船舶电力负荷因具有噪声多、随机性和非线性强的特点，在短期电力负荷预测中存在噪声干扰、特征提取困难和模型拟合度差的问题。故本文提出一种结合霜冰优化算法（Rime Optimization Algorithm，RIME）的变分模态分解（Variational Mode Decomposition，VMD）、时序卷积网络（Temporal Convolutional Network，TCN）和注意力机制的组合预测模型。首先，通过RIME-VMD分解，将复杂的船舶电力负荷信号分解为多个仅包含简单负荷特征的单独模态，以减少噪声的影响同时提高分解效率；其次，通过TCN模型结合Attention机制对各模态分量进行预测并将结果组合，使模型自适应捕捉电力负荷中的非线性特征，提高时序预测能力；最后，实验分析表明，本文提出的RIME-VMD-TCN-Attention模型误差指标MAE、MAPE、RMSE和R2均优于传统LSTM模型、GRU模型、单一TCN模型和未经模态分解的混合模型，具有更高的预测精度。
Due to the characteristics of high noise, randomness, and strong nonlinearity, ship power load has problems of noise interference, difficulty in feature extraction, and poor model fitting in short-term power load forecasting. Therefore, this paper proposes a combined prediction model of Variational Mode Decomposition (VMD), Temporal Convolutional Network (TCN), and attention mechanism combined with Rime Optimization Algorithm (RIME). Firstly, through RIME-VMD decomposition, the complex ship power load signal is decomposed into multiple individual modes containing only simple load characteristics, reducing the impact of noise and improving decomposition efficiency. Secondly, by combining the TCN model with the Attention mechanism to predict and combine the results of each modal component, the model can adaptively capture the nonlinear features in power load and improve the ability of time series prediction. Finally, experimental analysis shows that the error metrics MAE, MAPE, RMSE, and R2 of the RIME-VMD-TCN-Attention model proposed in this paper are superior to traditional LSTM models, GRU models, single TCN models, and hybrid models without modal decomposition, with higher prediction accuracy.
2025,47(18): 112-118 收稿日期：2024-12-20
DOI：10.3404/j.issn.1672-7649.2025.18.019
分类号：U665.12；TP183
基金项目：国家重点研发计划资助项目（2022YFC3102805）
作者简介：骆佳馨（2001 – ），女，硕士研究生，研究方向为船舶电气与自动化
参考文献：
[1] 工业和信息化部. 五部委关于印发船舶制造业绿色发展行动纲要(2024—2030年)的通知: 工信部联重装[2023]254号[A/OL].(2023-12-26)[2024-10-30].
[2] 王锡淮, 朱思锋. 基于支持向量机的船舶电力负荷预测[J]. 中国电机工程学报, 2004, 24(10): 36-39.
[3] SHI H, WANG L, SCHERER R, et al. Short-term load forecasting based on adabelief optimized temporal convolutional network and gated recurrent unit hybrid neural network[J]. IEEE Access, 2021, 9: 66965-66981.
[4] PARK D C, EL-SHARKAWI M A. Electric load forecasting using an artificial neural network[J]. IEEE Transactions on Power Systems: A Publication of the Power Engineering Society, 1991, 6(2): 442-449.
[5] 刘银波, 吕越. RBF神经网络在船舶电力负荷预测中的应用[J]. 舰船科学技术, 2021, 43(18): 121-123.
LIU Y B, LU Y. Application of RBF neural network in ship power load forecasting[J]. ship science and technology, 2021, 43(18): 121-123.
[6] 赵洋, 王瀚墨, 康丽, 等. 基于时间卷积网络的短期电力负荷预测[J]. 电工技术学报, 2022, 37(5): 1242-1251.
[7] 任建吉, 位慧慧, 邹卓霖, 等. 基于CNN-BiLSTM-Attention的超短期电力负荷预测[J]. 电力系统保护与控制, 2022, 50(8): 108-116.
[8] 穆晨宇, 薛文斌, 穆羡瑛, 等. 基于VMD-LSTM- Attention模型的短期负荷预测研究[J]. 现代电子技术, 2023, 46(17): 174-178.
[9] 杨晓伟. 微电网短期电力负荷预测及优化调度方法研究[D]. 银川: 宁夏大学, 2023.
[10] KONSTANTIN D, DOMINIQUE Z. Variational mode deco-mposition[J]. IEEE Transactions on Signal Processing, 2014 , 62(3): 531-544.
[11] 吴松梅, 蒋建东, 燕跃豪等. 基于VMD-PSO-多核极限学习机的短期负荷预测[J]. 电力系统及其自动化学报, 2022, 34(5): 18-25.
[12] ZHAO Y, LI C, FU W, et al. A modified variational mode decomposition method based on envelope nesting and multi- criteria evaluation[J]. Journal of Sound and Vibration, 2019, 468: 115099.
[13] SU H, ZHAO D, HEIDARI A, et al. RIME: A physics-based optim- ization[J]. Neurocomputing, 2023, 532: 183-214.
[14] 王俊, 李霞, 周昔东, 等. 基于VMD和LSTM的超短期风速预测[J]. 电力系统保护与控制, 2020, 48(11): 45-52.
[15] BAI S J, KOLTER JZ, KOLTUN V. An empirical evaluation of generic convolutional and recurrent networks for sequence modeling[J]. arXiv: 1803.01271, 2018.
[16] 王琪凯, 熊永康, 陈瑛, 等. 基于Attention机制优化CNN-seq2seq模型的非侵入式负荷监测[J]. 电力系统及其自动化学报, 2022, 34(12): 27-34+42.
