摘要
精确的发电量预测对电网的科学运行和经济效益具有重大意义.针对用于预测的神经网络初始权值与阈值选取不合理、网络收敛性和预测精度差的问题,提出了一种用灰狼优化算法(GWO)对BP网络的初始权值和初始阈值进行优化的算法.首先采用灰色关联度对发电量特征数据进行关联分析,确定出模型的输入层节点数.然后将网络的初始权值和初始阈值对应表示为狼群中每个个体的位置向量,其次引入领导者策略模拟狼群包围、捕食猎物的过程,完成参数优化.最后,根据误差反向传播训练得到最终权值和最终阈值,建立用于短期发电量预测的BP神经网络.实验结果表明,与传统BP和GA-BP算法相比,提出的GWO-BP预测算法的精度分别提高了0.63‰和0.32‰,为解决短期发电量准确预测的问题提供了一种新的方法.
The accurate power generation prediction plays an important role in the safe operation of power generation system.Aiming at the issues of determination on neural network initial weights and thresholds in the prediction model,this paper presents a grey wolf algorithm optimization(GWO)to optimize the initial weights and thresholds of BP network.According to the wolves′hierarchy social structure and the process for prey,individual wolf position vectors are introduced to represent the initial weights and thresholds.The parameter optimization can be completed by introducing the leader policy which can simulate the wolves′process for prey and capture.Finally the final weights and thresholds can be achieved by BP network training.And the GWO-BP model are derived to forecast the short-term power generation prediction.Experimental results show that GWO-BP prediction algorithm can reach more precise results than the traditional BP or GA-BP algorithm with 0.63‰and 0.32‰respectively.And the proposed GWO-BP prediction algorithm promises a new method to solve the short-term power generation prediction problems.
作者
侯勇严
杨澳
郭文强
张栋
师帅
HOU Yong-yan;YANG Ao;GUO Wen-qiang;ZHANG Dong;SHI Shuai(School of Electrical and Control Engineering, Shaanxi University of Science & Technology, Xi′an 710021, China;School of Electronic Information and Artificial Intelligence, Shaanxi University of Science & Technology, Xi′an 710021, China)
出处
《陕西科技大学学报》
北大核心
2022年第4期171-177,共7页
Journal of Shaanxi University of Science & Technology
基金
陕西省科技厅重点研发计划项目(2020SF-286)
陕西省教育厅产业化计划项目(18JC003)
陕西省西安市科技计划项目(2019216514GXRC001CG002GXYD1.1)。
关键词
灰狼优化算法
神经网络
发电量
预测
grey wolf optimization algorithm
neural network
power generation
prediction
