Optimized System Management, Monitoring and Predictive Maintenance for Photovoltaic Generators Using Innovative Module Sensor Solutions: ZeroDefect4PV Project

The ZeroDefect4PV project’s development team are using the latest sensor technology to boost large solar power plants’ operational efficiency significantly. This technology enables precise and proactive monitoring that not only reduces maintenance costs dramatically but also maximizes energy efficiency by detecting and correcting malfunctions at an early stage. This increases cost-effectiveness and upholds sustainable energy targets by minimizing downtimes, thus creating direct benefits for plant operators.

Large solar power plants often comprise tens of thousands of modules and components. The failure of individual modules reduces the overall efficiency of interconnected strings severely and is not easy to detect. This results in economic losses and diminished returns from renewable energies. Every kilowatt-hour of free renewable energy not injected into the grid must be compensated by expensive and possibly carbon-emitting plants, something inconsistent with the EU’s sustainability strategy.

The objective of the project is the development and testing of a state-of-the-art integrated sensor prototype for large solar plants. A sensor system for plants combined with an intelligent IoT-based communications architecture will enable the detection of solar generating units’ main inefficiencies. Rapid malfunction detection and predictive maintenance will ensure maximum injection values.

A local client-server architecture combined with interregional control system interfacing will serve as the basis for establishing the conditions for integrated power system management. The overlay sensor network generated will make it possible to explore the wide array of potential uses with AI data analysis, including data imputation, forecasting and anomaly detection. The theoretical concepts developed will be tested and validated in a pilot implementation.

Fraunhofer IFF’s subproject will concentrate on the intercommunication of the sensor prototypes developed and the generation of value-added services from incoming readings. To this end, existing research infrastructures, such as the Energy Operations Center EOC in Magdeburg, will be upgraded to process sensor readings intelligently. Innovative data processing mechanisms will be integrated and used in master control systems to develop schemes for using the newly generated sensor data.

The basic requirement for reliable data processing is a customizable generic data processing mechanism. This will comprise

1. Data imputation: Gaps in datasets hamper subsequent processes, such as corrupted averages. Data imputation assisted by AI closes such gaps with minimum deviations from the assumed true value. Post hoc values will be employed to train the mathematical models so that the data imputation implemented will be self-learning.
2. Anomaly detection: Data deliveries are described by mathematical features. By defining confidence intervals, outliers are indicative of anomalies that ought to be used to localize and classify malfunctions or power plant failures.
3. Predictive maintenance: The trained models will be used to be able to detect anomalies, even in advance. By identifying and classifying malfunctions rapidly, these predictive maintenance mechanisms will reduce the maintenance required.

Another of Fraunhofer IFF’s objectives is the development of schemes for utilizing the new sensor data in advanced control system functions. Standard interfaces will be used to connect the overlay sensor network produced with the Energy Operations Center’s system management. The precise and granular monitoring of the mesh network is intended to generate optimized system models and improve forecasting systems.

The approach’s feasibility and effectiveness will ultimately be demonstrated by a field test demonstrator. Cross-facility communications channels between Germany and Romania will establish highly realistic operating conditions.

By increasing large solar power plants’ full load hours, the proposed method will benefit Germany’s industrial base economically in the medium-term—especially in conjunction with the continuous expansion of solar power during the energy transition. Reduced maintenance expenditures and faster malfunction detection will boost energy yields.

Expected project deliverables

• Development and implementation of an integrated sensor system for the optimization of large solar plant monitoring and maintenance
• AI analytics for the precise detection and classification of system anomalies
• Methods for effective data processing, including data imputation and predictive maintenance
• Demonstrator under real-world operating conditions for the validation of system efficiency