Forecast of Power System Load in Short Term

In this paper, according to large quantity of historical statistics, we have established a model which could successfully forecast the power charge in two regions. Because of the different efforts between weekdays, weekends and holidays, we made a piecewise function[1] to decrease the error. The method of 2 times curve fitting was used to analyze the electric power charge of minimum and average per day by Matlab. Then an ARIMA (Auto Regressive Integrated Moving Average Model) connected with statistics between 2009 to 2014 was set up and verified available residual analysis. We also take climate factors into consideration. Being supported by huge data the model can predict variation tendency of electric power charge efficiently.