Study Of Short-term Load Forecasting Based On Wavelet Packet Analysis

With the rapid development of modern society, the quality of Chinese people’s life is improving. Power source has become an irreplaceable important energy in people’s daily life and national economic construction. The sustainable development of the national economy is to promote the rapid growth of electricity demand, resulting in a very rapid development of power industry.Power system load forecasting is an important part of the economical and safety operation of power system, and it has always been an important issue in the study of power system. With the complexity diversification of electricity users, the requirement for the quality of power is stricter, faster rate and better accuracy are required in short-term forecasting. As a tool for frequency analysis, wavelet analysis has developed very rapidly in recent years. It has distinctive advantages in the field of analyzing load data, which contains periodic component and random component. Then, many methods of short-term load forecasting based on wavelet analysis have been creatived.This thesis studies the issue of short-term load forecasting based on wavelet packet analysis. After the load data is composed by wavelet packet analysis, two short-term load forecasting models based on BP neural network and Markov chain are proposed. Using MATALB software to simulate an instance, the two methods are proved feasible. Through the study found that the peak-type Markov chain is superior to the traditional Markov chain. Peak-type Markov chain is applied in Power system short-term load forecasting firstly in this paper. Thus an improved short-term load forecasting algorithm is proposed, which is based on wavelet packet analysis and peak-type Markov chain. Programming by MATLAB software and simulating the same distance, and comparing the predicted results. The prediction accuracy of the improved algorithm is greatly better, so the improved short-term load forecasting algorithm in this thesis is proved correct and effective.