采用级联回归残差校正麻雀搜索算法优化长短期记忆网络的光伏发电功率预测方法

Photovoltaic Power Forecasting Method Based on a Long Short-Term Memory Network Optimized by Cascade Regression Residual Correction Sparrow Search Algorithm

摘要

针对光伏发电功率序列波动性强、非线性特征显著以及预测精度易受噪声干扰等问题，提出一种基于改进自适应噪声完备集合经验模态分解（ICEEMDAN）与级联回归（CR）残差校正麻雀搜索算法优化长短期记忆网络（SSA-LSTM）的光伏发电功率预测方法。首先，采用 ICEEMDAN 分解对原始光伏发电功率序列进行分解，得到多个固有模态函数（IMF），以提取不同频率尺度下的时序特征并削弱噪声干扰；其次，引入麻雀搜索算法（SSA）对长短期记忆网络（LSTM）模型的关键参数进行寻优，并利用优化后的 LSTM 模型对各 IMF 分量分别进行预测；最后，采用级联回归对各分量重构后的预测结果进行残差校正，进一步减小模型输出偏差，提高整体预测精度。仿真结果表明：按四季测试样本数加权平均计算，相较于传统 LSTM 模型，所提方法平均绝对误差（MAE）降低了 69.97%，均方根误差（RMSE）降低了 67.49%；相较于未校正模型，MAE 降低了 34.01%，RMSE 降低了 27.78%，决定系数R²大于 0.96。研究结果表明，该方法能够有效降低预测偏差，提高模型在不同季节条件下的稳定性和适应性，为光伏发电功率预测提供了高精度、高鲁棒性的技术手段。

Abstract

To address the strong volatility and significant nonlinear characteristics of photovoltaic power sequences,as well as the susceptibility of forecasting accuracy to noise interference,a photovoltaic power forecasting method integrating improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN),a sparrow search algorithm-optimized long short-term memory network (SSA-LSTM),and cascade regression (CR) error correction is proposed.The original power sequence is decomposed using ICEEMDAN to obtain multiple intrinsic mode functions (IMFs),thereby effectively separating multiscale temporal features and suppressing noise interference.The sparrow search algorithm (SSA) is introduced to adaptively optimize the key parameters of the LSTM model.Cascade regression is used to perform parallel error correction on the preliminary prediction results obtained after reconstruction of the components,thereby further reducing the model output bias and improving the overall prediction accuracy.It is shown by simulation results that,based on a weighted average according to the numbers of test samples in the four seasons,the mean absolute error (MAE) and root mean square error (RMSE) are reduced by 69.97% and 67.49%,respectively,compared with those of the conventional LSTM model.Compared with the uncorrected model,the e MAE and e RMSE are reduced by 34.01% and 27.78%,respectively,and all coefficients of determination (R ²) exceed 0.96.The results show that the forecasting bias can be effectively reduced and the stability and adaptability of the model under different seasonal conditions can be improved by the proposed method,providing a high-accuracy and highly robust technical approach for photovoltaic power forecasting.

关键词

光伏发电功率预测 / 级联回归残差校正 / 麻雀搜索算法 / 自适应噪声完备集合经验模态分解 / 长短期记忆网络

Key words

photovoltaic power forecasting / cascade regression / sparrow search algorithm / complete ensemble empirical mode decomposition with adaptive noise / long short-term memory network

参考文献

[1] Li Zhengmao,Wu Lei,Xu Yan,et al. Distributed tri—layer risk—averse stochastic game approach for energy trading among multi—energy microgrids[J]. Applied Energy, 2023,331：120282.

[2] Sun Wei,Wang Yuwei. Short—term wind speed forecasting based on fast ensemble empirical mode decomposition,phase space reconstruction,sample entropy and improved back—propagation neural network[J]. Energy Conversion and Management, 2018,157：1-12.

[3] 文劲宇,周博,魏利屾. 中国未来电力系统储电网初探[J]. 电力系统保护与控制, 2022,50（7）：1-10.

[4] Wen Jinyu,Zhou Bo,Wei Lishen. Preliminary study on an energy storage grid for future power system in China[J]. Power System Protection and Control, 2022,50（7）：1-10.

