Photovoltaic panel authenticity identification demonstration

How to detect PV modules using imaging spectroscopy?

Therefore, PV modules detection using imaging spectroscopy data should focus on the physical characteristics and the spectral uniqueness of PV modules. PV modules commonly consist of several layers, including fully transparent glass covers for protection, highly transparent EVA films, and the core PV cell.

Are PV modules correctly detected?

In general, most PV modules were correctly detected within the four subsets. Either for the PV power plant of subset A, the campus roofs of subset B, the residential roofs of subset C, the industrial area of subset D, both locations and shapes were correctly detected. Further statistical results were obtained in the following validation process.

Should imaging spectroscopy data be used for PV identification?

This study aims to create greater awareness of the potential importance of imaging spectroscopy data for PV identification. As a physics-based approach, it is robust, transferable, and can provide data on PV coverage on a regional or global scale in short time.

How robust is physics-based detection for PV power plants?

The robustness of the developed and tested novel physics-based detection approach for PV power plants paves the way for more refined investigations towards PV type differentiation and the analysis of the efficiency of such modules. W. Heldens and M. Schroedter-Homscheidt conceived the idea.

Are Avnir and vpep spectral indices useful for polyethylene surface & synthetic turf?

The aVNIR and REND (see Fig. 5) are physically meaningful spectral indices, but cannot constrain polyethylene surface and synthetic turf misclassified as PV modules. For this reason, the PolyEthylene Peak (PEP) index and the PolyEthylene Peak in Visible range (VPEP) were proposed explicitly for polyethylene surface and synthetic turf, respectively.

Can supervised random forest classification automatically identify distributed PV arrays?

Malof et al. (2016a) investigated an approach based on supervised random forest classification to automatically identify distributed PV arrays using color aerial images with a spatial resolution of 0.3 m × 0.3 m, and achieved 72% precision and 80% recall.