Abstract :
To make decision for power industry development, it is important to known changes of power demand cycle. Firstly ARMA model and its modeling process of time series were introduced, then according to autocorrelation and partial-autocorrelation coefficients of power demand growth rate from year 1980 to year 2005,AR (2) model was chosen to fit the time series of power demand in China. The maximum likelihood method was used to estimate the value of model parameter, the model and parameters were tested by significance test, and the fitting accuracy was analyzed by errors between actual and forecasting value. At last the growth rate of power demand and year 2006-2020 power demand cycle turning points in China were forecasted. The error average of the growth rate of power demand in China between actual and forecasting value is 0.1417, and the mean absolute error of the forecasting is 1.6253, the mean absolute error rate is 23.5%, year 2008 and year 2012 are power demand cycle turning points. The results show that it is a better method using ARMA model to forecast power demand cycle turning points, fitting model is remarkable, the method is reliable, the forecasting precision is high.
Keywords :
autoregressive moving average processes; correlation methods; decision making; electricity supply industry; load forecasting; maximum likelihood estimation; power system economics; time series; ARMA model; autoregressive moving average process; decision making; maximum likelihood estimation; parameter estimation; partial-autocorrelation coefficient; power demand cycle turning point forecasting; power industry development; time series; Autocorrelation; Data mining; Demand forecasting; Error analysis; Maximum likelihood estimation; Power demand; Power system modeling; Predictive models; Testing; Turning; ARMA; power demand cycle; turning points;
