2022 · Time Series
Statistical Analysis and Forecasting of Solar Energy Production
Time Series Forecasting ARIMA
Statistical Analysis and Forecasting of Solar Energy Production
BITS Pilani | Sep 2022 – Dec 2022
Python ARIMA Prophet Time Series Renewable Energy
Project Overview
Developed comprehensive time series forecasting system for solar energy production to support grid planning and energy management decisions.
Key Contributions
Model Development: Developed ARIMA and Prophet time series models for daily solar energy production forecasting using 3 years of historical generation data
Feature Engineering: Incorporated exogenous variables (temperature, cloud cover, humidity) improving forecast accuracy by 12% over univariate baseline
High Accuracy: Achieved MAPE of 8.5% on 30-day ahead forecasts, enabling improved grid scheduling and energy trading decisions
Technologies Used
- Languages: Python
- Libraries: statsmodels, Prophet, pandas
- Methods: ARIMA, SARIMA, Facebook Prophet
- Data: 3 years historical solar generation
Key Results
| Metric | Value |
|---|---|
| MAPE (30-day forecast) | 8.5% |
| Historical Data | 3 years |
| Improvement over Baseline | +12% |
| Exogenous Variables | 3 |
Forecasting Pipeline
Data Collection → Preprocessing → Stationarity Testing
→ Model Selection (AIC/BIC)
→ Parameter Optimization
→ Forecast Generation
→ Error Analysis → Deployment