• Results clearly suggest that the proposed IGWO-TSSCOPF method produces similar results in most of the trials and robustness of the proposed method is thus proved.
Conclusions and Future Scope
[In this chapter the major research findings of the work done by the author are summarized and suggestions have also been made to extend the current research work.]
T
he objective of this chapter is to summarize the main contributions and findings of the work carried out in this thesis and to suggest scope for future research work in this area. In the deregulated and competitive scenario of power system, the operation of modern power systems has become complicated. This thesis focuses on the most important aspects related to power system stability, particularly security assessment, dynamic stability assessment and their enhancement.Chapter 3 of this thesis has presented an improved version of Grey Wolf Optimizer (GWO), named as Intelligent Grey Wolf Optimizer (IGWO). In this chapter an ef- ficient sinusoidal function has been employed to improve the bridging mechanism between the exploration and exploitation phase of GWO. Exploration and exploita- tion capabilities of GWO are enhanced with this newly developed mechanism. Fur- ther, opposition based learning concept has been employed in initialization phase of the GWO along with this sinusoidal bridging mechanism. The combined effect of these two modifications is positive and the implication of these modifications can be observed through the results on various benchmarking functions. The performance of the proposed variant has been validated on standard 22 benchmark functions of different properties and nature. The effectiveness of the proposed algorithm IGWO
165
is throughly investigated and presented. The conclusions drawn from this work are summarized below.
1. There is a scope of improvement in existing meta-heuristic techniques for many complex optimization problems. The developed IGWO perform better than its previous model and the existing meta-heuristic models for the standard 22 bench- mark functions.
2. The modifications suggested in order to improved GWO algorithm are effectively contributing towards enhancing the convergence, accuracy and efficiency of the algorithm.
3. Proposed opposition based modification, disperses tentative solutions near the promising region so virtually reduces search space of meta-heuristics. This feature makes the algorithms more efficient.
4. The obtained results reveals that the proposed variant IGWO shows promising results on majority of the benchmark functions. The superiority of this variant has been validated by optimal values of standard deviation, mean and p-value less than the significance level.
In Chapter 4 of this thesis the application of ANN approaches for online static se- curity assessment and contingency analysis of power systems has been investigated.
Attempt has been made to employ supervised learning architecture for ranking and classification of the probable contingencies. In supervised learning approaches fea- ture selection process is very critical because these approaches are employed for online applications. By considering this fact meta-heuristic based distinct feature selection approach has been proposed and the application results are presented for two systems namely IEEE 30-bus system and IEEE 39-bus system. Two different ANN-based architectures i.e. Radial Basis Function Neural Network (RBFNN) and Feed Forward Neural Network (FFNN) have been investigated for online contin- gency classification. The RBFNN and FFNN have been trained using some selected features offered by the proposed IGWO based feature selection method. A large number of load patterns have been generated by randomly perturbing the real and reactive loads on all the buses to generate a data set that is representative of all pos- sible operating conditions. For each operating condition, a contingency is simulated.
Two Performance Indices (PIs) asP IM V A and P IV Q have been used to classify the contingency into different critical and non-critical classes. A supervised learning approach to fast and accurate power system security assessment and contingency analysis has been proposed in this chapter. Work has been done to improve the classification accuracy and reduce the computational complexity by incorporating the IGWO based feature selection method. The suitability of the proposed model for contingency analysis as a decision making tool for online application at the EMS has been investigated. Following conclusions are drawn from this chapter:
1. For feature selection task, new version of GWO algorithm named as IGWO has been proposed and employed for performing feature selection task. It has been ob- served that the proposed IGWO based feature selection technique shows promis- ing results, when it is compared with previously published approaches.
2. The results of comparative study for two power systems show that the proposed RBFNN yields results with higher accuracy in comparison to the existing con- temporary approaches.
3. The overall accuracy of the test results for unseen samples highlights the ca- pability and the suitability of the proposed approach for online application at EMS.
4. The proposed method is also capable of contingency ranking under uncertain loading condition.
5. The proposed RBFNN method gives excellent accuracy (more than 99%) for con- tingency classification even with a very small feature subset. Therefore, proposed RBFNN-based method may serve as a promising tool for online contingency clas- sification.
Chapter 5, of the thesis is devoted for online assessment of transient stability of power system. A supervised learning algorithm is investigated for the stability iden- tification which is fast and accurate. In this chapter, a new index named Transient Stability Index (TSI) is proposed for identification of the stability status of individ- ual generator in term of its synchronism. The proposed index is based on the Time Domain Simulation (TDS) of the swing equation. An application of the RBFNN
has been investigated for the online identification of generator criticality and TSA of the system for set of probable contingencies. The RBFNN is trained by taking the wide ranging dataset consisting of the randomly varied real and reactive loads at all the buses. The TSI values of all the generators and “Out of Step” time of system are obtained from the trained RBFNN for the unseen operating conditions. The effectiveness of the proposed scheme for online TSA is tested on IEEE 10-generator 39-bus New England system, IEEE 68-bus 16 generator and IEEE 50-generator 145-bus power systems at different loading levels with random perturbations. Thus the proposed unified TSA scheme can provide vital solution to identify the gener- ator criticality and instability problem to the operator at EMS. From this chapter following conclusions can be drawn:
1. A fast and highly accurate RBFNN based transient stability assessment of power system has been carried out for different size small to large power systems con- sidering different operating conditions for the probable contingencies.
