Short-term load forecasting using general regression neural network
Dongxiao Niu, Huiqing Wang, Zhihong Gu · 2005
This paper describes an optimal generalized regression neural network (GRNN) in which training of the network is optimization of the smoothing factors. And a modified differential evolution algorithm (MDE) is proposed to improve the searching efficiency of simple differential evolution algorithm (DE). The modified evolution algorithm advanced the performance of global optimizing through collecting population information during evolution and at the same time introducing deterministic operation, amending distribution of individuals adaptively. The eugenic evolution strategies used in this paper include maintaining population diversity, adding new deterministic simplex searching operation, modifying the probability operation, and others. The GRNN-MDE, which is based on DE and provides powerful capacity in non-linear modeling and predicting, is applied to modeling short-term power load forecasting, and the result is satisfied.