Development of Temperature Control in Climatic Test Chambers with LSTM-based Deep Neural Network Algorithm

Abstract

Environmentally conditioned climatic test chambers are test devices that can simulate temperature and relative humidity conditions in a wide range as standard and can be produced in various volumes. PID is the most used control method in climatic chambers. Since the parameters in the PID controller are determined for wide ranges, the system performance is insufficient since the parameters at intermediate values and different volumes need to be optimized, long waiting times and high energy consumption to reach the target temperature value may cause the PID controller to not fully meet the requirements under some conditions. This study presents an innovative method for the control of interior space heating system with LSTM (Long-Short Time Memory)-based deep neural network model in accordance with the created test recipe. The model is trained with the dataset created with the outputs of the PID controller. As a result of the method used, an error value of 0.0014 was obtained. The presented results show that the trained model successfully predicts the outputs of the heating system according to the target temperatures.