Research On Forecast And Seasonally Adjusted Model For Electricity Demand Based On Time Series Analysis

Forecasting of the electricity demand can provide a scientific basis for thegovernment to formulate development strategy and electricity generation indicators,which will help promote the development of the national economy with sustained,healthy and rapid status.The innovation of this paper is as follows: Firstly, using a new method based onseasonal exponential smoothing model and the average estimation error to fill in themissing electricity generated series; Secondly,conducting the Box-Cox transformationto eliminate the non-stable variance before modeling the electricity generated series,and then considering the converted sequence as the sample series for modeling.Thirdly, using the weighted combination of the seasonal exponential smoothing modeland the seasonal autoregressive moving average model to predict the future electricitydemand and regarding the original and the predicted sequence as the object ofseasonally adjusted; Fourth, using the seasonally adjusted method containing thethree-stage Spring Festival model to calculate the chain growth rate indicators of theelectricity demand.The main work can be divided into two parts: building forecast model andseasonally adjusted model for electricity demand series. In the process of buildingforecast model, the first step is to fill in the missing data of electricity demand seriesand conduct variance stability transformation of the complement sequence, then useadditive and multiplicative seasonal exponential smoothing models and the seasonalautoregressive moving average model to fit the sample sequence, and select the betterseasonal exponential smoothing model and the optimal seasonal autoregressivemoving average model based on the fitting and short-term prediction error indicators,and regard the mean of the predictive data as the electricity demand for the next twoyears using these two models; In the process of building seasonally adjusted model,the three-stage Spring Festival model will be used to removed the effects of theSpring Festival, and then use the seasonally adjusted method based on movingaverage principle to separate the trend-cycle component, the seasonal factors and theirregular component, and calculate and compare the chain growth rate of theseasonally adjusted series with the year-on-year growth rate of discount month for theunadjusted series finally.The combined time series model can predict the future China’s electricity demandand confidence interval of the next two years accurately, which is useful tomacroscopic the trend of the future electricity demand. The chain growth rateobtained by seasonally adjusted model is more sensitive than the year-on-year growthrate of discount month for the unadjusted series, with which we can capture the inflection points of the past and the future electricity demand. The methods used topredict and adjust the electricity demand are also applicable to other seasonal timeseries.