Application of improved adding-weight one-rank local-region method in electric power system short-term load forecasting

Adding-weight one-rank local-region method makes too many computations and cumulative errors while carrying out multi-step predictions, an improved adding-weight one-rank local-region forecasting model is presented in this paper. According to the prediction effectiveness of Euclid distance between two points away from prediction point in phase space, and synthetically taking into account the effect of distance and degree of incidence between nearest neighbor points and prediction point, an improved prediction is maken with weighted evolution of the neighbor points historically and the evolution of the center reference point to forecast next point directly. The results show that the improved model for short-term load not only reduce forecasting error, but also improve calculation speed. It is a novel prediction method for chaotic time series, and worth to be studied deeply.