基于CEEMD-WOA-LSTM的光伏发电功率预测
Prediction of Photovoltaic Power Generation based on CEEMD-WOA-LSTM
李恺丽 1王剑斌 1沈怡俊 1陈博1
作者信息
- 1. 浙江工业大学信息工程学院,浙江 杭州 310032
- 折叠
摘要
针对实际电力系统中光伏发电的波动性和不确定性,建立了基于CEEMD-WOA-LSTM的光伏发电功率预测模型.首先,采用皮尔逊相关系数法确定辐照度、湿度、温度和风速为光伏功率的关键影响因素,基于高斯混合模型聚类将数据集分为晴天、多云、雨天3种天气类型,以降低训练集与测试集之间的差异并提高预测模型的泛化能力,从而完成数据预处理.其次,采用互补集合经验模态分解对预处理后的数据进行分解并重构,降低其强随机性和复杂性,通过长短期记忆神经网络对分解所得的各本征模态函数分量进行功率预测,并利用鲸鱼优化算法优化网络参数以提升预测精度,从而叠加各分量的预测结果以确定最终预测值.最后,通过实验验证所提方法的有效性.结果表明:与现有方法相比,在不同天气条件下CEEMD-WOA-LSTM的预测精度均有所提高,且在复杂天气条件时展现出更好的稳定性和鲁棒性.
Abstract
A photovoltaic(PV)power prediction model based on CEEMD-WOA-LSTM was established to address the fluctuation and uncertainty issues of PV power generation in practical power systems.First-ly,the Pearson correlation coefficient method was used to determine irradiance,humidity,temperature,and wind speed as the key influencing factors of PV power.Based on the Gaussian mixture model cluste-ring,the dataset was divided into three weather types including sunny,cloudy,and rainy,to reduce the difference between the training and testing sets and improve the generalization ability of the prediction model,thus completing data preprocessing.Secondly,complementary ensemble empirical mode decom-position(CEEMD)was adopted to decompose and reconstruct the preprocessed data to reduce its strong randomness and complexity.Long short-term memory(LSTM)neural network was used to predict the power of each decomposed intrinsic mode function(IMF),and the whale optimization algorithm(WOA)was used to optimize network parameters to enhance prediction accuracy.Thus,the predicted results of each component were superimposed to determine the final predicted value.Finally,the effectiveness of the proposed method was validated through experiments.The results show that compared with existing methods,CEEMD-WOA-LSTM shows improved prediction accuracy under different weather conditions and demonstrates better stability and robustness under complex weather conditions.
关键词
光伏功率预测/CEEMD/LSTM神经网络/鲸鱼优化算法Key words
photovoltaic power prediction/CEEMD/LSTM neural network/whale optimization algorithm(WOA)引用本文复制引用
出版年
2025
NETL
NSTL
维普
万方数据