Short-term Power Load Forecasting Based On The Particle Swarm Optimization Algorithm

The power load forecasting is ,not only the most important foundation for the economic operation as well as security dispatching of power system ,but also a strong guarantee to the normal production of the society and a lower cost of power generation. The neural networks forecasting model is one of the most normal models used in short-term power load forecasting. On base of improving the shortcomings of Neural Networks Forecasting using the traditional BP algorithm, this paper aims to introduce a new model in short-term power load forecasting. The main study in this paper is as follows.(1) In view of the shortcomings of Neural Networks Forecasting using the traditional BP algorithm, including slow convergence and easy tendency to partial optimization, the particle swarm optimization algorithm with a faster convergence and a stronger global search capability is introduced to replace the BP algorithm to optimize the weight and threshold of neural network. Therefore, the Neural Networks Forecasting model based on the particle swarm optimization algorithm is built. In addition, by analyzing the relationship between temperature and load, the hourly temperature factors are introduced into the load forecasting in summer. Furthermore, the advantages and the effectiveness of the new model have been proved by taking experiment simulation somewhere in autumn and summer.(2) The standard particle swarm optimization algorithm can't converge to the global optimal solution and gets into partial optimization too easy; therefore, a hybrid algorithm is built by introducing the idea of sudden jump in simulated annealing, during the searching process of the particle swarm optimization algorithm. This algorithm not only inherits the advantages of the particle swarm optimization algorithm but also has the capability of probability in sudden jump. Thus, it can effectively jump out of local optimum. Furthermore, this algorithm is used to build a new neural network forecasting model based on hybrid algorithm instead of BP algorithm. And the effectiveness of the new forecasting model as well as the advantages of the hybrid algorithm has been proved by forecasting the power load somewhere in China.(3) Through optimal decomposition of wavelet packets, the complicated power load series is divided into several simple subsequences with stronger regularity. Then, respective forecasting is conducted upon all the subsequences, and the hourly temperature factors are introduced into the summer load forecasting of low-frequency subsequences. Finally, forecasting result is got by reconstructing the power load series, and the prediction precision can be improved. As all above, by practical application in a district, the feasibility of the method has been proved.