Study On Short-term Load Forecasting Based On Particle Swarm Optimization BP Neural Network

The level of load forecasting is one of the measures of modernization of power system management. Load forecasting, especially accurate short-term load forecasting is of great importance to power system. There are many factors that affect system load, such as history data of load, many non-load factors in which weather factor is the most important.The outlier identification is divided into two sequential parts: the robust day-load-curves cluster and the bad curve pattern classification. By analyzing the effects of Kohonen network clustering and BP network classification, the dissertation designs an outlier identification model comprising these two kinds of neural network and implements the tasks of bad data identifications and adjustments.This thesis analyzes every kind of factor which impacts load, and set up the load forecasting model that many factors are considered. In its input features, the load characteristic of near days and every kind of weather factors that considered. Then we unify the input variables, quantify the temperature, rain and shine on etc. To avoid entering into local minimum point for improper selection of initial parameters value of forward-back neural network, particle swarm optimization algorithm is introduced to determine initial parameters value of network , a short-term load forecasting method of power system based on PSO-BP is presented. Compared with traditional neural network ,the method presented in this thesis can quicken the learning speed of the network and improve the predicting precision. In this method, PSO is used to optimize connection weights of forward-back neural network until the learning error has tended to stability, Then we use BP algorithm with optimized weights to finish short-term load forecasting process. Experimental results show that can quicken the learning speed of the network and improve the predicting precision.