首页|基于GWO-GRU的光伏发电功率预测

基于GWO-GRU的光伏发电功率预测

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
针对长短期记忆网络(LSTM)应用于光伏发电功率预测时存在的耗时长或精准度低的问题,提出基于灰狼算法(GWO)优化门控循环单元(GRU)的光伏发电功率预测模型.通过GWO算法优化GRU模型的超参数,以近似最优参数建立光伏发电功率预测模型.结果表明,长时功率预测时,GWO-GRU模型的均方根误差更低、拟合系数更高、耗时更少,比传统LSTM模型的平均绝对误差降低10.20%;短时功率预测时,GWO-GRU模型在3种典型天气条件下不仅预测的平均误差最低、稳定性最强,而且比GWO-LSTM模型的平均用时节省17.24%.不同时长的功率预测表明,GWO-GRU相对于LSTM光伏功率预测效果更佳.
PHOTOVOLTAIC POWER PREDICTION MODEL BASED ON GWO-GRU
The long short-term memory network(LSTM)model has the problem of long time consumption or low accuracy when applied to the prediction of photovoltaic power generation.A photovoltaic power power prediction model based on the grey wolf algorithm(GWO)optimized gated recurrent unit(GRU)was proposed.The photovoltaic power prediction model is established by the approximate optimal hyperparameter,which is obtained by the GWO algorithm.The results show that in terms of long-term power prediction,the GWO-GRU model has lower root mean square error,higher fitting coefficients,and less time consumption,with an average absolute error reduction of 10.20%compared to traditional LSTM models.In terms of short-term power prediction,the GWO-GRU model not only has the lowest average prediction error and the strongest stability under three typical weather conditions,but also saves 17.24%of the average time compared to the GWO-LSTM model.Power predictions of different durations indicate that GWO-GRU performs better in predicting photovoltaic power compared to LSTM.

photovoltaic power generationpower forecastinggated recurrent unitgrey wolf optimizerlong short-term memorytime series

陈庆明、廖鸿飞、孙颖楷、曾亚森

展开 >

中山火炬职业技术学院光电信息学院,中山 528400

广东万和新电气股份有限公司万和研究院,佛山 528000

光伏发电 功率预测 门控循环单元 灰狼算法 长短期记忆网络 时间序列

广东省普通高校特色创新项目广东省高职院校产教融合创新平台项目2023年中山火炬职业技术学院校级课程思政示范课程

2022KTSCX3332020CJPT0162023KCSZ15

2024

10.19912/j.0254-0096.tynxb.2023-1248

太阳能学报
中国可再生能源学会

太阳能学报

CSTPCD北大核心
影响因子：0.392
ISSN：0254-0096
年,卷(期)：2024.45(7)
  • 14