On-Line dynamic security assessment of wind farm connected power systems using a class of intelligent algorithms
|dc.contributor.supervisor||Prof. Syed M. Islam|
|dc.contributor.supervisor||Dr Dilan Jayaweera|
Recently large-scale wind farms are integrated quite commonly into power systems. The stochastic operation of wind plants due to intermittency and intra-interval effects of the wind is a problematic issue to determine the amount of wind power to be injected into the system. If the wind power penetration is not within the limit of dynamic security of power system, cascading outage and a shutdown of major portion of power system is likely to occur. The only practical way to avoid the disaster is to alert the system operator in near-real time for preventive actions to be taken in case need arises. Novel approaches for on-line Dynamic Security Assessment (DSA) of wind farm connected power systems is a potential cure to the problem.Power system networks are non-linear with operating conditions that change from time to time. The significant challenge with the design and development of novel algorithms for on-line DSA is its effectiveness and very fast speed of execution over a wide range of operating conditions. Several algorithms like numerical integration, Lyapunov methods, Decision Tree (DT) and Artificial Neural Network (ANN) have been already proposed in power system literature.The numerical integration algorithm method is used to solve a set of first-order differential equations that described the dynamics of the system model. The method is very effective in providing exact answers relating to stability and in handling detail models in an off-line environment. Nevertheless the method is not suitable in an online environment due to excess computational requirement leading to increased processing time.The aim of Lyapunov methods is to moderate the computational load of the numerical integration method through a scalar function. This method is successful with the speed of solution and ability to compute the degree of stability. But the method some limitation since only classical generator model can be used.DT and ANN are two classifier methods implemented using training examples. DT relies on selected optimum attributes for efficient classification of operating points. Even if DT method chooses the best attributes, they are still many undetected operating conditions of which security status cannot be appropriately established. ANN is effective because it uses appropriate input and output variable sets, and large neurons and non-linear interconnection. The drawback of the ANN is due to the gradient descent algorithm training, which can be stuck at a non-optimal local minimum.Recently, D. Ruisheng, Vijay Vittal and N. Logic worked on a hybrid model integrating phasor measurement units and decision tree. The tool assessed four postcontingency security issues, including transient stability. Test results show that properly trained DTs perform very well to assess different security issues in a near real time. But because DT methodology only use linear split, some cases are still misclassified at the boundary range which is between the secure and the insecure region where nonlinear characteristics are exhibited. This makes the above tool not adequate to robustly solve on-line security issues.Case-Based Reasoning (CBR) is a simple general paradigm for reasoning from experience. The application of CBR to the field of dynamic security assessment of power systems is yet to be explored. CBR and DT have some complementary merits which can be used efficiently if they are integrated.This thesis aimed at designing and developing a new class of intelligent algorithms for an on-line dynamic security assessment of wind farm connected power systems. The main objectives of this thesis can be summarized as follow: 1. Design and development of CBR based modelled for an on-line dynamic security assessment of wind farm connected power systems. 2. Defining new hybrid criteria for integrating virtues of CBR and DT techniques to improve the classification accuracy and the speed of execution of operating points for an on-line dynamic security assessment of wind farm connected power systems. 3. Investigate the potential of CBR and hybrid CBR-DT methods for an on-line dynamic security assessment of wind farm connected power systems.Based on research findings, the thesis argues that: 1. CBR method is more efficient than existing techniques and can be successfully applied in real power networks. 2. Hybrid CBR-DT improves the classification accuracy and the speed of execution of operating points considerably.
|dc.title||On-Line dynamic security assessment of wind farm connected power systems using a class of intelligent algorithms|
|curtin.department||Department of Electrical and Computer Engineering|