<![CDATA[Planning the capacity and predicting the energy production of rural solar power plants requires a reliable approach because the system output is highly influenced by weather variability, while field monitoring data is often limited. The need for reliable electricity in rural areas remains a challenge, especially when the electricity load must be met continuously but the network infrastructure is not yet fully optimized. This study proposes a hybrid pipeline to estimate performance and capacity needs using NASA POWER DAV hourly meteorological data (2019–2024) for Surokonto Wetan Village, Kendal Regency, Indonesia. Stage 1 applies PVLib physics-based simulation (PVWatts with the SAPM temperature model) to generate hourly AC power and energy, then adjusts estimates with a Performance Ratio (PR)=0.80 to represent aggregate system losses. Stage 2 trains a fast prediction model using Random Forest Regression with meteorological and temporal features (GHI, DNI, DHI, air temperature, humidity, month, hour, and lag-1) and compares it with persistence and climatology baselines. The results show that for a requirement of 200 kWh/day, the required solar power plant capacity is 103.66 kWp (P10+margin). In hourly energy prediction, Random Forest showed the best performance with R²=0.9968 in daylight-only evaluation where GHI>0, far exceeding the persistence baseline (R²=0.7298) and climatology (R²=0.8611). These findings indicate that the combination of PVLib and Random Forest based on NASA POWER DAV data can support rapid and measurable estimation of potential, conservative capacity planning, and prediction of rural solar power plant production.]]>
