复杂工况和恶劣环境下的船舶电力负荷预测对船舶的能量管理意义重大，因此提出一种基于多源域迁移学习（MDT）、稀疏自编码器（SAE）和nCPSO算法优化一维卷积神经网络和双向长短时记忆神经网络（nCPSO-1DCNN-BiLSTM）的船舶电力负荷预测模型。首先利用SAE增强多工况数据特征，然后建立nCPSO-1DCNN-BiLSTM特征提取模型，最后在互相关法（CORAL）和联合最大平均差异法（JMMD）算法作用下实现域间自适应并利用多个源域数据进行目标域电力负荷预测。结果表明，所提方法在多种工况下较不迁移和单源域迁移模型精度均有所提升，对船舶设备的能量管理具有一定指导意义。
The prediction of ship power load under complex working conditions and harsh environments is crucial for energy management of ships. This study proposes a ship power load prediction model, nCPSO-1DCNN-BiLSTM, based on multi-source domain transfer learning (MDT), sparse autoencoder (SAE), and optimization with the nCPSO algorithm. Initially, SAE is used to enhance features of multi-condition data, followed by the establishment of an nCPSO-1DCNN-BiLSTM feature extraction model. Domain adaptation is achieved through the CORAL and JMMD algorithms, utilizing multiple source domain data for target domain power load prediction. Results show that the proposed method improves accuracy in various operating conditions for both non-migration and single source domain migration models, providing valuable insights for ship equipment energy management.
2025,47(7): 153-159 收稿日期：2024-6-18
DOI：10.3404/j.issn.1672-7649.2025.07.028
分类号：U665.12；TM743
基金项目：国家自然科学基金资助项目（52271279）；黄骅港务新能源拖轮技术研究项目（SBZX〔2023〕34号）
作者简介：邢承斌(1978-),男,高级工程师,研究方向为港口设备自动化、信息化
参考文献：
[1] 严新平, 李志雄, 袁成清, 等. 考虑船体变形耦合作用的船舶推进系统建模与控制[J]. 船海工程, 2011, 40(1): 60-63.
[2] 谷义恩, 杨智荣, 高海波, 等. 恶劣海况下船舶短期电力负荷预测方法[J]. 海洋工程装备与技术, 2023, 10(1): 117-123.
[3] 钱宇轩, 俞鑫, 柴婷逸, 等. 复杂工况下基于RBF神经网络的全电推进船舶电力负荷预测[J]. 舰船科学技术, 2023, 45(17): 97-101.
QIAN Y X, YU X, CHAI T Y, et al. Power load prediction of fully electric propulsion ships based on RBF neural network under complex working conditions[J]. Ship Science and Technology, 2023, 45(17): 97-101.
[4] 舒方舟, 王莹, 戴晓强, 等. 基于BOA-LSSVM的电力推进船舶电力负荷预测[J]. 舰船科学技术, 2023, 45(20): 159-166.
SHU F Z, WANG Y, DAI X Q, et al. Power load prediction of electric propulsion ships based on BOA-LSSVM[J]. Ship Science and Technology, 2023, 45(20): 159-166.
[5] 宋明达. 基于改进遗传算法优化Elman神经网络的短期电力负荷预测[D]. 衡阳: 南华大学, 2020.
[6] 魏熙朋, 林建辉, 易彩. 基于SAE和BiGRU的滚动轴承剩余寿命预测[J]. 计算机与数字工程, 2024, 52(2): 605-610.
[7] 欧阳卫年, 赵紫昱, 陈渊睿. 自样本特征构造的1DCNN-BiLSTM网侧光伏功率预测[J]. 电力系统及其自动化学报, 2024, 36(3): 151-158.
[8] MENG C, XU C, LEI Q, et al. Balanced joint maximum mean discrepancy for deep transfer learning[J]. Analysis and Applications, 2021, 19(3): 491-508.
[9] LI X, ZHANG Z, GAO L, et al. A new semi-supervised fault diagnosis method via deep CORAL and transfer component analysis[J]. IEEE Transactions on Emerging Topics in Computational Intelligence, 2021, 6(3): 690-699.
[10] 卢新彪, 叶春林, 陈艺森, 等. 一种基于迁移学习的PM2.5浓度预测混合模型[J/OL]. 系统仿真学报, 1-16[2024-06-05].
