Load Characteristics Analysis And Forecasting Research Of Power System

With the implementation of energy conservation policy and economic policy, as well as the implementation of the new holiday system, power system load characteristics have changed significantly, which brings the difficult in load forecasting. Therefore it is significant to study the load characteristic and forecasting deeply.In order to describe the changes of load characteristics correctly, a number of specific load curves and indictors are used. The common indictors are presented in this paper. Based on the investigation of the load characteristics of a province, this paper analyzes the load characteriatics by the curve and the load indictors from the load curves of year, month, typical day and typical industries deeply and draws some conclusions. The main factors impact load characteristics indictors are summarized.Power system load composition refers to analyze the classification of customer load, the proportion and the relationship between each other. It has great significance on load forecasting, load management and load modeling. The four commonly clustering methods are described: hierarchical clustering, dynamic clustering, fuzzy c-means, grey clustering. A solution model of load composition based on clustering ananlysis and least squares method is established.Meteorological conditions, especially the temperature, have been the mian factors that impact the short-term load demand. The correlation between temperature and electricity load is studied by correlation coefficient, scatter plot and regression analysis. Based on the historical load and temperature data, the impacts of the high temperature accumulation effect on power demand are analyzed in this paper. An adjusted model considering the daily highest temperature is established. The adjusted formulation could correctly reflect the impact of temperature accumulation on maximum demand. It can be easily interfaced with currently load analyzing and forecasting methods. Several methods are proposed to calculate the yearly maximum cooling load. They are verified by examples and the results show that the methods are feasible.There are many load forecasting methods based on artificial intelligence and also based on statistical methods. Electricity load can be seen as a series of random sequences, their characteristics can be analyzed by time series. In this paper, autoregressive moving average model(ARMA) is used to forecast the power load. Case studies are served to demonstrate the model's accuracy.