融合特征筛选与信号分解协同优化的xLSTM-Informer风电功率预测研究

doi:10.19595/j.cnki.1000-6753.tces.251380

摘要
图/表
参考文献
相关文章 (5)

全文: PDF (1582 KB)
输出: BibTeX | EndNote (RIS)

摘要面对风电随机性和波动性引发的电网风险,高精度风电功率预测是保障电网安全和能效的关键。针对现有方法在数据模态混叠、特征筛选局限及季节性适应不足等问题,提出“混合特征筛选-季节性自适应分解-xLSTM-Informer预测”框架。首先,结合Spearman相关系数与MIC方法分别提取气象因子的线性和非线性特征;其次,基于SGMD-rePE-VMD设计“分解-重构-分解”机制,利用辛几何模态分解（SGMD）初次分解,通过鲁棒排列熵（rePE）准则自适应重组分量,并采用变分模态分解（VMD）二次分解高频分量;最后,构建xLSTM-Informer混合模型,xLSTM捕捉长时依赖,Informer处理长序列全局特征,通过门控融合模块整合跨尺度信息。基于福建某风电场数据的实验结果表明：该组合方法四季的预测均方根误差（RMSE）均值为0.5804,R²均值为0.9916,显著优于其他基准模型及组合方法,验证了特征筛选有效性、序列分解优越性及xLSTM-Informer模型的核心作用,为风光储协同调度提供可靠技术支撑。

	服务

	把本文推荐给朋友
	加入我的书架
	加入引用管理器
	E-mail Alert
	RSS
	作者相关文章
	张振龙
	聂达文
	张新生
	任金哥

关键词 ：风电功率预测, 特征筛选, 信号分解, xLSTM模型, Informer模型

Abstract：In the field of wind farm power prediction, the non-stationarity of power sequences and the coupling effect of meteorological factors severely restrict the prediction accuracy. However, traditional prediction models (such as single LSTM or Informer) are difficult to effectively decouple the fluctuation mode features, and existing feature selection methods often ignore the nonlinear interaction between variables. Although some studies have introduced signal decomposition techniques in recent years, most of them have problems such as mode aliasing and unclear physical meaning of component reconstruction. At the same time, the redundant information in the feature selection stage significantly increases the computational burden. To address these challenges, an xLSTM-Informer prediction framework that integrates feature selection and signal decomposition for collaborative optimization is proposed. By combining the time-varying characteristics of meteorological factors and multi-scale power sequence features, it achieves high-precision wind power prediction.
Firstly, based on the historical power data of the wind farm and multi-dimensional meteorological monitoring data, the training sample set is constructed on a quarterly basis. Secondly, a two-stage feature processing procedure is executed: in the first stage, the Spearman-MIC hybrid feature selection strategy is adopted to simultaneously capture the linear and nonlinear correlation features of meteorological factors and eliminate redundant variables; in the second stage, the “SGMD-rePE-VMD” signal decomposition chain is established, where the symplectic geometry mode decomposition (SGMD) is used to decouple the original power sequence, the components are adaptively reorganized based on the robust permutation entropy (rePE) criterion, and the high-frequency components are further optimized by variational mode decomposition (VMD). Finally, the xLSTM-Informer hybrid model is designed: the xLSTM module captures long-term dependency features through an exponential gating mechanism, the Informer module processes global sequence correlations with ProbSparse attention, and the gated fusion network integrates cross-scale information. The model generates the final power value through the prediction of decomposed sub-sequences and result reconstruction.
The validation results of actual wind farm data show that when the prediction step is 15 minutes, the average root mean square error of the proposed model in four seasons is 0.5804 MW, and the coefficient of determination reaches 0.9516. As the prediction duration extends to 6 hours, the average prediction error of the model stabilizes at 1.982 MW. The key feature ablation experiments show that when Spearman-MIC feature selection is not used, the average root mean square error rises to 2.1266 MW; when VMD secondary decomposition is not implemented, the error climbs to 2.7870 MW. In the model comparison test, xLSTM-Informer has a significant advantage over the single LSTM (average RMSE of 2.1253 MW) and Informer (average RMSE of 1.870 8 MW).
Comprehensive experimental analysis leads to the following conclusions: (1) The proposed Spearman-MIC hybrid screening mechanism can effectively identify key meteorological factors (such as the MIC of wind speed at 100 meters height in all four seasons being greater than 0.41), reducing the prediction error by 13.2%. (2) The “SGMD-rePE-VMD” decomposition chain, by suppressing the modal aliasing phenomenon, improves the accuracy by 31.2% compared to single-stage decomposition models. (3) This framework only requires historical power and meteorological monitoring data, making it more suitable for actual wind farm deployment than physical models that rely on numerical weather forecasts. Meanwhile, its modeling method of fusing multi-source data into a three-dimensional feature tensor significantly enhances the expression ability of spatiotemporal features.

