研究目的
To provide a simple, robust, and low-cost Fault Detection and Classification (FDC) method for PV shading faults based on real electrical measurements (I-V curves) using Principal Component Analysis (PCA).
研究成果
The PCA-based method effectively detects and classifies PV shading faults with a classification success rate above 97%, using only electrical measurements without additional hardware. It is robust to environmental changes and cost-effective, demonstrating feasibility for PV system diagnosis.
研究不足
The method is applied offline for a single PV module; scalability to larger PV systems and real-time application may require further validation. Environmental variations (e.g., irradiance changes) could affect performance, though normalization mitigates this. The study focuses only on shading faults, not other types of PV faults.
1：Experimental Design and Method Selection:
The study uses a data-driven approach with PCA for fault detection and classification. Features are extracted from I-V curves under healthy and shading conditions, and PCA is applied for data representation and classification.
2：Sample Selection and Data Sources:
Experimental data is collected from a 250 Wp PV module (FL60-250MBP) under healthy and four shading configurations. Data includes I-V curves with 101 samples each, measured three times per condition.
3：List of Experimental Equipment and Materials:
PV module (FL60-250MBP), programmable DC electronic load (Chroma 63600), reference cell (RG100 by SOLEMS), 4-wire Pt100 probe, data acquisition system (NI), computer with LABVIEW software, survival blanket for shading.
4：Experimental Procedures and Operational Workflow:
I-V curves are acquired online using the electronic load. Shading is applied by covering PV cells with a survival blanket. Data is collected under varying irradiance and temperature conditions, with measurements repeated for redundancy.
5：Data Analysis Methods:
Features (voltage, current, power, efficiency) are extracted and normalized. PCA is applied to the data matrix, and classification accuracy is evaluated using confusion matrices and Euclidean distance metrics.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
