针对发电汽轮机运行参数高度耦合、故障特征参数预测困难的问题，综合运用中值回归经验模态分解、卷积神经网络和双向长短期记忆模型，提出一种适用于发电汽轮机的多特征故障预测方法。首先，通过中值回归经验模态分解（Median Regression Empirical Mode Decomposition, MREMD）对故障相关参数进行趋势提取；然后，以各参数运行趋势为训练集的输入特征向量，构建可以提取训练集时空特征的卷积-双向长短期记忆（Convolution-Bidirectional Long Short-Term Memory, CNN-BiLSTM）模型；为提高模型的收敛速度，采用麻雀搜索算法（Sparrow Search Algorithm, SSA）对模型超参数进行优化。经实际案例验证，该方法可有效通过故障相关参数的波动预测故障参数的发展趋势，为系统的预警和应急处置提供参考。
Aiming at the problem that the operating parameters of power generation turbines are highly coupled and the fault characteristic parameters are difficult to predict, using median regression empirical mode decomposition, convolutional neural network and bidirectional long short-term memory model, a multi-feature fault prediction method suitable for power generation turbines is proposed. Firstly, Median Regression Empirical Mode Decomposition is used to extracted tendency from fault-related parameters. then the running trend of each parameter is used as the input feature vector of the training set, constructed the Convolution-Bidirectional long short-term memory model that can extract the spatiotemporal features of the training set. In order to improve the convergence speed of the model, the Sparrow Search Algorithm (SSA) was used to optimize the hyperparameter of the model. Through practical case verification, this method can effectively predict the development trend of fault parameters through fluctuations in fault related parameters, which can provide some reference for the early warning and emergency response of system.
2024,46(11): 125-133 收稿日期：2023-07-24
DOI：10.3404/j.issn.1672-7649.2024.11.023
分类号：N37；TK221
基金项目：国家自然科学基金面上项目（51909254）；海军工程大学自主研发基金资助项目（425317T014）
作者简介：卓越(1999-),男,硕士研究生,研究方向为热力系统状态监测与故障诊断
参考文献：
[1] KARIM S, MAJID S K, MOJTABA K. Fault detection and diagnosis of an industrial steam turbine using a distributed configuration of adaptive neuro-fuzzy inference systems[J]. Simulation Modelling Practice and Theory, 2011, 19(5): 1280-1293.
[2] 郑奕扬, 倪何, 金家善. 基于MSOP的蒸汽动力系统单参数运行稳定性评估方法[J]. 上海交通大学学报, 2021, 55(11): 1438-1444.
ZHENG Yiyang, NI He, JIN Jiashan. An operation stability assessment method of a single-parameter in steam power system based on MSOP[J]. Journal of Shanghai Jiao Tong University, 2021, 55(11): 1438-1444.
[3] 孟忍, 董学平, 甘敏. 基于KPCA和BiLSTM的分解炉出口温度预测[J]. 合肥工业大学学报(自然科学版), 2023, 46(2): 169-174.
MENG Ren, DONG Xueping, GAN Min. Prediction of calciner outlet temperature based on KPCA and BiLSTM[J]. Journal of Hefei University of Technology (natural science), 2023, 46(2): 169-174.
[4] 邢红涛, 郭江龙, 刘书安, 等. 基于CNN-LSTM混合神经网络模型的NOx排放预测[J]. 电子测量技术, 2022, 45(2): 98-103.
XING Hongtao, GUO Jiangjiang, LIU Shuan, et al. NOx emission forecasting based on CNN-LSTM hybrid neural network[J]. Electronic Measurement Technology, 2022, 45(2): 98-103.
[5] 闾曾怡, 王高峰, 杨瑶, 等. 基于CNN-LSTM 深度学习的燃烧不稳定性检测方法研究[J]. 燃烧科学与技术, 2022, 28(2): 119-125.
LV Zengyi, WANG Gaofeng, YANG Yao , et al. Combustion instability detection method based on CNN-LSTM deep learning[J]. Journal of Combustion Science and Technology, 2022, 28(2): 119-125.
[6] XIANG Ling, WANG Penghe, YANG Xin, et al. Fault detection of wind turbine based on SCADA data analysis using CNN and LSTM with attention mechanism[J]. Measurement, 2021(175): 1-8.
[7] WANG Jianxing, GUO Lili, ZHANG Chengying, et al. Thermal power forecasting of solar power tower system by combining mechanism modeling and deep learning method[J]. Energy, 2020(208): 1-16.
[8] 肖鹏飞, 倪何, 金家善. 基于MWSA的热力系统单参数时序预测方法[J]. 上海交通大学学报, 2023, 57(1): 36-44.
XIAO Pengfei, NI He, JIN Jiashan. Sequential prediction method of single parameter for thermal system based on MWSA[J]. Journal of Shanghai Jiao Tong University, 2023, 57(1): 36-44.
[9] 王敏, 邹婕, 王惠琳, 等. 基于改进的AHP-CRITIC-MARCOS配电网设备风险评估方法[J]. 电力系统保护与控制, 2023, 51(3): 164-172.
WANG Min, ZOU Jie, WANG Huilin, et al. Improved AHP-CRITIC-MARCOS-based risk assessment method for distribution network equipment[J]. Power System Protection and Control, 2023, 51(3): 164-172.
[10] 董渊博, 茅大钧, 章明明. 基于CNN-LSTM的燃气轮机NOx排放预测研究[J]. 热能动力工程, 2021, 36(9): 132-138.
DONG Yuanbo, MAO Dajun, ZHANG Mingming. Research on NOx emission prediction of gas turbine based on CNN-LSTM[J]. Jouranl of Engineering for Thermal Energy and Power, 2021, 36(9): 132-138.
[11] CARLOS I G, FRANCESCO G, ANTONIO L. A comparison of power quality disturbance detection and classification methods using CNN, LSTM and CNN-LSTM[J]. Applied Sciences, 2020, 10(19): 6755.
[12] 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.
REN Jianji, WEI Huihui, ZOU Zhuolin, et al. Ultra-short-term power load forecasting based on CNN-BiLSTM-Attention[J]. Power System Protection and Control, 2022, 50(8): 108-116.
[13] 张加劲. 基于注意力机制和CNN-BiLSTM 模型的航空发动机剩余寿命预测[J]. 电子测量与仪器学报, 2022, 36(8): 231-237.
ZHANG Jiajing. Remaining useful life estimation of aeroengine based on CNN-BiLSTM and attention mechanism[J]. Jouranl of Electronic Measurement and Instrumentation, 2022, 36(8): 231-237.
[14] XUE Jiankai, SHEN Bo. A novel swarm intelligence optimization approach: sparrow search algorithm[J]. Systems Science & Control Engineering an Open Access Journal, 2020, 8(1): 22-34.
