基于混合改变惯性因子PSO-BP的短期电力负荷预测

针对短期电力负荷预测的精度和网络的收敛问题,通过分析BP、PSO固有缺点,采用周期改变惯性因子(PCW )和动态改变惯性因子(DCW)的双策略,同时对传统的流程增加了额外BP局部寻优,编制了基于MATLAB的混合改变惯性因子PSO-BP神经网络算法(PDPSO-BP),并对广东某城市短期负荷进行预测。结果表明,PDPSO-BP有效地改善了BP的泛化能力,PSO的搜索能力,整体加快了网络的收敛速度,提高了预测的精度,保持误差在3%以下,具有良好的预测效果,满足负荷预测的要求。

Short-term load forecast based on PSO-BP of mixed changing weight

Aiming at the problem of the precision of short- term load forecast and the convergence ofthe network , by analyzing inherent shortcomings of BP and PSO , the double strategies of periodicallychanging weight (PCW ) and dynamically changing weight (DCW ) were adopted , simultaneouslyadditional local optimization of BP was added in the traditional process, the PSO- BP neural networkalgorithm with mixed changing weight ( PDPSO- BP) was compiled based on MATLAB, and the short-term load forecast was made in a city of Guangdong. The results show that the PDPSO- BP effectivelyimproves the generalization ability of BP and the search ability of PSO, accelerates the convergencespeed of the network, improves the accuracy of prediction, and keeps the error below 3% , which hasbetter prediction effect and can satisfy the demands of load forecast.