Photovoltaic (PV) soiling loss refers to the buildup of dust on PV systems, which reduces their power output over time. This leads to financial losses and necessitates regular cleaning. To plan cleaning effectively, soiling loss must be continuously monitored. This can be achieved by analysing PV SCADA time-series data along with weather information. The Stochastic Rate and Recovery (SRR) model is among the most widely used for this. The SRR model detects PV cleaning events, which are indicated by improvements in the PV Performance Ratio. A cleaning event refers to a scenario in which a PV system is cleaned naturally (by rain or wind) or manually (by human intervention). The model uses a fixed- parameter approach, which, as highlighted by researchers in the literature, doesn’t work for PV systems that vary in size and location. Hence, using trial-and-error, researchers obtained optimal values of these two parameters. However, in a utility-scale PV plant with a large number of strings, this process becomes time-intensive and infeasible. Researchers at NCPRE have developed a fully automated, data-driven framework to detect PV cleaning events and quantify soiling loss across varying system sizes and locations. The work was carried out by Shoubhik De (PhD scholar) and Bipasha Ghosh (NCPRE intern) under the guidance of Prof. Narendra Shiradkar and Prof. Anil Kottantharayil. Rooftop and utility- scale systems from India and the US were analysed. The proposed approach significantly reduces manual effort in analysing soiling loss for large-scale PV SCADA datasets. In addition, a rainfall sensitivity analysis was performed to estimate the minimum rainfall required for cleaning. The framework can support PV plant Operations & Maintenance (O&M) teams in assessing the effectiveness of manual cleaning and improving decision-making. This work has been published Elsevier’s Solar Energy journal under the title “A data-driven approach to automate cleaning event detection in PV systems for accurate quantification of soiling loss across varying installation capacities”, and it can be accessed at https://doi.org/10.1016/j.solener.2026.114384.