A research team, spearheaded by Professor Lan Dongchen from Zhejiang University, has introduced a groundbreaking method that facilitates the quantitative analysis of internal power loss in photovoltaic (PV) modules through a single luminescence image. This innovative technique leverages a lightweight artificial intelligence (AI) system to pinpoint the primary degradation mechanisms within the modules. Subsequently, it utilizes a device physical model to deduce the power loss distribution, thereby circumventing the necessity for vast amounts of training data.

Upon validation with 300 crystalline silicon PV modules sourced from diverse locations, the method demonstrated an impressive overall error margin of merely 0.5% between the reconstructed and measured power outputs. The utilization of a single-image measurement approach effectively circumvents alignment errors typically associated with multi-image capture techniques, thereby enhancing the quantitative accuracy of the analysis.

Furthermore, the team has successfully validated the image collection process using drones at operational PV power stations. This advancement enables a detailed examination of internal cell power loss and the quantification of prevalent degradation types, thereby significantly elevating inspection efficiency and curtailing costs for large-scale power stations.

At present, this method has undergone preliminary validation on perovskite PV modules. The team envisions furthering its implementation in the operations and maintenance of power stations, adapting it to accommodate a broader spectrum of PV module types, and bolstering decision-making processes for large-scale PV power station operations and power generation performance evaluations.