Short-term Electric Load Forecasting Based On Wavelet De-noising And Artificial Neural Network

In this paper, short-term load forecasting method and main factors impacting the forecast are studied. The power system load forecasting means to analyze and forecast the variation trend of future load for a period of time based on the power load historical data and the various factors that affect power load, including economic, social and meteorological factors. Load forecasting is very important to many sectors of the power system. For example, the medium and long term forecast are concerned by planning and schedule department of power system and the short-term forecast is taken as an important working basis for the marketing (electricity consumption), market transactions, scheduling department and so on. The factors that affect the power load forecasting is varied. Overall, it is divided into the internal factors and the external factors. Through the classification of electricity consumption, it is discussed in the second chapter that different influencing factors need to be taken into consideration when are different prediction types are discussed.Based on the wavelet theory, the de-noising of raw load data is proceeded and tit is divided into five layers by the means of Haar wavelet. The soft threshold function is applied to denoise the raw load data as well and the ideal load data is obtained, which lays the foundation for improving the short-term power load forecasting accuracy. The fourth chapter analyzes the overall concept of disposing the meteorological factors in short-term power load forecasting. On the other hand, the intuitive law and the relevant factors between the temperature factors and load are analyzed. It is concluded that the average temperature is the main factor that affects the short-term power load. Finally, the BP neural network and Elman neural network are applied to forecast the short-term power load under different circumstances.Through the wavelet de-noising of the load data, it shows that more satisfactory results can be obtained combined with Elman neural network short-term load forecasting model and taken the average temperature into consideration as well.