[5] 康重庆,夏清,张伯明. 电力系统负荷预测研究综述与发展方向的探讨[J]. 电力系统自动化, 2004,28（17）：1-11.

[6] Kang Zhongqing,Xia Qing,Zhang Boming. Review of power system load forecasting and its development[J]. Automation of Electric Power Systems, 2004,28（17）：1-11.

[7] Ramkumar G,Sahoo S,Amirthalakshmi T M,et al. A short—term solar photovoltaic power optimized prediction interval model based on FOS—ELM algorithm[J]. International Journal of Photoenergy, 2021,2021（1）：3981456.

[8] Pan Mingzhang,Li Chao,Gao Ran,et al. Photovoltaic power forecasting based on a support vector machine with improved ant colony optimization[J]. Journal of Cleaner Production, 2020,277：123948.

[9] 马伟,乔颖,谢丽蓉,等. 考虑气象特征波动划分阈值的双目标短期风功率预测[J]. 高电压技术, 2022,48（10）：4154-4162.

[10] Ma Wei,Qiao Ying,Xie Lirong,et al. Short—term wind power prediction with dual targets considering the threshold of meteorological characteristic fluctuation partition[J]. High Voltage Engineering, 2022,48（10）：4154-4162.

[11] 薛阳,燕宇铖,贾巍,等. 基于改进灰狼算法优化长短期记忆网络的光伏功率预测[J]. 太阳能学报, 2023,44（7）：207-213.

[12] Xue Yang,Yan Yucheng,Jia Wei,et al. Photovoltaic power prediction model based on IGWO—LSTM[J]. Acta Energiae Solaris Sinica, 2023,44（7）：207-213.

[13] 武小梅,张琦,田明正. 基于 VMD—SE 和优化支持向量机的光伏预测方法[J]. 电力科学与工程, 2017,33（9）：29-36.

[14] Wu Xiaomei,Zhang Qi,Tian Mingzheng. Photovoltaic prediction method based on VMD—SE and optimized support vector machine[J]. Electric Power Science and Engineering, 2017,33（9）：29-36.

[15] 郑少凯. 基于深度学习与前端信号处理的 MEMS 陀螺仪温漂补偿研究[D].山东泰安：山东农业大学,2024.

[16] 袁建华,谢斌斌,何宝林,等. 基于 DTW—VMD—PSO—BP 的光伏发电功率短期预测方法[J]. 太阳能学报, 2022,43（8）：58-66.

[17] Yuan Jianhua,Xie Binbin,He Baolin,et al. Short term forecasting method of photovoltaic output based on DTW—VMD—PSO—BP[J]. Acta Energiae Solaris Sinica, 2022,43（8）：58-66.

[18] 杨宁宁,王怡昕,吴朝俊,等. 基于 SSA—VMD—LIESN 的短期风电功率预测方法研究[J]. 太阳能学报, 2025,46（5）：440-447.

[19] Yang Ningning,Wang Yixin,Wu Chaojun,et al. Research on short—term wind power prediction method based on SSA—VMD—LIESN[J]. Acta Energiae Solaris Sinica, 2025,46（5）：440-447.

[20] 岳有军,吴明沅,王红君,等. 基于 CNN—GRU—ISSA—XGBoost 的短期光伏功率预测[J]. 南京信息工程大学学报, 2024,16（2）：231-238.

[21] Yue Youjun,Wu Mingyuan,Wang Hongjun,et al. Short term photovoltaic power prediction based on CNN—GRU—ISSA—XGBoost[J]. Journal of Nanjing University of Information Science & Technology, 2024,16（2）：231-238.

[22] 葛泉波,李凯,张兴国. 基于多关键点检测加权融合的无人机相对位姿估计算法[J]. 自动化学报, 2024,50（7）：1402-1416.

[23] Ge Quanbo,Li Kai,Zhang Xingguo. Relative pose estimation algorithm for unmanned aerial vehicles based on weighted fusion of multiple keypoint detection[J]. Acta Automatica Sinica, 2024,50（7）：1402-1416.

