| Hi-index | 0.00 |
In the emerging restructured power system, the congestion management (CM) has become extremely important in order to ensure the security and reliability of the system. In addition to this, lack of CM can impose a hindrance in electricity trading. This paper presents a novel, growing radial basis function neural network (GRBFNN)-based approach for CM. For achieving CM, Nodal congestion price (NCP) forecasting is performed in real time competitive power market. NCP forecasting is an effective way of price-based preventive CM as it directly indicates the presence as well as the severity of the congestion in the system. In present paper, GRBFNN has been developed for NCP forecasting dividing the whole power system into various congestion zones. An unsupervised learning vector quantization (VQ) clustering algorithm is applied as feature selection technique for the developed GRBFNN and for partitioning the power system into different congestion zones. For each congestion zone a separate neural network has been developed to ensure faster training and accurate forecasting results. The proposed approach of CM is implemented on an RTS 24-bus system. The results obtained are compared with a different constructive algorithm-based RBF network called as general regression neural network (GRNN) and two back-propagation algorithms based ANNs. Comparison results show that proposed GRBFNN is more computationally efficient with better predictive ability.