Short-term Daily Load Forecasting in Distribution Power System Based on VMD & BA-LSSVM Combined Method

Short-term load forecasting is an important basis for optimal control of power distribution network, which directly affects the reliability and economy of power network operation. In this paper, variational mode decomposition (VMD) is used to decompose hourly load sequence of a day to four subsequences with specific frequency scales. The previous 14-day loads are processed with VMD and their results are combined into four matrices. Four matrices and environmental datas which include five factors are normalized and put into four different LSSVMs for the respective forecast. Different parameters in LSSVMs are optimized by bat algorithm (BA). Four LSSVM results are summed and the final prediction result is yielded. Simulation verifies the effectiveness of this method.