Research On Short Term Load Forecasting Methods Of Regional Grid

The Short-Term Load Forecasting(STLF)is a daily routine in the operations of power system. Accurate STLF plays an important role for planning, economical scheduling and security analysis in production, which directly influences the profit of the electric utility enterprises. Since the complexity of power system load and its further complicated trend, traditional load forecasting techniques more and more difficult to predict the power sector to meet the precision requirements, the actual work requirements to develop precision and intelligence with the highest predicted Level of new forecasting techniques.Traditional Load forecast method hadmany deficiencies such as Poor Precision and crudeness and in caPability for non-linear relations; With the artificial neural network (ANN) method introduced in the load forecast and the shortage of the traditional methods in dealing with nonlinear problems was effectively overcome, new vitality was injected to the load forecast. Among these neural network model, RBF neural network approximation has been used in this paper because of its good overall performance, the simple training method and does not fall into local optimal value of the defect.Correctly understand and analyze the impact of related factors on the load and to reflect its inherent laws of quantitative has been a key issue of short-term load forecasting, in particular, appears in the prediction uncertainty of the load factor for some values change, so can not grasp the load forecasting model changes in trends in the load, thereby reducing prediction accuracy. Research on the impact of load factor include: selection of load factors, load factors in the analysis of uncertainty, uncertainty analysis of load factors and less data load processing. In this paper, take advantage of weather and load on the basis of historical data on the impact of load-depth analysis of key factors, combined with the structural characteristics of RBF neural networks, scientific and reasonable to set the input parameters related to the node, and strive to maximize the impact of factors and accurately reflect mapping between the load and achieve the best configuration of neural network structure optimization.This Paper has adopts Particle Swarm Optimization Algorithm to train the neural networks which is a kind of global Optimization technology; According to the main factors to consider date type, temperature, weather conditions and other factors that affect the load forecast, the establishment of the PSO based on particle swarm optimization and radial basis function RBF neural network short term load forecasting model, using particle swarm algorithm to RBF Neural network training to optimize and improve the credibility and reliability of the model.By the actual load forecasting tests of Heyuan, Guangdong Power Grid showed that the prediction accuracy is satisfactory, which proves the effectiveness and practicality of this approach.