CNN-LSTM architecture for predictive indoor temperature modeling

Indoor temperature modeling is a crucial part towards efficient Heating, Ventilation and Air Conditioning (HVAC) systems. Data-driven black-box approaches have been an attractive way to develop such models due their unique feature of not requiring detailed knowledge about the target zone. However, noisy non-linear nature problem remains bottleneck especially for long prediction horizons. In this paper, we introduce Convolutional Neural Networks-Long Short Term Memory (CNN-LSTM) architecture combine exceptional extraction convolutional layers with Long (LSTM)’s capability learning sequential dependencies. We experimentally collected dataset compared three approaches: Multi-Layer Perceptron (MLP), LSTM CNN-LSTM. Models are evaluated 1-30-60-120 min horizons closed-loop scheme. The CNN-LSTM outperformed all other showed better robustness against error accumulation. It managed predict room R2>0.9 in 120-min horizon.