Decentralized Wide-Area Neural Network Predictive Damping Controller for a Large-scale Power System

摘要

This paper has presented a design method for the measurement based decentralized wide-area neural network predictive damping controller for Static Var Compensator (SVC) to damp out low-frequency inter-area oscillations in a large-scale interconnected system. Since Model Predictive Control (MPC) is based on the fixed predictive model parameters, Minimal Resource Allocation Network (MRAN) and Extended MRAN (EMRAN) online sequential learning algorithms have been utilized for online identification of plant model parameters at each sampling instant. The main advantages of the proposed approach are a) Online learning based Radial basis function neural network (RBFNN) identifier is better than offline learning based RBFNN identifier for identifying the plant parameters during large disturbance conditions. b) MPC with on-line model identifier captures more effectively the plant dynamics, uncertainty and interactions between the subsystems at each sampling instant. The performance of the proposed wide-area damping control scheme has been tested using the wNAPS 179-bus large-scale system and supported by non-linear time-domain simulation. The Performance of the online learning algorithm based Neural Network Predictive Control (NNPC) has been compared with the offline learning algorithm based NNPC controller. The results show that the damping of inter-area oscillation is more effective under different types of fault conditions and the global stability of the system is preserved.