基于多策略融合的光伏系统故障诊断方法

doi:10.16182/j.issn1004731x.joss.25-0471

摘要/Abstract

摘要：

为解决光伏系统故障频发的问题，提出一种基于改进旅鼠算法优化多模态融合故障诊断模型。通过马尔可夫转换场将光伏电流、电压一维时序信号转换为二维图像，利用多尺度卷积神经网络挖掘原始波形的空间特征；采用BiGRU提取原始波形的时序动态特征，通过特征融合层实现时空特征的互补增强。引入改进旅鼠算法对BiGRU隐藏层神经元数量、模型的学习率等参数进行自适应优化，结合注意力机制强化故障敏感特征的权重分配。仿真实验结果表明：所提模型仿真与实测数据的诊断准确率分别达到97.9%和95.4%，相较对比模型，诊断准确率提升最高4.1%，为光伏系统智能化运维提供了新的技术路径。

关键词: 马尔可夫场, 双向门控循环单元, 改进旅鼠算法, 故障诊断, MPPT(maximum power point tracking)故障, 逆变器故障

Abstract:

To address the problem of frequent PV system faults, a multimodal fusion fault diagnosis model based on the optimization of the improved lemming algorithm was proposed. The one-dimensional time series signals of PV currents and voltages were converted into two-dimensional images by Markov transformation field, and the spatial features of the original waveforms were mined by using multiscale CNN (MCCNN); BiGRU was used to extract the temporal dynamic features of the original waveforms, and complementary enhancement of the temporal and spatial features was realized by the feature fusion layer. The improved lemming algorithm was innovatively introduced to adaptively optimize parameters such as the number of neurons in the hidden layer of BiGRU and the learning rate of the model, and the weight assignment of fault-sensitive features was enhanced by combining the assistive technology. The results of simulation experiments have shown that the diagnostic accuracies of the proposed model simulation and the measured data reach 97.9% and 95.4%, respectively.Compared with the comparative models, the diagnostic accuracy is improved by up to 4.1%. The proposed model provides a new technical path for intelligent operation and maintenance of PV systems.

Key words: Markov transition field, BiGRU, improved lemming algorithm, fault diagnosis, maximum power point tracking (MPPT) fault, inverter fault

中图分类号:

TP391.9

李斌,王于绰 . 基于多策略融合的光伏系统故障诊断方法[J]. 系统仿真学报, 2025, 37(12): 3018-3032.

Li Bin,Wang Yuchuo . Fault Diagnosis Method for Photovoltaic Systems Based on Multi-strategy Fusion[J]. Journal of System Simulation, 2025, 37(12): 3018-3032.

图/表 19

图1

图2

图3

图4

图5

表1

图6

表2

图7

表3

图8

表4

光伏系统工作状态

标签	工作状态	主要影响参数
K0	正常运行
K1	光伏阵列故障	$U p v, I p v, U d c$
K2	逆变器故障	$U d c, I a b c, U a b c$
K3	并网异常	$I a b c, U a b c$
K4	MPPT轻度故障	$U d c, I a b c, U a b c$
K5	MPPT严重故障	$U p v, U d c, I a b c, U a b c$
K6	光伏阵列及逆变器故障	$U p v, I p v, U d c, I a b c, U a b c$
K7	光伏阵列及MPPT轻度故障	$U p v, I p v, U d c, I a b c, U a b c$

