TY  - JOUR
AU  - Messias, Leonardo Alves
AU  - Domingos, José Luis
AU  - Mendes, Thiago Augusto
AU  - da Costa, Bruno Barzellay Ferreira
AU  - Maciel, Ana Carolina Fernandes
AU  - Mendes, Saymon Fonseca Santos
AU  - de Aquino Gomes, Raphael
PY  - 2026
DA  - 2026/04/03
TI  - Photovoltaic Power Forecasting Without Local Data: A Spatially-Aware Approach Using Neighboring Plants
JO  - Journal of Energy and Power Technology
SP  - 007
VL  - 08
IS  - 02
AB  - The increasing demand for renewable energy sources has intensified the adoption of photovoltaic systems. This study proposes predictive models for solar power generation that operate without dependence on on-site meteorological stations. The proposed approach integrates generation data from geographically distributed plants and accounts for the distance to meteorological stations when constructing climatic variables. Two machine learning techniques, Random Forest (RF) and Long Short-Term Memory (LSTM) networks, were evaluated. The RF model achieved R2 > 0.90 with lower MAE and RMSE values for 24-hour prediction windows, whereas the LSTM model demonstrated superior performance for extended horizons (48 hours). Moreover, the proposed models effectively identified anomalies and maintained robust predictive accuracy even when utilizing data from meteorological stations located up to 151.9 km away. Overall, the proposed approach produced results comparable to those reported for traditional models and recent state-of-the-art methods that rely on local meteorological data, as indicated by R2, MAE, and RMSE values reported in the literature.
SN  - 2690-1692
UR  - https://doi.org/10.21926/jept.2602007
DO  - 10.21926/jept.2602007
ID  - Messias2026
ER  -