Power load prediction method based on kernel extreme learning machine with t-distribution variation bat algorithm

To improve accuracy of power system load forecasting, according to nonlinearity and uncertainty of load sequence, a power load forecasting method combined with kernel extreme learning machine (KELM) and adaptive variation bat optimization algorithm (AMBA) is proposed in this paper. The model determines mutation probability of the current optimal individual according to the population fitness variance and the value of the current optimal solution and the t-distribution variation is performed on the global optimal individual, and bat individual after mutation is subjected to secondary optimization. Then AMBA is used to optimize the network parameters of kernel function extreme learning machine, so that make full use of advantages of fast learning speed and generalization ability of KELM to realize rapid prediction of load. AMBA-KLEM model is used to predict load of Ningxia Power Grid, and the results are compared with ELM model and KELM model. The results show that the model has higher load prediction accuracy.