Study On Forecasting Method Of Power Consumption In Urban Power Grid Based On Modeling In Different Dimensions

Forecasting of power consumption in urban power grid is a basic work in power market.An exact forecasting result can not only promote the development of power market, but alsoprovide decision support of the generator output and Transformer operating for power supplyenterprise. Power grid can be more safe and economic operated by an exact Forecasting ofpower consumption.In the power forecasting process, there are three problems need to be resolved. First,because the nonlinear and volatility are the dual characteristics of seasonal powerconsumption data, it is very difficult for a single forecasting model to accurately describe thisnonlinear trend. Thus, the accuracy requirements of Forecasting are hard to meet. Second,seasonal power consumption data not only change in the regular pattern, but also change withrandomness because of the affecting by internal and external factors. The model cannot beaccurately constructed unless the regular pattern and randomness can be reflected by othercorrelate variables in modeling process. Three, the best fitting model is not equal to theoptimal Forecasting model because of the randomness of the seasonal power consumptiondata. So, Forecasting process only depending on the best fitting model may lose some keyinformation and lead to a worse result if other fitting models are abandoned.This paper proposes two class of modeling method to resolve the problem above. One issingle dimension forecasting method based on the changing trend of the power consumptiondata, which contain4models that can reflect the trend and volatility of the seasonal powerconsumption data in different way. The other one is multi-dimension forecasting methodbased on power consumption of different industry and the related factors, which contain2models that remedy the defect of the single dimension forecasting method. Then, allforecasting result of the above models is optimum combined by the variance–covariancemethod. So, all the information of all in the forecasting process is farthest utilized, improvingthe accuracy and reliability of forecasting results.In order to make the method easier to utilize and popularize, a software is developed byusing the MATLAB GUI based on the method mentioned above. Finally, one example inGuang Dong power grid is analyzed through the software, which shows that the result ofmethod proposed is more accurate than any single model. The performance of the methodproposed in this paper is superior and the software developed has very high practical value.