Short-term electrical load prediction based on a system identification model in an actual educational building

According to the U.S. Energy Information Administration (EIA), the U.S. buildings sector consumed nearly ۷۴% of total electricity sales in ۲۰۲۱. Predicting the electricity load has become increasingly important in building energy management, directly affecting control and economic operation. Many methods for predicting energy consumption in the building sector exist, but they have limitations. This study used five modeling techniques, system identification, support vector regression machine, artificial neural network, autoregressive with exogenous, and multiple linear regressive, to forecast hourly electricity consumption in an actual educational building in Tehran, Iran. The first educational building was modeled using EnergyPlus and used bills to validate electricity consumption. In the second part, we obtain input/output data to perform system identification and four other techniques. Based on the results of this study, the system identification technique is more efficient and reliable than other techniques for predicting electricity loads. The system identification technique can achieve ۹۶% Coefficient of Determination (R^۲) and less than ۱۰% Normalized Root Mean Square Error(NRMSE) in electricity load forecasting for an actual educational building

کلیدواژه ها:

system identification model ، building electricity consumption ، SVM ، ARX ، MLR ، ANN ، forecasting ، machine learning