Chaotic Multi-step Forecasting Algorism Applied in Short-Time Electric Power Load Forecasting

In the chaotic local adding-weight linear method, Euclid distance is used as correlation measurement between phase points. Because Euclid distance just indicates space distance between phase points, the inherent relevant information can not be mined adequately, so that the enhancement of forecasting precision is restricted. The article uses the angle between vectors as phase points' correlation measurement, then in the process of linear regression parameters identification, introduces the vector modulus and the angle between vectors as optimized aims into the least square method. By means of the new algorism, reference neighborhood correlated closely with datum phase point is picked out and better linear regression parameters are identified, so that the disadvantage of traditional algorism based on Euclid distance is overcame. In a forecasting example about power grid data of a Chinese southern city, the algorism of the article achieves good forecasting effect. Especially, the algorism performs well to sudden load change.