On line transient stability assessment (TSA) is an important application in power systems operation. This function must be performed continuously as actual power system operating conditions (demand, topology, generation) change in time. Solution of this problem requires the evaluation of system dynamic behavior following credible contingencies.Activation of discrete supplementary controls (DSC) must be included if a complete solution is desired. In longitudinal power systems, with active constraints in energy transmission, transient stability problems are common between distant generation centers and load areas.
Transient stability is characterized by a fast development, generally of first oscillation type, because of the radial characteristic of the transmission system. System security under these co nditions requires the implementation of adequate preventive measures.Discrete supplementary controls have shown to be effective means to control transient stability problems in longitudinal power systeims, allowing better usage of transmission systems.Several interesting approaches have been proposed in the past to deal with the transient stability assessment problem, showing the importance of this subject. Some of these techniques are: Fast time domain simulation using powerful computers, direct methods, decision trees, expert systems, pattern recognition probabilistic methods, and artificial neural networks (ANN) .To the authors knowledge, the idea to apply ANN to solve the dynamic security assessment problem was first presented by Sobajic and Pao. A good agreement between estimated and actual critical clearing time (CCT) was obtained in a small test power system, considering different operating conditions and topological changes. Results using semi-linear feedforward, and functional link ANN were reported.
Security assessment with a projection based ANN, using steady state stability studies, was proposed by Aggoune et al. A small test system was employed.A hybrid approach was reported by Miranda, et al to perform a power system generation dispatch with security constraints using a conventional optimization approach combined with a feedforward neural network. A backpropagation (BP) algorithm with adaptive learning step was used in the training phase, tests were performed in a CIGRE test system with about 1400 operating points. Chang, et al. proposed a hybrid strategy to evaluate transient stability of medium size power systems.
A pattern recognition method was employed to assess transient stability and security transfer limits between interconnected systems.A feedforward network trained with an adaptive step BP algorithm was used in this application.Fast pattern recognition and classification of dynamic.
security states were reported to be obtained by Zhou et al. Power system vulnerability was the output variable from a feedfonvard network trained with a BP algorithm.
No comments:
Post a Comment