摘要
为提高光伏发电功率预测精度,提出一种基于多策略改进麻雀搜索算法(multi strategy improved sparrow search algorithm,MSISSA)优化长短时记忆(long short-term memory,LSTM)网络的预测模型。首先,采用Logistic-Sine-Cosine混沌映射对麻雀种群初始化进行改进,增加初始种群的均匀性和遍历性。然后,针对在使用麻雀算法(sparrow search algorithm,SSA)优化LSTM网络时陷入局部最优及精度不足的问题,利用鱼鹰捕食策略、Levy飞行策略和柯西变异策略优化种群中麻雀的位置。最后,通过MSISSA优化得到LSTM网络的最优隐含层节点数、训练次数和学习率。实验证明,该模型比LSTM、GWO(grey wolf optimizer)-LSTM、WOA(whale optimization algorithm)-LSTM和SSA-LSTM模型具有更高的预测精度,尤其在阴天和雨天。该模型有助于电网稳定运行和电力系统调度,具有较高的实用价值。
To improve the accuracy of photovoltaic power generation prediction,a prediction model is proposed based on the long and short term memory(LSTM)network optimized by multi-strategy improved sparrow search algorithm(MSISSA).Firstly,the logistic-sine-cosine chaotic mapping is used to improve the initialization of the sparrow population,increasing the uniformity and ergodicity of the initial population.Then,to address the problems of falling into local optima and insufficient accuracy when optimizing the LSTM network using the sparrow algorithm,the osprey feeding strategy,the Levy flight strategy and the Cauchy mutation strategy are utilized to optimize the positions of the sparrows in the population.Finally,the optimal number of hidden layer nodes,training frequency,and learning rate of the LSTM network are obtained through MSISSA optimization.Experiments demonstrate that this model has higher prediction accuracy than the LSTM,GWO(grey wolf optimizer)-LSTM,WOA(whale optimization algorithm)-LSTM and SSA-LSTM models,especially on cloudy and rainy days.This model is helpful for the stable operation of power grids and the scheduling of power systems,and thus has high practical value.
作者
王玲芝
李晨阳
李程
刘婧
WANG Lingzhi;LI Chenyang;LI Cheng;LIU Jing(School of Automation,Xi'an University of Posts and Telecommunications,Xi'an 710121,China;Weinan Power Supply Company,State Grid Shaanxi Electric Power Co.,Ltd.,Weinan 714000,China)
出处
《武汉大学学报(工学版)》
北大核心
2025年第8期1356-1366,共11页
Engineering Journal of Wuhan University
基金
国家自然科学基金项目(编号:62073259、52177194)
西安邮电大学创新基金项目(编号:CXJJYL2023064)。
关键词
光伏功率预测
麻雀搜索算法
混合混沌映射
鱼鹰捕食策略
Levy飞行策略
柯西变异策略
长短时记忆
photovoltaic power prediction
sparrow search algorithm
hybrid chaotic mapping
osprey predation strategy
Levy flight strategy
Cauchy mutation strategy
long and short term memory
