Study and Application of Dynamic Collocation of Variable Weights Combination Forecasting Model

Abstract

Short-term load forecasting plays a very important role in operating, controlling, and planning of power system. As the load forecasting is vulnerable to various environmental factors, the short-term load forecasting is uncertain and variable. The traditional single forecasting model used to forecast the load of power grid can't comply with the requirements of the power grid management. Combination forecasting model can largely make up for the one-sidedness of the single forecasting methods. In the implementation of combination model, the fixed load forecasting methods also make forecasting results inaccurate, and have a series of problems such as low credibility. In this paper, the thought of dynamic combination is applied in the orderly power consumption management platform, and a combined optimal forecasting model is constructed through automatic screening of the forecasting methods and dynamic collocation of weights. Practice has proved that the combined forecasting method has higher forecasting accuracy than the single forecasting method and it is not only has high forecasting accuracy, but also has good extendibility, quick speed of data processing, simplicity of operation and diversity of display mode.