigate the methods, to estimate the security level for a pre-simulated contingencies data set using Radial Basis Function (RBF) network . A back propagation trained multi perceptron for power system contin- gency screening and static security assessment has been used in [53–59]. For line flow contingency ranking, a 3-layer perceptron ANN has been designed using back propagation learning technique in [52]. A method based on two-phase optimization neural network has been presented in [66] to compute the degree of insecurity and the voltage and angle at all the buses of the system corresponding to closest secure point.

In [67], ANN is being applied for ranking of critical contingencies. Application of Multilayer Perceptron (MLP), RBF Networks and Self-Organizing Feature Map (SOFM) is proposed in [68]. Chow et al. proposed a Hopfield model to solve the contingency classification problem in [149]. This method has demonstrated its feasibility in test cases, yet its assessment accuracy is highly dependent on the amount of training data. Application of Support Vector Machine (SVM) for power system static security assessment is proposed in [69–71, 73, 154]. However, despite their prominent properties, they are unsuitable for large data sets. Devarajet al.[74]

developed a set of FFNN to estimate the voltage stability level at different load conditions for the selected contingencies. Verma et al. [153] developed PIs based cotingency selection and ranking approach employing FFNN with different type of feature selection techniques. The use of MLP-ANN for contingency analysis, screening, ranking considering dynamic security has been investigated in [75–80].

2.1.1.3 Other AI-based and hybrid methods

Many Artificial Intelligence approaches and hybridization of different methods have also been explored by different authors to implement assessment problem more prac- tical by mitigating the limitations of previously discussed methods. For fast voltage contingency ranking, of the most severe contingency for online applications in Energy Management System (EMS), model tree and hybrid decision tree based approaches were used in [81,82]. A Decision Tree (DT) for real time static security assessment is proposed in [83]. A hybrid model of ANN and Fast Fourier Transform (FFT) is used in [84] for contingency screening. Fuzzified multilayer perceptron network has been proposed in [85–87, 90] for voltage security based contingency analysis and ranking.

Sobajicet al.[155] has developed a rule-based method for assessment of both single and multiple line-outages contingencies. A genetic based ANN for static security assessment has been proposed in [93]. An approach of using query-based learning in neural networks has been proposed in [94]. In [156], Particle Swarm Optimization (PSO) based method has been proposed for classification of power system security states.