The short-term electric load forecasting grid model based on MDRBR algorithm

Abstract

In load forecasting of a bulk power system, the geographical scope of forecasting region is large, the main electrical load effecting factors in sub districts are different greatly. So it is very significance to establish different load forecasting model according to self-feature of each sub district of a large area, by which the forecasting load is closer to the fact load than establishing one model in whole to forecast electrical load. This paper presents a grid model in terms of geographical division for short-term load forecasting in a bulk power system. The subset model in each geographical grid, considering its own historical loads and meteorological conditions, is more effective and could lead to more accurate results. Therefore, every subnet model is established based on the mining default rules on rough sets (MDRBR) algorithm. First, the MDRBR algorithm is discussed, and the constructing process of the multi-layered rule-network of daily load forecasting is then analyzed in detail. Furthermore, the whole process of load forecasting based on the MDRBR algorithm is presented. Finally, an example using actual historical data shows that the grid forecasting model can yield high accurate results, reduce noises effectively, and is efficient in computation and rule searching.