| 引用本文: | 魏书荣,殷世杰,闫梦飞,等.基于改进生成对抗网络的海上风电机组故障数据增强及诊断[J].电力系统保护与控制,2025,53(1):114-124.[点击复制] |
| WEI Shurong,YIN Shijie,YAN Mengfei,et al.Offshore wind turbine fault data enhancement and diagnosis based on an improved generative adversarial network[J].Power System Protection and Control,2025,53(1):114-124[点击复制] |
|
| 摘要: |
| 海洋复杂运行环境下，风电机组故障多样，故障有效样本数据明显不足，严重影响了故障诊断效果。为解决海上风电运行数据及故障样本积累不足的问题，提出了一种基于GRA-rACGAN生成对抗网络的数据增强方法，可有效扩充海上风机异常工况数据，并通过实际运行数据进行诊断验证。首先，对SCADA系统采集的数据进行灰色关联分析(grey relation analysis, GRA)，筛选出与海上风电机组运行状态高度相关的状态变量，对数据进行归一化处理，将特征的最小最大范围添加为每个样本的两个附加属性，避免异常数据干扰，提高数据生成能力。然后，将筛选出的状态变量数据集输入至改进型辅助分类器，采用生成对抗网络进行学习，扩充故障数据。最后，以海上风机实际运行数据的增强结果作为样本进行故障诊断，检验故障数据增强方法的可靠性。通过对海上风电场的实际运行数据实测结果表明，本模型相比于传统数据增强技术可以有效地生成故障样本，提高故障诊断的准确率与稳定性，为海上风机故障的准确预警提供技术支撑。 |
| 关键词: 海上风机 数据增强 灰色关联分析 辅助分类器生成对抗网络 故障诊断 |
| DOI：10.19783/j.cnki.pspc.240523 |
| 投稿时间：2024-04-29修订日期：2024-07-24 |
| 基金项目:国家自然科学基金项目资助(52377063)；上海市教委自然科学重大项目资助(2021-01-07-00- 07-E00122)；上海科技创新行动计划资助(22dz1206100)；上海高校特聘教授(东方学者)(TP2020066) |
|
| Offshore wind turbine fault data enhancement and diagnosis based on an improved generative adversarial network |
| WEI Shurong1,YIN Shijie1,YAN Mengfei2,ZHOU Hailin1 |
| (1. Engineering Research Center of Offshore Wind Technology, Ministry of Education, Shanghai University of Electric Power,
Shanghai 200090, China; 2. Shinan Power Supply?Company, SMEPC, Shanghai 201100, China) |
| Abstract: |
| Given the complex operational environment of the ocean, wind turbine faults are diverse, and the effective sample data of faults are obviously insufficient. This seriously affects fault diagnosis results. To solve the problem of insufficient accumulation of offshore wind turbine operational data and fault samples, a data enhancement method based on a GRA-rACGAN generative adversarial network is proposed. This can effectively expand offshore wind turbine abnormal working condition data and carry out diagnosis validation through actual operational data. First, grey relation analysis (GRA) is performed on the data collected by the SCADA system to screen out the state variables that are highly correlated with the operating state of the wind turbines, normalize the data, and add the minimum and maximum ranges of the features as two additional attributes for each sample to avoid the interference of abnormal data and to improve the ability of data generation. Then, the filtered state-variable dataset is fed into an improved auxiliary classifier that employs a generative adversarial network for learning and expanding the fault data. Finally, the reliability of the fault data enhancement method is tested using the enhancement results of actual offshore wind turbine operation data as a sample for fault diagnosis. The measured results of the actual operation data of offshore wind farms show that this model can more effectively generate fault samples and improve the accuracy and stability of fault diagnosis than traditional data enhancement techniques, providing technical support for the accurate early warning of offshore wind turbine faults. |
| Key words: offshore wind turbine data enhancement grey relation analysis ACGAN fault diagnosis |
