Power System Load Forecasting Based On Wavelet And Neural Network Theory

The load forecasting of power system includes long-term load forecasting,medium-term load forecasting and short-term load forecasting. The precise short-term load forecasting plays a very important role to the scheduling arrangement and operation plan. From the direction of big field, it can improve the power quality of the power system, also it can ensure the safety and reliability of the power system. Moreover, precise short-term load forecasting make it possible for the power system to operate at lower cost.This paper first digs into the whole load characteristics of the power system including the periodicity of one year,one month,one week and one day. Then the relations between temperature and load numerical value is further discussed. Also, the paper does a lot of pretreatment to the training sample. By the way, the paper introduces the neural network theory, after this, the prediction model based on BP learning algorithm is built. It determines the structure of network through a lot of experiments, and it gets satisfactory prediction results.Although the prediction result of error back propagation (BP) algorithm is just ideal, the BP algorithm is easy to fall into the local minimum points and its convergence time is long, so the paper gives the genetic algorithm which is used to optimize the weight and threshold of BP network referring to genetics theory. Code,do genetic operation and iterative optimization constantly to achieve the best weights and threshold. To predict load by BP neural network optimized by genetic algorithm, the prediction load and actual results are more closer. At the end of the article, the paper systematically introduces the wavelet theory, the PSO algorithm and support vector machine theory. Based on these theories, the total forecast system about the wavelet analysis is given. The comprehensive prediction results and prediction error shows that the method gives higher prediction precision and efficiency, it can satisfy the demand of error and requirements what the paper expect.