Power Demand Forecast Based on Optimized Neural Networks by Improved Genetic Algorithm

Power demand forecast is the basis for making power development plan. Through analyzing the factors, which affect power demand, one model for forecasting power demand has been established, and its data are standardized firstly. Then by designing the structure of BP neural networks and applying the improved genetic algorithm, the network structure and weights of neural networks for power demand are optimized. Finally through training the data from 1980 to 2004 in China, a non-linear relation model between power demand and its influential factors is obtained. The method avoids the shortcomings such as the slow speed of obtaining the optimal solution by genetic algorithm and easily trapping into local optimal solution by the neural networks. The result shows that the method is accurate and feasible