Table 9
Diagnostic workflow for fault detection and analysis in photovoltaic (PV) systems, detailing each stage from raw data acquisition to comprehensive fault diagnosis. The integration of early-morning voltage analysis (V*oc,meas) in Machine Learning (ML) algorithms enables differentiation and identification of simultaneous faults such as diode short-circuiting, partial shading, and increased series resistance.
| Stage | Description | Output |
|---|---|---|
| Stage 1: Data Acquisition | Collection of raw PV data, including voltage (V), current (I), and power (P). | Raw PV dataset |
| Stage 2: Data Filtering | Filtering data under specific conditions, such as low irradiance (<100 W/m2) and specific time intervals (05:00–08:00). | Filtered dataset |
| Stage 3: Data Correction | Applying Standard Test Condition (STC) correction to normalize voltage and current values. | V*oc,meas under STC |
| Stage 4: V*oc,meas Extraction | Extraction of the corrected open-circuit voltage to assess early-day voltage behavior. | Boolean value (0 or 1) |
| Stage 5: ML Algorithm Input | Input Boolean values from V*oc,meas routine into the Machine Learning fault detection algorithm. | ML-ready Boolean data |
| Stage 6: Fault Diagnosis | Using the ML algorithm to diagnose potential faults (e.g., diode SC, partial shading, or increased series resistance). | Fault identification and severity |
| Stage 7: Multiple Faults Analysis | Cross-analyzing early-day and full-day voltage behavior to diagnose the simultaneous presence of multiple faults. | Comprehensive fault diagnosis |
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.