The Practical Method Research For Power Demand Forecasting

Power demand forecasting is an important problem in power maiketing, power system programming and decision-making of power utilities. Aiming at the practical demand of power marketing programming in the author's company, this paper, based on current forecasting techniques and data mining techniques, presents a practical method for power demand forecasting based on the theory of time series and the model of linear regression. The method is validated using the real data of consumed power of various customers during 1999-2004. Then a linear regression model is constructed according to the different features of various customers to forecast the power demand in future 3-5 years, and the forecasted results of 2005 has been used as an important basis to program the work target for superior departments.Many scholars study the theme of load forecasting and power demand forecasting, however because many factors influence the power demand, so currently there is no power demand forecasting software systems used widely. The current research focuses on the forecasting models and theory methods, and is immature for practical application. Considering the practical demand, this paper classifies the power demand into industry customers, agriculture customers, traffic customers, city customers and dynamical customers, analyzes various factors influencing the power demand using time series method as data mining techniques, and finds out the rules of various factors. Then realizes accurate forecasting results based on the regression models.In the practical example, according to the consumed power data from different area and customers during 1999-2004, the paper analyzes many factors influencing the power demand, such as GDP of different industries, economy developing ratio of different industries and the average temperature in history years and months. By the data mining technique, the paper finds out the rules of heavy industry, light industry, commerce power consume, GDP of secondary industry, GDP of third industry and weather condition, by which the power demand is influenced. And the rules of power demand of above various customers are forecasted, the living power consume is forecasted by calculating every month and accumulating them to achieve the years power consume. Finally, the method presented in the paper is compared with the method of Grey forecasting, and the results show that the method is effective and its forecasting precision is better than any other methods applied currently.