Neural Network Based Power System Load Forecasting

Power system load forecasting plays a vital role on power system economic, security and reliable operation,which has become an important component of the modern energy management system,in which the significance of short-term load forecasting for power system operation and people's daily production and consumption is particularly significant. Power system load forecasting is an important prerequisite for optimal control and regulation of the entire power network. The load forecasting error will lead to the increase in operating and production costs, therefore , the precise power system load forecasting are of practical significance for power system control, operation and planning.This paper analyzes the characteristics and research status of power system load forecasting. Based on the above analysis,in the paper,the model for power system load forecasting is optimized with BP artificial neural network . By this method , it can better describe the multiple-input multiple-output , the complex nonlinear and complex characteristics of uncertainty of power system load. Genetic algorithm is a global search algorithm based on the natural selection law and the genetic law. On the basis of analyzing the characteristics of the GA algorithm and BP neural network , for the BP neural network is vulnerable to the lack of local minimum, the paper leads to a algorithm to train the neural network in use of the combination of GA and BP. Based on the combination of GA and BP, the specific algorithm and implementation process are given in order to overcome the defects of BP neural network.In this paper, the simulation software called matlab is used to optimize the network structure, with a history data of power system load and the maximum temperature, minimum temperature as the network input, with the GA algorithm to optimize the weights of BP neural networks to predict the power system load. This paper uses the simulation software in load of Tianjin power system. Test results show that the model for power system load forecasting is feasible, to a certain extent, which improves accuracy and speed of load forecasting.