Key words： Wind power prediction feature screening signal decomposition xLSTM model Informer model

收稿日期: 2025-08-06

PACS:

TM614

基金资助:国家自然科学基金青年项目（72504216）和陕西省创新能力支撑计划软科学项目（2024ZC-YBXM-010）资助

通讯作者: 张新生男,1978年生,教授,博士生导师,研究方向为风力发电系统及控制。E-mail：zhangxs@xauat.edu.cn

作者简介: 张振龙男,1980年生,副教授,硕士生导师,研究生方向为风力发电功率预测。E-mail：669745984@qq.com

引用本文:

张振龙, 聂达文, 张新生, 任金哥. 融合特征筛选与信号分解协同优化的xLSTM-Informer风电功率预测研究[J]. 电工技术学报, 0, (): 251380-. Zhang Zhenlong, Nie Dawen, Zhang Xinsheng, Ren jinge. Research on xLSTM-Informer Wind Power Prediction with Synergistic Optimization of Feature Selection and Signal Decomposition. Transactions of China Electrotechnical Society, 0, (): 251380-.

链接本文:

https://dgjsxb.ces-transaction.com/CN/10.19595/j.cnki.1000-6753.tces.251380 https://dgjsxb.ces-transaction.com/CN/Y0/V/I/251380

[1] Li Tian, Ai Lijuan, Yang Qingshan, et al.Short-term wind power prediction based on multiscale numerical simulation coupled with deep learning[J]. Renewable Energy, 2025, 246: 122951.
[2] 李宇佳, 陈富豪, 阎洁, 等. 基于迁移学习和自编码器的极端天气自适应短期风电功率预测[J]. 电力系统自动化, 2025, 49(16): 85-95.
Li Yujia, Chen Fuhao, Yan Jie, et al.Adaptive short-term wind power forecasting under extreme weather events based on transfer learning and autoencoder[J]. Automation of Electric Power Systems, 2025, 49(16): 85-95.
[3] 杨茂, 张书天, 王勃. 基于因果正则化极限学习机的风电功率短期预测方法[J]. 电力系统保护与控制, 2024, 52(11): 127-136.
Yang Mao, Zhang Shutian, Wang Bo.Short-term wind power forecasting method based on a causal regularized extreme learning machine[J]. Power System Protection and Control, 2024, 52(11): 127-136.
[4] 杨茂, 陈郁林. 基于EMD分解和集对分析的风电功率实时预测[J]. 电工技术学报, 2016, 31(21): 86-93.
Yang Mao, Chen Yulin.Real-time prediction for wind power based on EMD and set pair analysis[J]. Transactions of China Electrotechnical Society, 2016, 31(21): 86-93.
[5] Meng Anbo, Zhang Haitao, Dai Zhongfu, et al.An adaptive distribution-matched recurrent network for wind power prediction using time-series distribution period division[J]. Energy, 2024, 299: 131383.
[6] 叶瑞丽, 郭志忠, 刘瑞叶, 等. 基于小波包分解和改进Elman神经网络的风电场风速和风电功率预测[J]. 电工技术学报, 2017, 32(21): 103-111.
Ye Ruili, Guo Zhizhong, Liu Ruiye, et al.Wind speed and wind power forecasting method based on wavelet packet decomposition and improved Elman neural network[J]. Transactions of China Electrotechnical Society, 2017, 32(21): 103-111.
[7] Krishna Rayi V, Mishra S P, Naik J, et al.Adaptive VMD based optimized deep learning mixed kernel ELM autoencoder for single and multistep wind power forecasting[J]. Energy, 2022, 244: 122585.
[8] 宋江涛, 崔双喜, 樊小朝, 等. 基于SGMD-SE与优化TCN-BiLSTM/BiGRU的超短期风功率预测[J]. 太阳能学报, 2024, 45(10): 588-596.
