A Predictive Control Model Based on Environmental Parameter Modeling (Temperature and Humidity) for HVAC Systems to Enhance Thermal Comfort and Reduce Energy Consumption

Abstract

This study proposes a predictive control model for HVAC systems based on the modeling of environmental parameters, such as temperature and humidity, to improve thermal comfort and reduce energy consumption. This research is a modeling-based study that employs MATLAB software along with a metaheuristic algorithm. To evaluate and compare the methods, two metrics—RMSE and R²—were used. Both the MLP (Multi-Layer Perceptron) method and the Genetic Algorithm were applied to assess their accuracy as effective approaches for time-series prediction.

The results indicate that, in terms of R², the Genetic Algorithm’s performance in predicting room temperature and humidity affected by the HVAC system is comparable to that of the MLP method. Overall, the MLP network demonstrated the highest prediction accuracy. The Genetic Algorithm, however, showed lower accuracy for humidity prediction. To investigate this further, the accuracy of the MLP model was evaluated without disrupting the data sequence. In this case, the R² value was 0.559, and the RMSE was 0.12. The notable reduction in R² compared to when the data sequence was preserved in the MLP network, along with the low R² of the Genetic Algorithm, highlights its limitations for this application.