表4

图9

表5

图10

表6

表7

图11

图12

参考文献 23

[1]	Shouquat Hossain MD, Wadi Al-Fatlawi A, Kumar Laveet, et al. Solar PV High-penetration Scenario: An Overview of the Global PV Power Status and Future Growth[J]. Energy Systems, 2024: 1-57.
[2]	Xie Xiangying, Meng Fanteng, Luo Xin, et al. Fault Diagnosis and Detection Based on Efficiency Loss Test of Photovoltaic Power Station[C]//IOP Conference Series: Earth and Environmental Science. Bristol: IOP Publishing, 2020: 012060.
[3]	Baltacı Özge, Kıral Zeki, Dalkılınç Konuralp, et al. Thermal Image and Inverter Data Analysis for Fault Detection and Diagnosis of PV Systems[J]. Applied Sciences, 2024, 14(9): 3671.
[4]	陈勇, 刘威. 基于无人机图像识别的光伏组件故障诊断系统研究[J]. 通讯世界, 2024, 31(9): 109-111.
[5]	Hong Yingyi, Pula Rolando A. Diagnosis of Photovoltaic Faults Using Digital Twin and PSO-optimized Shifted Window Transformer[J]. Applied Soft Computing, 2024, 150: 111092.
[6]	Meng B, Loonen R C G M, Hensen J L M. Leveraging Dynamic Power Benchmarks and CUSUM Charts for Enhanced Fault Detection in Distributed PV Systems[J]. Energy Conversion and Management, 2024, 314: 118692.
[7]	Ahmed Faris Amiri, Oudira Houcine, Chouder Aissa, et al. Faults Detection and Diagnosis of PV Systems Based on Machine Learning Approach Using Random Forest Classifier[J]. Energy Conversion and Management, 2024, 301: 118076.
[8]	Chahine Khaled. Tree-based Algorithms and Incremental Feature Optimization for Fault Detection and Diagnosis in Photovoltaic Systems[J]. Eng, 2025, 6(1): 20.
[9]	王梦圆, 徐潇源, 严正. 适用小样本的并网光伏阵列故障诊断方法[J]. 电网技术, 2024, 48(2): 587-596.
	Wang Mengyuan, Xu Xiaoyuan, Yan Zheng. Simultaneous Fault Diagnosis of PV Array Operation Considering Small Sample Sizes[J]. Power System Technology, 2024, 48(2): 587-596.
[10]	公铮, 丁家伟, 刘允浩, 等. 基于增量学习的光伏并网逆变器故障诊断[J]. 电力系统及其自动化学报, 2025, 37(2): 48-57.
	Gong Zheng, Ding Jiawei, Liu Yunhao, et al. Fault Diagnosis of Photovoltaic Grid-connected Inverter Based on Incremental Learning[J]. Proceedings of the CSU-EPSA, 2025, 37(2): 48-57.
[11]	黄晨昊, 高伟. 基于超声波传感器与孤立森林的光伏系统串联电弧故障诊断方法[J]. 电气技术, 2025, 26(5): 10-16, 26.
	Huang Chenhao, Gao Wei. Series Arc Fault Diagnosis Method for Photovoltaic System Based on Ultrasonic Sensor and Isolation Forest[J]. Electrical Engineering, 2025, 26(5): 10-16, 26.
[12]	李岩, 刘鑫月, 乔俊杰, 等. 基于1DCNN和D-S多信息融合的光伏系统直流母线串联电弧故障检测[J]. 电工电能新技术, 2024, 43(5): 58-67.
	Li Yan, Liu Xinyue, Qiao Junjie, et al. Series Arc Fault Detection of DC Bus of Photovoltaic System Based on 1DCNN and D-S Multi-information Fusion[J]. Advanced Technology of Electrical Engineering and Energy, 2024, 43(5): 58-67.
[13]	钱倍奇, 陈谦, 李宗源, 等. 基于马尔可夫转换场与多头注意力机制的电能质量扰动分类方法[J]. 电网技术, 2024, 48(2): 721-729.
	Qian Beiqi, Chen Qian, Li Zongyuan, et al. Power Quality Disturbances Classification Based on Markov Transition Field and Multi-head Attention[J]. Power System Technology, 2024, 48(2): 721-729.
[14]	刘沛津, 付雪峰, 孙浩峰, 等. 一种融合CNN与Transformer的高鲁棒性目标跟踪算法[J]. 系统仿真学报, 2024, 36(8): 1854-1868.
	Liu Peijin, Fu Xuefeng, Sun Haofeng, et al. A Highly Robust Target Tracking Algorithm Merging CNN and Transformer[J]. Journal of System Simulation, 2024, 36(8): 1854-1868.
[15]	仝航, 杨燕, 江永全. 检测脑电癫痫的多头自注意力机制神经网络[J]. 计算机科学与探索, 2023, 17(2): 442-452.
	Tong Hang, Yang Yan, Jiang Yongquan. Multi-head Self-attention Neural Network for Detecting EEG Epilepsy[J]. Journal of Frontiers of Computer Science and Technology, 2023, 17(2): 442-452.
[16]	丰胜成, 郭继成, 付华, 等. 基于ISSA与GRU分位数回归的风电功率概率密度预测[J]. 电工电能新技术, 2023, 42(10): 55-65.
	Feng Shengcheng, Guo Jicheng, Fu Hua, et al. Wind Power Probability Density Prediction Based on ISSA and GRU Quantile Regression[J]. Advanced Technology of Electrical Engineering and Energy, 2023, 42(10): 55-65.
[17]	余波, 高学军, 王灿, 等. 基于行波特征与KOA-CNN-BiGRU-AM的柔直输电线路故障诊断[J]. 电力工程技术, 2025, 44(2): 185-196.
	Yu Bo, Gao Xuejun, Wang Can, et al. Fault Diagnosis of MMC-MTDC Based on Traveling Wave Characteristics and KOA-CNN-BiGRU-AM[J]. Electric Power Engineering Technology, 2025, 44(2): 185-196.
[18]	林云. 基于Transformer多头注意力机制的文本分类模型[J]. 现代信息科技, 2025, 9(6): 100-104.
	Lin Yun. Text Classification Model Based on Transformer Multi-head Attention Mechanism[J]. Modern Information Technology, 2025, 9(6): 100-104.
[19]	Xiao Yaning, Cui Hao, Ruba Abu Khurma, et al. Artificial Lemming Algorithm: a Novel Bionic Meta-heuristic Technique for Solving Real-world Engineering Optimization Problems[J]. Artificial Intelligence Review, 2025, 58(3): 84.
[20]	李鹏, 陈守静, 杨山山, 等. 基于Logistic映射的果蝇算法优化Otsu图像分割方法[J]. 国外电子测量技术, 2022, 41(7): 9-17.
	Li Peng, Chen Shoujing, Yang Shanshan, et al. Otsu Image Segmentation Method Optimized by Fruit Fly Optimization Algorithm Based on Logistic Mapping[J]. Foreign Electronic Measurement Technology, 2022, 41(7): 9-17.
[21]	李丹丹, 朱石磊, 李仲康, 等. 多策略融合的改进金豺优化算法[J]. 传感器与微系统, 2025, 44(1): 127-130.
	Li Dandan, Zhu Shilei, Li Zhongkang, et al. Multi-strategy Fusion Improved Golden Jackal Optimization Algorithm[J]. Transducer and Microsystem Technologies, 2025, 44(1): 127-130.
[22]	傅洪全, 田丰伟, 周磊, 等. 光伏发电系统中的设备故障与对策分析[J]. 集成电路应用, 2024, 41(7): 182-183.
	Fu Hongquan, Tian Fengwei, Zhou Lei, et al. Analysis of Equipment Faults and Countermeasures in Photovoltaic Power Generation Systems[J]. Application of IC, 2024, 41(7): 182-183.
[23]	石凯, 李光明, 吴甫, 等. 光伏发电系统电气设备故障原因及处理措施分析[J]. 太阳能, 2023(3): 58-67.
	Shi Kai, Li Guangming, Wu Fu, et al. Analysis on Causes and Treatment Measures of Electrical Equipment Failure in PV Power Generation System[J]. Solar Energy, 2023(3): 58-67.