Song Jiangtao, Cui Shuangxi, Fan Xiaochao, et al.Ultra-short-term wind power prediction based on sgmd-se and optimized tcn-BiLSTM/BiGRU[J]. Acta Energiae Solaris Sinica, 2024, 45(10): 588-596.
[9] 韩博, 李长青, 刘卫亮, 等. 基于二次分解和BiGRU的超短期光伏发电功率预测[J]. 动力工程学报, 2025, 45(1): 62-69, 79.
Han Bo, Li Changqing, Liu Weiliang, et al.Ultra-short-term photovoltaic power forecasting based on secondary decomposition and BiGRU[J]. Journal of Chinese Society of Power Engineering, 2025, 45(1): 62-69, 79.
[10] Gong Zhipeng, Wan Anping, Ji Yunsong, et al.Improving short-term offshore wind speed forecast accuracy using a VMD-PE-FCGRU hybrid model[J]. Energy, 2024, 295: 131016.
[11] 魏振宇, 姜雪松, 杨立发. 基于误差补偿及IDBO-BiLSTM的风电功率短期预测[J]. 科学技术与工程, 2025, 25(6): 2397-2405.
Wei Zhenyu, Jiang Xuesong, Yang Lifa.Short term prediction of wind power based on error compensation and IDBO-BiLSTM[J]. Science Technology and Engineering, 2025, 25(6): 2397-2405.
[12] Sun Wei, Wang Xiaoxuan, Tan Bin.Multi-step wind speed forecasting based on a hybrid decomposition technique and an improved back-propagation neural network[J]. Environmental Science and Pollution Research, 2022, 29(33): 49684-49699.
[13] 王玲芝, 李晨阳, 刘婧, 等. 基于GRO-SSA-LSTM的短期光伏发电功率预测[J]. 太阳能学报, 2025, 46(2): 401-409.
Wang Lingzhi, Li Chenyang, Liu Jing, et al.Short term photovoltaic power prediction based on gro-ssa-lstm[J]. Acta Energiae Solaris Sinica, 2025, 46(2): 401-409.
[14] Gao Jiajing, Xing Hongmei, Wang Yongsheng, et al.Ultra-short-term wind power prediction based on hybrid denoising with improved CEEMD decomposition[J]. Renewable Energy, 2025, 251: 123352.
[15] Wang Shuangxin, Li Meng, Zhao Long, et al.Short-term wind power prediction based on improved small-world neural network[J]. Neural Computing and Applications, 2019, 31(7): 3173-3185.
[16] Yu Bingbing, Wang Yonggang, Wang Jun, et al.A hybrid model for short-term offshore wind power prediction combining Kepler optimization algorithm with variational mode decomposition and stochastic configuration networks[J]. International Journal of Electrical Power & Energy Systems, 2025, 168: 110703.
[17] Zhang Yagang, Chen Bing, Pan Guifang, et al.A novel hybrid model based on VMD-WT and PCA-BP-RBF neural network for short-term wind speed forecasting[J]. Energy Conversion and Management, 2019, 195: 180-197.
[18] Yassen M A, El-Kenawy E M, Abdel-Fattah M G, et al. Explainable artificial intelligence for wind power forecasting model based on long short-term memory[J]. Neural Computing and Applications, 2025, 37(19): 14589-14611.
[19] Zhang Wanqing, Lin Zi, Liu Xiaolei.Short-term offshore wind power forecasting - A hybrid model based on Discrete Wavelet Transform (DWT), Seasonal Autoregressive Integrated Moving Average (SARIMA), and deep-learning-based Long Short-Term Memory (LSTM)[J]. Renewable Energy, 2022, 185: 611-628.
