Summary
The lack of reliable access to electricity in remote rural areas remains one of the main barriers to social, economic, and productive development in Latin America. This paper presents an innovative artificial intelligence-based approach for designing optimal energy microgrids capable of adapting to variable climatic conditions. Using Conditional Generative Adversarial
Read more Read lessNetworks (cGANs) combined with Deep Q-Network (DQN) reinforcement learning, we developed a model that intelligently selects between solar, wind or hybrid energy configurations, learning from real meteorological data collected in La Marina, a district of Tuluá municipality, Valle del Cauca, Colombia. The model achieves 7.1% Mean Absolute Error in energy generation prediction compared to 15.7% for AutoRegressive Integrated Moving
Average (ARIMA) and 11.2% for Long Short-Term Memory (LSTM), while providing 87.3% average system reliability. A pilot 20 kWp system validated predictions with 1.6 percentage point accuracy, demonstrating practical applicability for rural electrification in mountainous regions.
Additional informations
| Publication type | Session Materials |
|---|---|
| Reference | C6_12032_2026 |
| Publication year | |
| Publisher | CIGRE |
| Country | Colombia |
| Study committees | |
| File size | 643 KB |
| Price for non member | 30 € |
| Price for member | 30 € |
Authors
ORTIZ TORRES Luis ferney - gers; CERON Jaime alberto - Univalle; GOMEZ LUNA Eduardo - Univalle
