Study On Short Term Electricity Loads Forecasting Based On Local Fuzzy Inference

Electric power load forecasting is an important part in the power system and the budget of the electricity user. With the development of the society, the electricity demand and maximum load are growing rapidly. For daily load is concerned, the valley of the peak value differences get bigger in a day and peak points become more, so they bring great difficulty to the power load forecasting, especially for the short-term load forecasting such as week-load or day-load forecasting. The electricity diversity led to daily electricity curve changes bigger, so many of the former models gradually no longer have the advantages in the prediction accuracy, such as regression coefficient, time sequence method etc. Recently the neural network is adopted by much of the load forecast literature, because it has learning ability, but the neural network can’t deal with uncertain problems. However, fuzzy inference has the advantages of processing the uncertain problem, so fuzzy sets and fuzzy reference knowledge is also widely used in electric power load of mathematical modeling.This paper presents the short-term power load forecasting model prediction method based on fuzzy inference. The method first will divided one day24hours into a number of time zones according to the load curve distribution characteristic, in order to guarantee the prediction accuracy in the period of big difference between peak and valley. For each time zone will establish a load forecasting model, and then according to the load data use the similarity search means to find the suitable load attribute as input vector, then use a supervised cluster analysis to get a classic subset load model and then use fuzzy reference to get forecasting results. Because this paper combines the supervised cluster analysis and fuzzy reference method so as to ensure the learning ability and the ability of processing uncertainties. Chapter5practical examples show that the short-term power load forecasting method based on fuzzy reasoning has strong practicability and accuracy.