Development of a control algorithm aiming at cost-effective operation of a VRF heating system

Abstract

This study aims to develop a control algorithm that can operate an intermittently working variable refrigerant flow (VRF) heating system in a cost-effective manner. An artificial neural network (ANN) model, which is designed to predict the heating energy cost during the next control cycle, is embedded in the control algorithm. By comparing the predicted energy costs for the different setpoint combinations for the system parameters, such as the air handling unit (AHU) supply air temperature, condensing warm fluid temperature, condensing warm fluid amount, and refrigerant condensing temperature, the control algorithm can determine the most cost-effective setpoints to optimally operate the heating system. Two major processes are conducted development of the predictive control algorithm in which the ANN model is embedded, and performance tests in terms of prediction accuracy and cost efficiency using computer simulation. Results analysis reveals that the ANN model accurately predicts the energy cost, presenting a low coefficient of variation of the root mean square error value (7.42%) between the simulated and predicted results. In addition, the predictive control algorithm significantly saves on the heating energy cost by as much as 7.93% compared with the conventional heuristic con