Study Of Mid-long Term Load Forecasting

With the development of China’s power industry, the traditional managementmode and management system can no longer meet the needs and the use of efficientelectric power marketing decision support system become more and more important.The scientific forecast is the foundation and premise of correct decisions. The moduleof load forecasting is an essential part of electric power marketing decision-making,which has the important support function. In the module, medium and long term loadforecasting is the basis of power system planning and construction. The accuracy ofthe mid-long term load prediction will directly affect the regular operation, theinvestment of company and layout of power grid.First of all, this paper introduce the classification and characteristics of electricityload.Secondly, with the understanding of the factors of load variation and associatedwith the commonly used prediction methods,4suitable for mid-long term loadforecasting models are suggested, they are based on seasonal exponent gray loadforecasting model(SE-GM), divided autoregressive moving average model loadforecasting based on seasonal exponent (SE-ARIMA),and neural network loadforecasting model based on seasonal exponent (SESI-ANN).Then, with empirical experiments, taking the electric power monthly data fromthe2006March to2011February of a power company located in Guangdong as asample, the data from2011March to2012February a total of12months as aconductor, the results shows that with seasonal factors included, above all model forprediction of average accurate rate rose respectively9.1%,1.2%,2.and4.1%;Finally, a kind of combinational model for predicting electric power is putforward, mean precision of the optimized model reached97.6%and the mean squareerror is reduced to0.0007, that become the optimization model and be applied to inthe decision support system of a grid.