Page 59 - IJAMD-2-1

International Journal of AI for
Materials and Design
ML-based MPC for multizone BAC

Practical implications for Equation II:

The weighting scheme reflects a design philosophy where occupant comfort is prioritized, but energy efficiency remains a
secondary goal. The slack variable ϵ ensures feasibility under unexpected disturbances (e.g., occupancy spikes), though its
high penalty weight (W ) minimizes its use in practice.
ϵ
Equations III and IV
The optimization is bounded by operational and comfort limits:
Q cool,lb ≤ Q ≤ Q cool,ub (III)
cool
−0.5 ≤ PMV ≤ 0.5 (IV)

Contextual details for Equations III and IV:
• Cooling power bounds (Q cool, lb , Q cool, ub ) are determined by the HVAC system’s capacity
• The PMV range (±0.5) corresponds to the ASHRAE Standard 55 “neutral” comfort zone, ensuring 80 – 90% occupant
satisfaction.
Additional notes:
• Control interval: The MPC operates at 5-min intervals, balancing responsiveness with computational tractability
• Slack V=variable (ϵ): Introduced as a soft constraint to avoid infeasible solutions during transient disturbances
(e.g., abrupt weather changes)
• Thermal comfort: The PMV model assumes steady-state conditions and standardized clothing/metabolic rates; dynamic
adjustments for occupant activity are not modeled here.

Volume 2 Issue 1 (2025) 53 doi: 10.36922/ijamd.8161