基于遗传算法优化神经网络的中期负荷预测  

Medium-Term Load Forecasting Based on Genetic Algorithm Optimised Neural Network

作　　者：闫泓全 李梓萍 李校良 孙楚词 金书池 Yan Hongquan;Li Ziping;Li Xiaoliang;Sun Chuyin;Jin Shuchi(Liaoning Technical University,Huludao Liaoning 125000,China)

机构地区：[1]辽宁工程技术大学,辽宁葫芦岛125000

出　　处：《现代工业经济和信息化》2024年第8期171-172,175,共3页Modern Industrial Economy and Informationization

摘　　要：针对某地中期电力负荷预测,BP神经网络因其强大的非线性映射能力被广泛应用于负荷预测。但BP算法本身存在一些缺陷,如易陷入局部最小值和较慢的收敛速度。而遗传算法以其优秀的全局搜索能力可以很好地解决这一问题,利用GA的全局优化能力,对BP网络的初始连接权重和偏置值进行系统的调整,从而显著减少了网络在学习过程中落入次优解的可能性,提升了收敛速度。研究结果表明,相较于单一的BP神经网络和其他对比模型,该集成模型在预测精度和稳定性方面均有显著提升。Medium-term power load forecasting is extremely important for grid planning and generation planning as well as power market operation.For the medium-term power load forecasting in a certain place,BP neural network is widely used in load forecasting because of its powerful nonlinear mapping ability.However,the BP algorithm itself has some defects,such as easy to fall into local minima and slower convergence speed.The genetic algorithm with its excellent global search ability can well solve this problem,using the global optimisation ability of GA to systematically adjust the initial connection weights and bias values of the BP network,which significantly reduces the possibility of the network falling into the suboptimal solution during the learning process.The convergence speed is improved.The results of the study show that the integrated model has significant improvement in both prediction accuracy and stability compared to the single BP neural network and other comparative models.

关 键 词：遗传算法 BP神经网络 负荷预测 优化算法 

分 类 号：TM715[电气工程—电力系统及自动化]