Soiling in photovoltaic systems is a recurring problem that reduces energy generation and demands efficient operation and maintenance (O&M) strategies. In this context, this paper proposes a machine learning-based approach to identify dirt levels and generate cleaning alerts using operational and weather data. Initially, the models were evaluated with a decision threshold ranging from 0.5 to 0.7, using only operational features. Subsequently, the inclusion of weather features was tested, which improved the models’ performance and enabled the selection of the best models for the exhaustive features search step. The models analyzed in this step were Extra Trees, Histogram-based Gradient Boosting, Extreme Gradient Boosting, and Random Forest. Exhaustive analysis further improved model performance, as indicated by global metrics and ROC curves. The Extra Trees model with a threshold of 0.5 showed the best performance and was selected as the final configuration, achieving an accuracy of 0.9884 and an AUC-ROC of 0.9957. Finally, the selected model was applied to determine daily soiling levels and trigger alerts based on temporal persistence, indicating its potential to support predictive O&M decisions and cleaning actions in PV systems.

Although Brazil is in the last mile of achieving universal electricity access, the Amazon region represents the country’s most critical electrification frontier, where remote energy expansion has largely failed to integrate clean alternatives. This study presents the design and techno-economic assessment of a hybrid off-grid microgrid tailored for the Amazon, incorporating a multi-source architecture that combines micro-hydro (MH), wind power (WT), solar photovoltaics (PV), diesel generation (DG), and battery energy storage systems (BESS). Three locations in Amazonas, Roraima, and Tocantins were evaluated under three scenarios: a hybrid system with DG, a fully renewable system without DG, and a green scenario that accounts for demand-side flexibility and selected technical and economic conditions. Optimization was performed using the HOMER tool for a typical isolated community. The hybrid configuration achieved a DG generation share of 0.7 (RSS, DG–WT–PV–MH–BESS), 0.9% (Jalapão, DG–WT–PV–MH–BESS), and 6.8% (Labrea, DG–PV–BESS). The fully renewable scenario resulted in Levelized Cost of Energy (COE) values ranging from 0.146 USD/kWh (RSS) to 0.388 USD/kWh (Lábrea), while the green scenario yielded COEs between 0.089 USD/kWh (RSS) and 0.190 USD/kWh (Lábrea). This study reveals the role of MH in reducing storage requirements, installed capacity, and system cost. The results are promising for presenting a competitive cost of energy.