Abstract:

For the combination forecasting of the medium and long term electricity consumption,there exists a difficultyin selecting weights as well as the problems of poor adaptability and low noise immunity.Aiming at the aboveissues,an odds-matrix combination forecasting method based on the odds-matrix algorithm is proposed by taking intoaccount the widely-used combination forecasting technique.In this method,the odds-matrix algorithm is employedto carry out the quantitative analysis on the validity of single forecasting models,and the probability distributionfunction of the weight is adopted to describe the advantages and disadvantages of each method.Then,singleforecasting models are selected and combined based on the weight.Finally,the proposed method is tested by usingthe actual data.The results show that the proposed method achieves a higher accuracy in comparison with othercommonly-used combination forecasting methods,which means that it possesses stronger adaptability and betternoise immunity.

Key words: electric load forecasting, odds-matrix algorithm, combination forecasting, model selection