2. The online prediction of transient instability within a few cycles from Fault Clear- ing Time (FCT) is determined.
3. Proposed TSI has been employed to do the assessment of system stability, to rank the generators as per the criticality and for identification of the candidate gener- ators for application of the control actions such as coherent group identification and generator rescheduling.
4. The proposed method is based on the prediction of TSI values of each generator for unseen operating condition using RBFNN for a given contingency and through the predicted value of TSI the stability of the system is assessed.
5. For all the unseen operating cases the average error in predicting the TSI using RBFNN is very less and the proposed RBFNN is able to classify stability rapidly with high accuracy.
6. Both informations about ranking obtained from the TSI value of each generat- ing unit and TSI based coherent group information are important for selecting the units for change the generation during generation rescheduling as preventive control action.
7. The proposed RBFNN based method provides better results than the existing methods and they are independent of minor changes in topology of power systems.
In Chapter 6, the applications of the proposed IGWO is exhibited in the area of optimal power flow with respect to both system security constraints and transient stability constraints. An OPF problem has been developed as a constrained opti- mization problem by incorporating different stability constraints i.e. transmission, generation and stability constrains. Penalty factor based approach has been em- ployed in order to enforce all inequality constraints. The solutions obtained by the proposed IGWO-TSSCOPF have been compared with those obtained by other heuristic methods in the literature. The validation of the performance has been carried out with the help of nonlinear TDS. The application results have been pre- sented for two systems as IEEE 30-bus system and IEEE 39-bus system. It has been observed that, the performance of this proposed method is promising. Based on this chapter, following conclusions may be drawn:
1. Penalty factor based objective function reduces handling of constraints and make the problem simple by considering all inequality constraints.
2. Proposed opposition based modification, disperses tentative solutions near the promising region so virtually reduces the problem search space of meta-heuristics.
This feature makes the algorithm more efficient.
3. The statistical analysis reveals that TSSCOPF using proposed IGWO algorithm is improved to a good degree of optima searching ability.
4. The solution obtained by the proposed IGWO-TSSCOPF have been compared with those obtained by other heuristic methods in the literature. It has been observed that, the performance of this proposed method is promising. The vali- dation of the performance has been carried out with the help of nonlinear TDS.
5. The application results reveal that proposed method effectively reduces the gen- eration cost by considering the stability and security of the power system.
Salient Contributions
Major contributions of the thesis may be summarized as below:
1. Developed the improved version of the Grey Wolf Optimizer (GWO) named In- telligent Grey Wolf Optimizer (IGWO) to solve large-scale complex optimization problems and tested over standard 22 benchmark functions.
2. Explored two applications of the IGWO to solve dimension reduction (feature selection) problem for supervised learning engine and Transient Stability and Security Constrained Optimal Fower Flow (TSSCOPF) problem for enhancing the stability and security of the power system.
3. Proposed appropriate modification in scalar Performance Indices (PIs) for con- tingency analysis and static security assessment.
4. Developed a new Transient Stability Index (TSI) to identify the stability status of each generator in terms of their synchronism.
5. Investigated different ANN architectures for fast and accurate contingency anal- ysis, online static security assessment and online transient stability assessment capable of performing well under uncertain loading condition while handling the contingencies.
6. Proposed a unified approach for contingency analysis, online static security assess- ment, online transient stability assessment and security & stability enhancement of power systems.
Future Research Scope
Now a days power systems are very complex and having a large number of compo- nents at all the levels i.e. generation, transmission and distribution. So a unified scheme is required for monitoring and control of system against the power system instabilities. The online assessment of the power system stability is a key issue for the operator. This thesis attempts to address major issues like contingency anal- ysis, static security assessment, transient stability assessment and TSSCOPF as a
method of enhancement of transient stability and security of power systems. The following are suggestions for future research direction in these areas:
1. The present research work is focused around the conventional generation re- sources. However, the problem may be extended with the inclusion of renewable energy resources with the conventional generation resources.
2. In this work, only ANN based supervised learning engines have been employed.
This may be extended to compare various other computational intelligence meth- ods like Deep Learning Programming. This may lead to a new solution for more accurate and fast assessment of power system stability.
3. In this work, focus is on the assessment part of the stability alone. Therefore the methods of preventive control and emergency control can be investigated.
4. This work can be extended for voltage stability assessment and control, and the preventive and emergency control for voltage instability can also be investigated.