基于改进哈里斯鹰算法优化核极限学习机的短期电网负荷预测研究

Short-term Grid Load Forecasting Based on CEHHO-KELM

利用改进的哈里斯鹰算法对核极限学习机进行优化,构建了CEHHO-KELM电力负荷预测模型。首先,在充分考虑了经济、时间、气候以及电网自身影响的基础上,采用灰色关联分析法筛选主要影响因素作为预测模型输入向量。然后,将优化的哈里斯鹰算法融合到核极限学习机的参数优化中,建立了CEHHO-KELM电力负荷预测模型。将某省电力负荷数据及其影响因素数据用于实证分析。仿真结果表明,CEHHO-KELM算法相较于HHO-KELM、LSSVM、KELM算法,能够较好地搜索核极限学习机的参数、更好地平衡全局和局部性能,从而使得KELM预测模型具有更高的预测精度。

In this paper,the CEHHO-KELM power load forecasting model is constructed by optimizing the Kernel Limit Learning Machine(KELM)with the improved Harris Hawk Algorithm(HHO).Firstly,the main influencing factors are selected as the input vectors of the forecasting model by using gray correlation analysis,taking into account the four influencing factors of economy,time,climate and the characteristics of the grid itself.Then,the optimized HHO is integrated into the parameter optimization of the KELM to establish the CEHHO-KELM electric load forecasting model.The electricity load data of a province and its influencing factor data are used for empirical analysis.The simulation results show that the CEHHO-KELM algorithm can better search the parameters of the nuclear limit learning machine and better balance the global and local performance compared with the HHO-KELM,LSSVM,and KELM algorithms,thus making the KELM prediction model have higher prediction accuracy.