Multi-zone temperature prediction in a commercial building using artificial neural network model

Predicting temperature in buildings equiped with Heating, ventilation and air-conditioning (HVAC) systems is a crucial step to take when implementing a model predictive control (MPC). This prediction is also challenging because the buildings themselves are nonlinear, have many uncertainties and strongly coupled. Artificial neural networks (ANNs) have been used in previous studies to solve such a modeling problem. Unlike most of the studies that have only considered small-scale, single zone modeling task, this paper presents a novel ANN modeling method for the modeling inside a real world multi-zone building. By comparing ANN models with different input variables, it was found that the prediction accuracies can be greatly improved when the thermal interactions were considered. The proposed models were used to perform both single-zone and multi-zone temperature prediction and achieved very good accuracies.