Photovoltaic Monitoring Sensor System Monitors PV Systems at Module Level

Related Vendors

FAULHABER Drive Systems

Taiwan Machine Tool & Accessory Builders' Association (TMBA)

Until now, large-scale photovoltaic power plants have usually only been monitored at string or inverter level. As a result, defects in individual solar modules often remain undetected for a long time. A sensor system now provides high-resolution diagnostic data down to module level.

The failure of individual solar modules, defective bypass diodes or electrical connection errors: In large photovoltaic systems with tens of thousands of components, undetected local defects quickly add up to massive yield losses.

The problem with current system technology is that established monitoring methods usually only evaluate the data that the inverter supplies across all strings. Individual module statuses remain invisible in this approach. Optical inspections via drone or infrared thermography can detect hotspots, but fail in the event of electrical defects, degradation or faults on the backsheet. To detect these blind spots, a consortium led by the Fraunhofer IFF is currently developing a highly granular sensor system for predictive maintenance at module level in the ZeroDefect4PV project.

The Hardware is Directly on the Panel

At the heart of the system are data collection units (DCU), which are installed directly on the back of the individual photovoltaic modules. These sensor nodes continuously record the direct voltage and direct current of the respective solar panel. In addition, the module temperature is measured in order to deduce thermal loads and fault conditions. External environmental data such as solar radiation is provided by a separate weather station.

From the Mesh Network to the Cloud

The system solves the communication challenge in large-scale PV parks via a hierarchical radio network: the prototype sensors are organized as a master-slave architecture in a mesh sensor network. They use the ESP-NOW protocol to communicate with each other. The aggregated data is then transmitted to higher-level gateways using the long-range, low-power LoRaWAN protocol. From there, the synchronized values are sent to a central data platform for evaluation.

While conventional inverter measurements only register a general drop in performance on a string (mismatches), the new system classifies the specific source of the fault. The algorithms are trained on the basis of fault patterns and detect anomalies in the current and voltage curve.

The AI can not only differentiate between thermal anomalies, mechanical damage (cell cracks), soiling or an electrical defect in the bypass diodes, but can also precisely localize the problem. "Unlike measuring the inverter, our system classifies the faults and precisely identifies where they occur," explains Dr. Christoph Wenge, researcher at the Fraunhofer IFF. "A wide variety of faults can occur in the solar panels connected in series. These can occur not only on the modules themselves, but also in the bypass diodes, in the cable or in the mounting system."

Technicians in the control room then receive concrete recommendations for action via implemented assistance functions, such as the targeted replacement of a specific module.

Field Tests Under Real Conditions

The system is currently leaving the laboratory phase. Master and slave sensors have already been tested for measurement accuracy and communication reliability. Pilot tests are currently underway at a Fraunhofer IFF research facility in Magdeburg (Germany), where the AI is being trained with targeted shading by leaves, among other things. At the same time, hardware validation is starting in a PV field in Turkey, while the training of the AI models is being flanked by the project partner BEIA Consult International with real inverter data in Bucharest.

"With the pilot installations, we are validating our system under real conditions," says Dr. Wenge, summing up the current status. "This allows us to iteratively optimize hardware, communication and data models and ultimately evaluate the scalability for large PV parks." (heh)

Tips & tricks for plastics professionals

Future trends in additive manufacturing