[24] 代少升,熊昆,吴云铎,等. 多视角约束级联回归的视频人脸特征点跟踪[J]. 计算机应用, 2022,42（8）：2415-2422.

[25] Dai Shaosheng,Xiong Kun,Wu Yunduo,et al. Video facial landmark tracking by multi—view constrained cascade regression[J]. Journal of Computer Applications, 2022,42（8）：2415-2422.

[26] 曹瑞钧,郭其一. 基于 ICEEMDAN—FE 和支持向量机的三电平逆变器的故障诊断技术[J]. 机车电传动, 2023（1）：97-103.

[27] Cao Ruijun,Guo Qiyi. Fault diagnosis technology for three—level inverter based on ICEEMDAN—FE and SVM[J]. Electric Drive for Locomotives, 2023（1）：97-103.

[28] 杨洋,郭兴明,郑伊能,等. 基于 ICEEMDAN—MSE 的左室舒张功能障碍心音信号的识别研究[J]. 仪器仪表学报, 2022,43（1）：274-281.

[29] Yang Yang,Guo Xingming,Zheng Yineng,et al. Study on left ventricular diastolic dysfunction heart sound signals identification based on ICEEMDAN—MSE[J]. Chinese Journal of Scientific Instrument, 2022,43（1）：274-281.

[30] Xue Jiankai,Shen Bo. A novel swarm intelligence optimization approach：sparrow search algorithm[J]. Systems Science & Control Engineering, 2020,8（1）：22-34.

[31] 任俊,魏霞,黄德启,等. 基于眼睛状态多特征融合的疲劳驾驶检测[J]. 计算机工程与设计, 2022,43（11）：3187-3194.

[32] Ren Jun,Wei Xia,Huang Deqi,et al. Fatigue driving detection based on multi—feature fusion of eye state[J]. Computer Engineering and Design, 2022,43（11）：3187-3194.

[33] 石启萌. 融合深度学习与级联回归的头影测量分析系统[D].天津：天津大学,2021.

[34] 陈庆明,廖鸿飞,孙颖楷,等. 改进的蜣螂优化算法及光伏发电功率预测应用[J]. 太阳能学报, 2025,46（9）：445-454.

[35] Chen Qingming,Liao Hongfei,Sun Yingkai,et al. Improved dung beetle optimization algorithm and application to photovoltaic power generation prediction[J]. Acta Energiae Solaris Sinica, 2025,46（9）：445-454.

[36] 徐昌,许野,王晓晖,等. 基于 KPCA 和 HC 的 IPSO—LSTM 光伏出力预测模型研究[J]. 太阳能学报, 2025,46（5）：362-374.

[37] Xu Chang,Xu Ye,Wang Xiaohui,et al. Research on combined photovoltaic output prediction model based on KPCA,HC and IPSO—LSTM[J]. Acta Energiae Solaris Sinica, 2025,46（5）：362-374.

[38] 姚钦才,向文国,陈时熠,等. 基于 ICEEMDAN—KPCA—ICPA—LSTM 的光伏发电功率预测[J]. 动力工程学报, 2025,45（3）：374-382.

[39] Yao Qincai,Xiang Wenguo,Chen Shiyi,et al. Photovoltaic power forecasting based on ICEEMDAN—KPCA—ICPA—LSTM[J]. Journal of Chinese Society of Power Engineering, 2025,45（3）：374-382.

[40] Wang Tianlei,Gu Shaowei,Liu Renju,et al. Cuckoo catfish optimizer：a new meta—heuristic optimization algorithm[J]. Artificial Intelligence Review, 2025,58（10）：326.

[41] Dragomiretskiy K,Zosso D. Variational mode decomposition[J]. IEEE Transactions on Signal Processing, 2014,62（3）：531-544.

[42] 高岩,吴汉斌,张纪欣,等. 基于组合深度学习的光伏功率日前概率预测模型[J]. 中国电力, 2024,57（4）：100-110.

[43] Gao Yan,Wu Hanbin,Zhang Jixin,et al. Day—ahead probabilistic prediction model for photovoltaic power based on combined deep learning[J]. Electric Power, 2024,57（4）：100-110.

基金资助

国家自然科学基金资助项目(52577138)