Deep-learning-based prediction model for indoor temperature parameters and energy consumption of radiant cooling system

It is estimated that air conditioning systems account for approximately 40% of a building's operational energy consumption. Therefore, a rational energy scheduling strategy must be based on an accurate prediction of the future state of the building. The prediction model used neural network technology, a 1DCNN-LSTM-GRU neural network model is built, which couples a One-dimensional convolution layer with a Long Short-Term Memory layer and a Gated Recurrent Unit layer. The validation on the test set leads to the conclusion that 1DCNN-LSTM-GRU can do a good job of predicting the indoor temperature and energy consumption parameters, and it is also found that the accuracy of the model prediction decreases gradually as the prediction interval expands.