基于SSA-VMD-GCN-GM的短期负荷预测

Short-term Load Forecasting Based on SSA-VMD-GCN-GM

摘要

针对时间序列波动性大、负荷预测精度较差等问题,提出奇异谱分析、变分模态分解、图卷积神经网络和灰色模型结合的短期负荷预测方法.首先,采用SSA对负荷序列进行降噪,以减少负荷数据噪声干扰.然后,利用VMD分解去噪后的序列,加快计算效率,得到分解精度更高的数据.接着,使用互信息筛选模型输入参数,并构造图数据,利用GCN网络进行预测,并将预测值送入全连接层作为初步预测结果.最后,通过GM对负荷误差值进行修正,将修正后的误差预测结果与初步预测结果相加作为模型最终预测值.算例仿真证明:模型预测效果良好.

Abstract

A short-term load forecasting method is proposed, which combines singular spectrum analysis (SSA), variational modal decomposition (VMD), graph convolutional neural network (GCN) and gray model (GM) to address the problems of large fluctuation in time series and poor load forecasting accuracy. First, SSA is used to denoise the load sequence and reduce the noise interference in load data. Then, VMD is used to decompose the denoised sequence, which improves the decomposition accuracy and accelerates the computation. Next, the input parameters are selected using the mutual information and graph data are constructed, predictions are made using the GCN network, and the predicted values are fed to a fully connected layer as preliminary prediction results. Finally, GM is used to correct the load error values, and the corrected error prediction results are added to the preliminary prediction results as the final prediction value of the model. Through the simulation of arithmetic cases, the proposed model show good performance in load forecasting.

关键词

奇异谱分析 / 变分模态分解 / 图卷积神经网络 / 短期负荷预测 / 灰色模型

Key words

singular spectrum analysis / variational modal decomposition / graph convolutional neural network / short-term load prediction / gray model

参考文献

[1] 廖旎焕, 胡智宏, 马莹莹, 等. . 电力系统短期负荷预测方法综述[J]. 电力系统保护与控制, 2011(01): 147-152.

[2] 杨胡萍, 余阳, 汪超, 等. . 基于VMD-CNN-BIGRU的电力系统短期负荷预测[J]. 中国电力, 2022(10): 71-76.

[3] 万昆, 柳瑞禹. 区间时间序列向量自回归模型在短期电力负荷预测中的应用[J]. 电网技术, 2012(11): 77-81.

[4] 王典礼. 指数平滑日负荷预测[J]. 水电能源科学, 1986(01): 91-97.

[5] 黎祚, 周步祥, 林楠. 基于模糊聚类与改进BP算法的日负荷特性曲线分类与短期负荷预测[J]. 电力系统保护与控制, 2012, 40(03): 56-60.

[6] 李焱, 贾雅君, 李磊, 等. . 基于随机森林算法的短期电力负荷预测[J]. 电力系统保护与控制, 2020(21): 117-124.

[7] Zhang S, Chen R, Cao J, et al. A CNN and LSTM-based multi-task learning architecture for short and medium-term electricity load forecasting[J]. Electric Power Systems Research, 2023, 222: 109507.

[8] 姬莉霞, 赵耀, 马郑祎, 等. . 基于iForest-BiLSTM-Attention的数据库负载预测方法[J]. 郑州大学学报(理学版), 2022, 54(06): 66-73.

[9] 汪鸣, 彭舰, 黄飞虎. 基于门控循环图卷积网络的交通流预测[J]. 计算机应用研究, 2022, 39(08): 2301-2305.

[10] Song Y, Tang D, Yu J, et al. Short-term forecasting based on graph convolution networks and multiresolution convolution neural networks for wind power[J]. IEEE Transactions on Industrial Informatics, 2022, 19(02): 1691-1702.

[11] 杜启亮, 向照夷, 田联房, 等. . 用于动作识别的双流自适应注意力图卷积网络[J]. 华南理工大学学报(自然科学版), 2022, 50(12): 20-29.

[12] 史含笑, 王雷春. 结合LSTM和自注意力机制的图卷积网络短期电力负荷预测[J/OL]. 计算机应用: 1-9[2023-05-09]. http://kns.cnki.net/kcms/detail/51.1307.TP.20230411.0901.002.html.

[13] 董雷, 陈振平, 韩富佳, 等. . 基于图卷积神经网络与K-means聚类的居民用户集群短期负荷预测[J/OL]. 电网技术: 1-11[2023-05-09]. http://kns.cnki.net/kcms/detail/11.2410.TM.20221013.1648.003.html.

[14] Rafi S H, Deeba S R, Hossain E. A Short-Term Load Forecasting Method Using Integrated CNN and LSTM Network[J]. IEEE Access, 2021, 9: 32436-32448.

[15] 王荣茂, 谢宁, 于海洋, 等. . 基于EMD-LSTM模型的台区负荷短期预测方法[J]. 实验室研究与探索, 2022, 41(01): 62-66.

[16] Liao Z, Huang J, Cheng Y, et al. A novel decomposition-based ensemble model for short-term load forecasting using hybrid artificial neural networks[J]. Applied Intelligence, 2022, 52(10): 11043-11057.

[17] 叶剑华, 曹旌, 杨理, 等. . 基于变分模态分解和多模型融合的用户级综合能源系统超短期负荷预测[J]. 电网技术, 2022, 46(07): 2610-2622.

[18] He F, Zhou J, Feng Z, et al. A hybrid short-term load forecasting model based on variational mode decomposition and long short-term memory networks considering relevant factors with Bayesian optimization algorithm[J]. Applied Energy, 2019, 237: 103-116.

[19] Jiang Y M, Liu M S, Li Y Y, et al. Enhanced neighborhood node graph neural networks for load forecasting in smart grid[J]. International Journal of Machine Learning and Cybernetics, 2023, 1: 20. https://doi.org/10.1007/s13042-023-01796-8.

[20] Zhu D, Wang Y, Zhang F. Energy price prediction integrated with singular spectrum analysis and long short-term memory network against the background of carbon neutrality[J]. Energies, 2022, 15(21): 8128.

[21] Neeraj N, Mathew J, Agarwal M, et al. Long short-term memory-singular spectrum analysis-based model for electric load forecasting[J]. Electrical Engineering, 2021, 103: 1067-1082.

基金资助

国家自然科学基金青年科学基金资助项目(51809097)

湖北省重点研发计划项目(2021BAA193)