[20] Cavalcante L, Bessa R J, Reis M, et al.LASSO vector autoregression structures for very short-term wind power forecasting[J]. Wind Energy, 2017, 20(4): 657-675.
[21] Yu Linlin, Meng Gaojun, Pau G, et al.Research on hierarchical control strategy of ESS in distribution based on GA-SVR wind power forecasting[J]. Energies, 2023, 16(4): 2079.
[22] Cai Long, Gu Jie, Ma Jinghuan, et al.Probabilistic wind power forecasting approach via instance-based transfer learning embedded gradient boosting decision trees[J]. Energies, 2019, 12(1): 159.
[23] Nakhchi M E, Win Naung S, Rahmati M.Wake and power prediction of horizontal-axis wind farm under yaw-controlled conditions with machine learning[J]. Energy Conversion and Management, 2023, 296: 117708.
[24] Wu Zhenlong, Fan Xinyu, Bian Guibin, et al.Short-term wind power forecast with turning weather based on DBSCAN-RFE-LightGBM[J]. Renewable Energy, 2025, 251: 123217.
[25] 苏向敬, 程子凡, 聂良钊, 等. 基于AGCN-LSTM模型的海上风电场功率概率预测[J]. 电力系统自动化, 2024, 48(22): 140-149.
Su Xiangjing, Cheng Zifan, Nie Liangzhao, et al.Power probability prediction for offshore wind farm based on AGCN-LSTM model[J]. Automation of Electric Power Systems, 2024, 48(22): 140-149.
[26] Lee S, Almomani M H, Ali Alomari S, et al.A novel deep learning framework with artificial protozoa optimization-based adaptive environmental response for wind power prediction[J]. Scientific Reports, 2025, 15: 18746.
[27] 刘雨佳, 樊艳芳, 白雪岩, 等. 基于特征交叉机制和误差补偿的风力发电功率短期预测[J]. 电工技术学报, 2023, 38(12): 3277-3288.
Liu Yujia, Fan Yanfang, Bai Xueyan, et al.Short-term wind power prediction based on feature crossover mechanism and error compensation[J]. Transactions of China Electrotechnical Society, 2023, 38(12): 3277-3288.
[28] 李云松, 张智晟. 考虑综合需求响应的Transformer-图神经网络综合能源系统多元负荷短期预测[J]. 电工技术学报, 2024, 39(19): 6119-6128.
Li Yunsong, Zhang Zhisheng.Transformer based multi load short-term forecasting of integrated energy system considering integrated demand response[J]. Transactions of China Electrotechnical Society, 2024, 39(19): 6119-6128.
[29] 金亮, 宋居恒, 马天赐, 等. 电磁轨道发射器高速下电流密度场预测[J]. 电工技术学报, 2024, 39(19): 5914-5928, 5936.
Jin Liang, Song Juheng, Ma Tianci, et al.Current density field prediction method for electromagnetic rail launcher at high speed[J]. Transactions of China Electrotechnical Society, 2024, 39(19): 5914-5928, 5936.
[30] 刘素贞, 任佳乐, 袁路航, 等. 磷酸铁锂电池荷电状态估计的多源数据特征提取[J]. 电工技术学报, 2025, 40(11): 3349-3361.
Liu Suzhen, Ren Jiale, Yuan Luhang, et al.Multi-source data feature extraction method for state of charge estimation of LiFePO₄ battery[J]. Transactions of China Electrotechnical Society, 2025, 40(11): 3349-3361.
[31] 张新生, 聂达文, 陈章政. 基于MIC特征提取与ICEEMD-RIME-DHKELM的建筑业碳排放预测模型[J]. 环境工程, 2025, 43(4): 46-58.
