Research On Improved Method For Short-term Load Forecasting

At present,unsatisfied prediction precision and mathematization of methods are two important problems in the short-term load forecasting.The complex algorithms are suitable for theory study,but hard to be applied to the practice.Focusing on the demand of projects,several methods are brought up based on the characteristic analysis of load and methods.Different load forecasting methods are used according to day type.For the normal day load,the combination forecasting method of Second order Self-adaptive Method and Inverted Index Model is proposed.For Holiday load,the weighted growth factor point by point method is employed.As inflection point forecast is difficult,the inflection point forecast based on local similarity as a new idea is put forward.The practice shows that the prediction accuracy can be improved by the improvements mentioned above which are easy to implement,quick to calculate and have a very good value in engineering practice.