Cooling load estimation using machine learning techniques

Estimating cooling loads in heating, ventilation, and air-conditioning (HVAC) systems is a complex task. This is mainly due to its dependence on numerous factors which are both intrinsic and extrinsic to buildings. These include climate, forecasts, building material, fenestration etc. In addition, these factors are non-linear and time-varying. Therefore, capturing the effect of these parameters on the cooling load is a complex task. This investigation combines forward modelling, i.e., physics based model simulated using energyPlus with deep-learning techniques to build a cooling load estimator. The forward model captures all the time-varying factors influencing the cooling loads. We use the long short-term memory (LSTM), a deep-learning method to provide forecasts of cooling loads. The advantage of the proposed approach is that cooling load estimations can be provided in real-time thus providing sort of soft-sensor for estimating cooling loads in buildings. The proposed approach is illustrated on a building of suitable scale and our results demonstrates the ability of the tool to provide forecasts.