| Introduction | Correlation Analysis | Forecasting | Conclusion |
Concluding Remarks
This study started was initiated based on the hypothesis that weather can be used to understand electricity demand based on various consumer proxies. The overall results indicated that the weather has varying correlation across different electricity consumers. Industrial electricity consumption has low correlation to weather parameters which are negative in magnitude for temperature and pressure variables. The electricity consumption in a commercial and apartment block are uniquely negatively correlated with humidity which has a positive correlations with consumption in an industrial setting. Rolling correlations indicated that the relationship between energy consumptions and weather parameters was not stable and changed across time.
Weather variables correlated among themselves indicate a strong correlation between normal outdoor temperature(T) and the dew point temperature(Td). For this reason one of these variables must be eliminated from the subsequent regression models to control for potential multicollinearity.
Regression models indicate that Pressure (P), Humidity (U), Wind speed (Ff) and Dew point temperature (Td) are all significant variables in explaining the consumption of electricity. These tested models were however observed to have low predictive power and therefore could not generate reliable forecasts on electricity consumption but they do provide information on how select weather parameters affect electricity consumption.
To improve the predictive power to these models, additional variables would be required to added to the equations.