Short-term nodal congestion price forecasting in a large-scale power market using ANN with genetic optimization training

Authors: MAJID MOAZZAMI, RAHMAT ALLAH HOOSHMAND

Abstract: In a daily power market, price and load forecasting are the most important signals for the market participants. In this paper, an accurate feed-forward neural network model with a genetic optimization Levenberg-Marquardt back propagation training algorithm is employed for short-term nodal congestion price forecasting in different zones of a large-scale power market. The use of genetic algorithms for neural network training optimization has a remarkable effect on the accuracy of price forecasting in a large-scale power market. The necessary data for neural network training are obtained by solving optimal power flow equations that take into account all effective constraints at any hour of the day in a single month. The structure of the neural network has 2 input signals of active and reactive powers for every load busbar in every hour of the programming model. These 2 signals are always available. In this study, an IEEE 118-bus power system is used to test the proposed method authenticity. This system is divided into 3 zones, and a neural network with genetic algorithm training optimization is employed for every zone. Performance of the proposed method is compared with ARIMA and GARCH time series for the same data. The simulation results show that the proposed algorithm is robust, efficient, and accurate. Therefore, the algorithm produces better results than the ARIMA and GARCH time series for short-term nodal congestion price forecasting.

Keywords: Price forecasting, nodal congestion price, artificial neural network, genetic algorithm

Full Text: PDF