The Household Energy Consumption Analysis And Prediction Based On Wavelet Analysis

Residents household energy consumption can be reflected the civilization degree of a country.Our country residents household energy consumption has became the second largest energy consumption.Therefore,the study of household energy consumption can helps for the further study of the whole energy consumption system,regulate and control energy consumption model from on macroscopic,improve the relationship between the energy department and other departments.Because of household energy consumption,it is not only affected by the economic system itself,but also by such as cultural background,regions and political structure and some external factors of economic system.Therefore,in order to better grasp the evolution regularity and development trend of household energy consumption,it is not enough to analysis in one time level.Wavelet analysis,that is a method for signal processing in engineering field,can analysis of the non-stationary time series data from on the time scale of several layers.Based on multi-scale decomposition of the data, which can be according to the data characteristics of different scales for effective modeling,and forecasting,and then reconstruct,thus reducing waste of information, improve the efficiency and accuracy of prediction.In view of these considerations,we use household energy consumption as the research object in this paper,we select the date for 1983—2012 of Chinese household energy consumption.By use for reference the application of international and domestic wavelet analysis methods in economic and other field,using the theory of wavelet analysis to analysis and forecast it.First,by the means of decomposition and reconstruction in algorithm of multi-resolution analysis for wavelet theory. To analysis of the characteristic time scale,main cycle and mutation characteristics for household energy consumption,and get some results.In the aspect of characteristic time scales,household energy consumption growth rate has 25 to 30 years and 3 to 10 years these two major characteristic time scales,with 25 to 30 year characteristic time scale is very obvious.In the main cycle,through wavelet variance figure, found the household energy consumption growth rate has four main cycle of 3 years, 6 years, 8 years and 28 years.In the mutation characteristics,through the analysis of the four major cycles underthe mathematical model of wavelet transform,we found the four major cycles of household energy consumption growth rate sequence has 1988,1998 and 2008 these three point mutations.And judge the future household energy consumption for slowing growth trend.Thereafter,we compare with traditional ARMA model and wavelet analysis modeling to predict household energy consumption.Then using the Root Mean Square Error, Mean absolute Error, Mean Absolute Percentage Error and Theil Inequality Coefficient of these four indicators to evaluate result of model prediction.Results shows that the wavelet model established by the prediction is better.