Power System Load Forecasting Based On PSO Chaotic Neural Networks

Load forecasting is the basis of power operation and long-term plan. It is widely used in the dispatching and operation planning of power systems, and the accuracy of the load forecasting is helpful to the security, economy of power systems and quality of the power supply. The features of short-term load forecasting can be generalized as followings: many data need to be forecasted, the physical factors which influence forecast are complicated and random, and high precision of forecasting is demanded. With the establishment and development of the power market, STLF has been one of the important symbols for the power system modern management.The primary problem of the load forecasting is the selecting of the forecasting technique, namely, how to build a forecasting model which is applicable to the studied district. The study of STLF has been deepened and various methods have been putted forward, along with the development of the technology. The accuracy has been improved continuously, because of the advanced method from the classical Statistic analysis methods to the modern intelligent methods. However, due to the complexity of the forecasting problem, they have some shortages inevitably.Particle Swarm Optimization (PSO) is an evolutionary computation technique based on the swarm intelligence, which is originated from artificial life and evolutionary computation. In order to fundamentally improve the short-term power load forecasting neural network forecasting accuracy and taking into account the power load characteristics of Chaos. In this paper, a method based on particle swarm optimization (PSO) and chaos neural network(CNN)is presented for short-term load forecasting. The former PSO is used to train connection weights and threshold parameters—layer feed forward neural network until the learning error tends to be stable.Then using the weights and threshold to accomplish load forecasting .The impact of climate and temperature is processed with chaos technique and considered as input data of the network. Experimental results show that the proposed method can quicken the learning speed of the network and improve the predicting precision compared with the traditional artificial neural network. And were made based on the group of elementary particles and improved PSO model predicted that the two results are compared. Improvement PSO draw more suitable for optimizing neural network used for load forecast.