End User
PPC
Description
This service quantifies energy losses in PV systems due to soiling (accumulation of dust, dirt, etc.) on the panels. By analyzing historical data, including PV generation and soiling sensor measurements, the AI models estimate the percentage of power reduction caused by soiling. The service provides insights into the optimal timing of cleaning interventions, maximizing energy yield and minimizing operational costs.
Core Capabilities
Optimization & Decision Support
Business Need
The AI-Based PV Soiling Loss Assessment service addresses the significant problem of reduced energy production in photovoltaic (PV) systems caused by soiling. Soiling, the accumulation of dust, dirt, and other pollutants on PV panels, leads to decreased energy output, directly resulting in lost revenue for PV plant operators. Furthermore, without accurate soiling loss assessment, scheduling cleaning activities becomes suboptimal, leading to either unnecessary cleaning costs incurred by cleaning too frequently or prolonged energy losses caused by not cleaning often enough. This lack of accurate assessment also hinders the ability to precisely evaluate the true performance of PV systems, making it difficult to identify other potential issues or accurately predict future energy production.
Key Performance Indicators
Accuracy of Soiling Loss Estimation: Mean Absolute Error (MAE) or Root Mean Squared Error (RMSE) between the predicted soiling loss and the actual soiling loss (if a reliable ground truth measurement is available, from PVSyst calculations).
Accuracy of Soiling Loss Estimation: R-squared value to indicate how well the model fits the data.
Data Provided
Historical generation data of the inverters - CSVs - unknown size - no documentation available - 15 min resolution
Soiling sensor data - CSVs - unknown size - no documentation available - 15 min resolution
TEF
TEF RES