为捕捉序列中相隔较远信息之间的联系,设计了基于门控循环单元深度学习算法的舰船用电数据挖掘方法。利用采集终端实时采集发电机组、甲板机械等设备的原始舰船用电数据;将原始用电数据转换为标准格式,根据舰船电力系统电能质量分析与故障诊断等需求,运用深度学习算法中的门控循环单元,在原始用电数据内捕捉序列中相隔较远信息之间的联系,挖掘舰船用电数据规律特征,为舰船用电管理优化提供决策支持。实验证明:该方法可有效实时采集原始舰船用电数据,并有效挖掘舰船用电数据规律特征;应用该方法后,可有效提升舰船电力系统负荷预测精度。
Designed a deep learning algorithm for ship electricity data mining based on gated recurrent units to capture the connections between distant information in the sequence. Real time collection of raw ship electricity data from generator sets, deck machinery, and other equipment using collection terminals; Convert the original power consumption data into a standard format, and use the gate controlled loop unit in deep learning algorithms to capture the connections between distant information in the sequence within the original power consumption data, based on the requirements of power quality analysis and fault diagnosis of ship power systems. Explore the regular features of ship power consumption data, and provide decision support for optimizing ship power management. Experimental results have shown that this method can effectively collect real-time raw ship electricity data and effectively mine the regular features of ship electricity data; After applying this method, the accuracy of load forecasting for ship power systems can be effectively improved.
2025,47(9): 170-174 收稿日期:2024-12-28
DOI:10.3404/j.issn.1672-7649.2025.09.029
分类号:TP39
基金项目:山东省教育厅本科教学改革研究项目(M2022328)
作者简介:王晓辉(1983-),女 ,硕士,副教授,研究方向为数据分析及数据挖掘
参考文献:
[1] 周寅正, 陈俐. 基于模型预测控制的双机组混合动力船舶能量管理研究[J]. 中国舰船研究, 2024, 19(z1): 74-83.
[2] 詹锦皓, 李维波, 李齐, 等. 基于比例伪时序算法的舰船电力风险评估系统[J]. 中国舰船研究, 2022, 17(1): 176-186.
[3] 易庚, 何琳, 刘锦明, 等. 基于迁移学习算法的电力数据挖掘模型[J]. 沈阳工业大学学报, 2023, 45(5): 510-515.
[4] 张悦, 宋运忠. 一种基于类内类间协同的用电数据聚类选择[J]. 武汉大学学报(工学版), 2022, 55(5): 493-502.
[5] 柳劲松, 王颖, 熊春晖, 等. 基于日负荷曲线的用电模式提取与聚类分析[J]. 电工电能新技术, 2023, 42(5): 57-63.
[6] 盛锦超, 杜明晶, 李宇蕊, 等. 结合柯西核的分类型数据密度峰值聚类算法[J]. 计算机工程与应用, 2022, 58(18): 162-171.
[7] 宋铁维, 施伟锋, 毕宗, 等. 基于2D-ResNet的船舶电力系统电能质量扰动识别[J]. 电力系统保护与控制, 2022, 50(10): 94-103.