Forward-Looking Setpoint Prediction for HVAC Systems via Look-Ahead Predictive Control

Abstract

To address the challenges of comfort and energy efficiency in modern buildings posed by pure feedback PID control, Model Predictive Control (MPC) has emerged as a solution. With the advancement of technology, mitigating the heavy computation and deployment burden of conventional MPC in real time is crucial. This paper introduces a Look-ahead Predictive Control scheme that shifts the temperature setpoint ahead by one hour, allowing heating, ventilation and air-conditioning systems to react early without invoking a full MPC solver. Under identical thermodynamic conditions, the new strategy is benchmarked against standard Proportional–Integral–Derivative (PID) control on a MATLAB model run for eight consecutive hours. Key indicators—energy use, integral absolute error, tight-band Integral Absolute Error (IAE) within 0.2°C, maximum overshoot, and both full-process and steady-state windows—are extracted and contrasted. Look-ahead control cuts peak overshoot by roughly 51% and lowers steady-state IAE by 1.3%, while full-process IAE, energy consumption and Tight-IAE rise by 8.3%, 2.5% and 8.5%, respectively. Stabilization time, defined as the first continuous 10-minute interval within 0.5°C, shows no reduction. Overall, the setpoint-based predictive approach offers a practical trade-off: it markedly improves thermal comfort at only a slight energy penalty, without resorting to a complete MPC implementation.