Urban Residential Load Combined Forecast Model Based on Data Mining Techniques and Panel Data Theory

As a result of rapid urban economic development and improvement in living standards, urban residential electricity consumption in China is increasing quickly. Although the factors which influence the urban residential load are complex, an objective analysis followed by the setting up of a logical urban residential load forecast model can offer a scientific basis for decisions regarding urban power planning and demand-side management. Firstly, based on data mining techniques, association rules mining of residential load were carried out on the relevant data for nine typical cities in China during the period 1992-2006 and the primary factors influencing the residential load were obtained, which avoids the forecast error coming from a subjective choice of factors. Next, the urban residential load was analyzed based on these factors and an urban residential load forecast model was set up based on panel data theory. The model considered not only the time effect of data but also the cross section effect, which can overcome the limitation of data deficiency. Finally, the combined forecast model is proved efficient and reliable in urban residential load forecast by comparing the forecast errors of this model with those of other models based on case studies of typical cities in China.