Zhang Xinsheng, Nie Dawen, Chen Zhangzheng.Research on a prediction model for carbon emissions of construction industry based on MIC feature extraction and ICEEMD-RIME-DHKELM[J]. Environmental Engineering, 2025, 43(4): 46-58.
[32] Zhang Guangyao, Wang Yi, Li Xiaomeng, et al.Enhanced symplectic geometry mode decomposition and its application to rotating machinery fault diagnosis under variable speed conditions[J]. Mechanical Systems and Signal Processing, 2022, 170: 108841.
[33] 刘晓东, 刘朦月, 陈寅生, 等. EEMD-PE与M-RVM相结合的轴承故障诊断方法[J]. 哈尔滨工业大学学报, 2017, 49(9): 122-128.
Liu Xiaodong, Liu Mengyue, Chen Yinsheng, et al.Rolling bearing fault diagnosis based on EEMD-PE coupled with M-RVM[J]. Journal of Harbin Institute of Technology, 2017, 49(9): 122-128.
[34] 屈龙腾, 李沛兴, 乔壮壮. 基于变分模态分解和希尔伯特变换的直流纹波检测[J]. 电气技术, 2020, 21(8): 66-72, 86.
Qu Longteng, Li Peixing, Qiao Zhuangzhuang.DC ripple detection based on variational mode decomposition and Hilbert transform[J]. Electrical Engineering, 2020, 21(8): 66-72, 86.
[35] 刘晓军, 熊健, 王艺博, 等. 考虑不确定变量变分模态分解及绿证-碳联合交易的综合能源系统经济优化调度[J]. 电工技术学报, 2025, 40(13): 4276-4291.
Liu Xiaojun, Xiong Jian, Wang Yibo, et al.Economic optimization of integrated energy system scheduling considering uncertainty variables variational mode decomposition and green certificate-carbon joint trading[J]. Transactions of China Electrotechnical Society, 2025, 40(13): 4276-4291.
[36] 李辰辉, 褚继峰, 龙潇, 等. 基于弧触头接触振动特征分析的高压SF₆断路器电寿命在线监测方法[J]. 电工技术学报, 2024, 39(15): 4883-4895.
Li Chenhui, Chu Jifeng, Long Xiao, et al.Online monitoring method of electrical life of high voltage SF₆ circuit breaker based on analysis of arcing contact vibration characteristics[J]. Transactions of China Electrotechnical Society, 2024, 39(15): 4883-4895.
[37] Xiang Nan, Zhang Tianyu, Liu Yu.State of health prediction for Lithium-Ion batteries using extended long Short-Term memory network and frequency enhanced channel attention mechanism with Variational mode Decomposition[J]. Measurement, 2025, 249: 117084.
[38] Zhou Haoyi, Zhang Shanghang, Peng Jieqi, et al.Informer: beyond efficient transformer for long sequence time-series forecasting[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2021, 35(12): 11106-11115.
[39] 吉兴全, 赵国航, 叶平峰, 等. 基于QMD-HBi GRU的短期光伏功率预测方法[J]. 高电压技术, 2024, 50(9): 3850-3859.
Ji Xingquan, Zhao Guohang, Ye Pingfeng, et al.Short-term PV forecasting method based on the QMD-HBiGRU model[J]. High Voltage Engineering, 2024, 50(9): 3850-3859.
[40] Fang Chengzhi, Chen Yushen, Deng Xiaolei, et al.Denoising method of machine tool vibration signal based on variational mode decomposition and Whale-Tabu optimization algorithm[J]. Scientific Reports, 2023, 13: 1505.