A Comparative Study of Machine Learning Models in Predicting Energy Consumption

Climate change and the integration of renewable energy resources into the electricity grid in recent years are making electricity consumption very vulnerable. Traditional time series forecasting methodologies make predictions based on historical trends and patterns and are inadequate to address the changing environmental parameters in real-time. Machine learning methods, on the other hand, are more flexible in taking into consideration of parameters such as temperature, wind power, and market price, which are likely to fluctuate in real-time and influence electricity consumption. In this research, we propose to develop an accurate prediction model for hourly electricity consumption using machine learning. We will use California consumption data from US Energy Information Administration and hourly local climatological data (LCD) from National Oceanic and Atmospheric Administration for the study. Six machine learning models that include Linear Regression, K-Nearest Neighbors (KNN) Regression, Gradient Boosting, Random Forest Regression, Long Short-Term Memory (LSTM) – a Recurrent Neural Network, and Support Vector Machine (SVM) will be used, and their prediction accuracy will be compared and analyzed.