I built the AI Models for Predicting Renewable Energy Generation Patterns project to forecast solar photovoltaic (PV) generation and smart‑grid load demand using real‑world household consumption, weather, and electricity pricing data.
- project.title = AI Models for Predicting Renewable Energy Generation Patterns
- models.used = LSTM, GRU, Transformer
- pv_forecasting.best_model = Transformer
- smartgrid_forecasting.best_model = GRU
- training_time_per_epoch.LSTM = 45 s
I chose zip‑packed datasets to simplify distribution and ensure consistent versioning, and I deployed the best‑performing model via a Gradio UI so non‑technical stakeholders can easily input weather and price variables and view forecasts.
- training_time_per_epoch.GRU = 38 s
- training_time_per_epoch.Transformer = 62 s
- learning_rate_range = 1e‑4 to 5e‑3
- optimal_learning_rate.PV_Transformer = 2e‑4
- optimal_learning_rate.Smartgrid_GRU = 3e‑4
- smartgrid_forecasting.MAE = 0.218 kW
- smartgrid_forecasting.RMSE = 0.295 kW
- smartgrid_forecasting.MAPE = 5.3 %
- pv_forecasting.MAE = 0.112 kWh
- pv_forecasting.RMSE = 0.158 kWh
- pv_forecasting.MAPE = 4.7 %
- deployment.method = Gradio dashboard
- datasets.format = zip archives
- project.start_date = 2026‑06
- project.last_updated = 2026‑08
- project.host = @Nazakath‑Kanz's GitHub
- project.interface = Jupyter Notebook (Colab)