Please click here to access the live Quarto Document
- Investigating the relationship between temperature and monthly energy consumption in Ontario, modeling dynamic regression model with ARIMA.
Temperature:
- Historical data shows an increase in temperature percentiles (10%, 50%, and 90%) over the past 30 years, with a strong upward trend and clear seasonality.
Energy Consumption:
- The decomposition of energy consumption data reveals a decreasing trend with strong seasonal components, indicating regular fluctuations throughout the year.
1. Temperature Forecasting:
- Utilized ARIMA(1,0,0)(2,1,2) model to forecast future temperatures based on ACF and PACF plots.
- Applied seasonal adjustments to address non-stationarity in the data.
2. Model Selection for Energy Consumption:
- ARIMA Model based solely on historical data.
- Dynamic Regression Model that includes temperature as an explanatory variable.
3. Model Evaluation:
- Compared performance using metrics such as AICc and BIC and conducted rolling-forecast-origin cross-validation to assess predictive accuracy.
- The dynamic regression model outperforms the ARIMA model by 1-2% in terms of error, highlighting temperature’s significance as a predictor for energy consumption in Ontario.
- Accurate forecasts for energy consumption can facilitate planning for maintenance and repairs of electricity-generating facilities.
- Future developments could explore optimization techniques using dynamic programming to enhance production efficiency and reduce costs based on available supply data.