Short-term Load Forecasting Based On Fuzzy Clustering Analysis And Neural Network Algebra Algorithm

The short-term load forecasting(STLF) of power system plays an important role in the safely assigning and economically running, also is an important basis of power dispatch, electric purchasing and transmitting plan. The percision of forecasting will directly affect the stability, economic running and the quality of supplying power in power system. As the development of power market, the important of STLF are more and more taken into account. This paper will study and discuss the subject, and present a combined methed for STLF based on fuzzy clustering analysis and neural network algebra algorithm. It fully carries out the characteristics of artificial neural networks, that is, it can approach to any nonlianear function and self-study. Basides, fuzzy clustering analysis is good at dealing with uncertainty,inaccuracy problem. As the electric power system is always influenced by many factors, therefore this system takes the elements of temperature, date type, weather status and so on into the forecasting model, the forecasting results show that the propsed method is more accurate.First of all, the paper interprets the backgroud, significance and development of SLTF. Various algorithms are discussed, also decise the mian content of the paper.Secondly, through the analysis of the basic theory, the factors of influencing the load and the forecasting process, it takes the mian factors such as temperature, date type, weather status into the load forecasting model. It introduces the basic principle of the artificial neural network, also describes the process of study, the deficiencies and the corresponding improvement method of back propagation(BP) algrithm. As the defects of slow learning convergence, easily falling into local minimum and can not determine the number of hidden node before training and so on exists in the BP algrithm, the paper introduces neural network algebra algorithm and discusses the basic theory, the process of study and the advantages of this algrithm, it completely overcomes the shortcoming of the BP algorithm.Thirdly, the paper intrduces fuzzy clustering analysis. It normalizes the historical data, then establishs the fuzzy equivalent matrix in order to cluster the samples into several categories and finds out the category coincident with the daily load to be forecasted. The same samples is uesd as the input of neual network, and trains by the neural network algebra algorithm to forecast hourly load of working day and weekend day.Lastly, the results of BP neural network algorithm, neural network algebra algorithm and the method proposed in this paper are compared. The forecasting results show that the proposed method possesses better forecasting accuracy and the forecasting is satisfactory. It also verifies the validity and practicability of the proposed method.