Short-term power load forecasting based on IVL-BP neural network technology

Accurate forecasting of power load has been one of the important issues in the electricity industry. Recently, along with the privatization and the deregulation, accurate forecasting of power load draws more and more attentions. There are many difficulties in the application of BP neural network which is a very useful tool for the forecasting, such as the defining for the network structure and the local solution which is easy to fall into. To solve these problems, the back-propagation (BP) neural network short-term load forecasting method based on improved variable learning rate back propagation (IVL-BP) is presented in this paper. Though introducing two threshold parameters for the amount of the mean square increasing and decreasing, the learning algorithm is sensitive to the error and convergence speed. Then use genetic algorithm to train network parameters until the error tending to some stable value. Then conduct BP algorithm with the optimized weights to achieve short-term load forecasting. The experimental results have shown that the load forecasting system based on this method has higher accuracy and real-time.