Development of a Machine Learning-Based Indoor Temperature Prediction Model and Optimal Control Algorithm for AHU Discharge Air Temperature in CAV System

The CAV (Constant Air Volume) system is widely used for cooling and heating buildings by maintaining a constant airflow while varying the discharge temperature.However, this system has limitations in responding promptly to various external environmental conditions and indoor thermal load changes.This study aims to control the discharge temperature of a real-time CAV system based on indoor temperature prediction using an Artificial Neural Network (ANN).A coupled simulation using EnergyPlus and Python was conducted, analyzing data from August 1st to 31st during the summer season.The developed ANN showed a prediction error with a cv(RMSE) of approximately 7.6% and an R of 0.73.By controlling the AHU discharge temperature in real-time, indoor comfort was maintained with a PMV range of -0.2 to +0.7, and a PPD of less than 10% for most of the time.The energy consumption showed a variation of approximately 2 MWh compared to cases with fixed discharge temperatures.