测试函数	理论值	测试函数	理论值
F1	100	F4	1 700
F2	1 100	F5	2 200
F3	700	F6	2 400

函数	指标	IALA	ALA	BWO	SSOA	HHO	WOA	IVY
F1	最优值	1.07×10²	1.10×10²	4.13×10⁹	4.84×10⁹	1.64×10⁵	8.26×10⁵	2.04×10²
	平均值	3.56×10²	7.85×10²	7.72×10⁹	1.09×10¹⁰	6.82×10⁵	9.98×10⁶	8.76×10⁷
	标准差	7.42×10²	1.11×10³	2.07×10⁹	2.81×10⁹	9.20×10⁵	1.69×10⁷	4.50×10⁸
F2	最优值	1.21×10³	1.22×10³	2.20×10³	2.43×10³	1.57×10³	1.50×10³	1.23×10³
	平均值	1.62×10³	1.62×10³	2.56×10³	3.32×10³	2.08×10³	2.26×10³	2.20×10³
	标准差	2.25×10²	2.31×10²	2.80×10²	2.95×10²	2.60×10²	2.94×10²	3.97×10²
F3	最优值	7.09×10²	7.15×10²	7.73×10²	7.97×10²	7.20×10²	7.40×10²	7.31×10²
	平均值	7.25×10²	7.27×10²	7.99×10²	8.27×10²	7.92×10²	7.83×10²	7.87×10²
	标准差	6.12	6.40	9.87	1.66×10¹	2.17×10¹	3.32×10¹	2.60×10¹
F4	最优值	1.73×10³	1.74×10³	7.16×10³	2.48×10⁵	3.98×10³	3.15×10³	4.09×10⁴
	平均值	1.81×10³	1.82×10³	2.15×10⁵	5.42×10⁵	5.55×10⁴	2.24×10⁵	4.50×10⁵
	标准差	5.78×10¹	6.70×10¹	1.15×10⁵	1.25×10⁵	4.43×10⁴	5.09×10⁵	2.41×10⁵
F5	最优值	2.20×10³	2.21×10³	2.36×10³	2.70×10³	2.31×10³	2.27×10³	2.30×10³
	平均值	2.29×10³	2.30×10³	2.62×10³	3.18×10³	2.42×10³	2.41×10³	2.32×10³
	标准差	2.64×10¹	1.64×10¹	1.90×10²	3.46×10²	3.87×10²	3.35×10²	2.39×10¹
F6	最优值	2.50×10²	2.73×10³	2.59×10³	2.84×10³	2.75×10³	2.54×10³	2.50×10³
	平均值	2.64×10³	2.75×10³	2.73×10³	2.95×10³	2.82×10³	2.77×10³	2.71×10³
	标准差	3.43×10¹	7.34	8.02×10¹	5.12×10¹	4.00×10¹	4.76×10¹	9.57×10¹

模型	原模型	无AT模型
训练量	640	640
测试量	160	160
训练时间/min	6.2	5.4
测试时间/s	1.7	1.1
参数量	2.76×10⁷	2.48×10⁷
准确率/%	98.1	92.5

参数	理论值
功率/W	1 050
开路电压/V	109.23
短路电流/A	13.82
工作电压/V	90.84

参数	理论值
功率/W	350
开路电压/V	36.41
短路电流/A	13.82
工作电压/V	30.28
功率容差/%	±5