Thermal load prediction and operation optimization of office building with a zone-level artificial neural network and rule-based control

Abstract

Precise and quick thermal load prediction for buildings is imperative in realizing the flexibility of building energy systems. Operation optimization based on the prediction results can effectively mitigate the energy consumption and operation cost. Here, an artificial neural network (ANN) model is built to predict the load demand and energy consumption of an office building. Thermal zones division is considered to improve the prediction accuracy. Measured data are collected to establish and validate the ANN model. The total number of samples for training, validation and testing are 13674, 3978 and 3977, respectively. A rule-based control optimization model is introduced and combined with ANN model to optimize the operation of heating, ventilation and air-conditioning system. Building energy flexibility is activated by optimization model in a time-of-use electricity price scenario. The results indicate that ANN model has high precision and the coefficient of variation of root mean squared errors corresponding to the load demand prediction and energy consumption prediction are 10.76% and 15.59%, respectively. Furthermore, the optimization results can reduce the operation cost by 39.22% and 44.41% in the heating and cooling season, respectively.