Data driven models applied in building load forecasting for residential and commercial buildings
A significant portion of the operating costs of utilities comes from energy production. Machine learning methods are widely used for short-term load forecasts for commercial buildings and also the utility grid. These forecasts are used to minimize unit power production costs for the energy managers for better planning of power units and load management. In this work, three different state-of-art machine learning methods i.e. Artificial Neural Network, Support Vector Regression and Gaussian Process Regression are applied in hour ahead and 24 –hour ahead building energy forecasting. The work uses four residential buildings and one commercial building located in Downtown, San Antonio as test-bed using energy consumption data from those buildings monitored in real-time. Uncertainty quantification analysis is conducted to understand the confidence in each forecast using Bayesian Network. Using a combination of weather variables and historical load, forecasting is done in a supervised way based on a moving window training algorithm. A range of comparisons between different forecasting models in terms of relative accuracy are then presented.