Optimization of Control Parameters of Vehicle Air-Conditioning System for Maximum Efficiency 2020-01-1242
Modern automotive heating, ventilation, and air-conditioning (HVAC) systems have multiple and often redundant actuators. In order to maximize HVAC efficiency, it is necessary to design a control system that optimally synthesizes multiple control actions while satisfying control set points and system hardware-related constraints. To this end, an optimization approach to control system design is proposed in this paper and demonstrated for a generic HVAC system. The paper first outlines a nonlinear 12th-order HVAC dynamics model that is based on the moving-boundary method. Then, the HVAC control system is defined, which combines proportional-integral (PI) feedback controllers commanding the compressor speed and expansion valve setpoints, and open-loop actions of condenser and blower fans. Next, a three-stage optimization approach is proposed consisting of the following steps: (i) rough optimization of all the aforementioned control inputs in an open-loop manner, (ii) feedback controllers’ parameter optimization, and (iii) refined optimization of the two open-loop control inputs while having the closed-loop controllers running. The first-stage optimization is aimed at finding a rough estimate of control vector containing all control inputs, which maximizes the system efficiency (i.e. HVAC coefficient of performance, COP) while satisfying relaxed control set point and safety-related constraints (related to regulation of superheat temperature). In the second stage, the closed-loop controller parameters are optimized for the operating point obtained in the first stage and using a discrete-time model of closed-loop system, with the aim to provide a favorable trade-off between the control error suppression and control effort/energy. The third stage includes refined optimization of control system open-loop inputs to maximize efficiency, where firmer thermal comfort and safety-related constraints are provided by closed-loop control system. The optimizations are conducted by using state-of-the-art multi-objective genetic algorithm MOGA-II incorporated within modeFrontier environment. Finally, the optimization procedure is carried out for multiple control set points, in order to provide control parameter maps to be used in the form of a gain-scheduling algorithm. Analysis of optimization computational efficiency shows that a substantially improved computational speed is achieved when using the proposed, comprehensive three-stage optimization approach when compared to using only the first stage one.