유전 알고리즘과 외란 관측기를 이용한 최적 서보 제어기 기반의 가변속 냉동시스템의 강인한 온도 제어

Abstract

Variable speed refrigeration systems (VSRSs) have been applied in different industrial and commercial fields because of their excellent energy saving performance and partial load response. A VSRS basically comprises a variable speed compressor, an electronic expansion valve (EEV), and heat exchangers. As all components of the system are connected to various pipes and valves, the system exhibits strong inherent nonlinear characteristics such as dead time in the operational ranges. Moreover, heat load fluctuates continuously. Thus, obtaining a practical linear nominal model and accurately controlling the target temperature is difficult. Even the sophisticated mathematical model used for VSRS (i.e., the high-order linear state space model) has many deficiencies when the controller is operating because of model uncertainties and disturbances including noise. For high dimensional models, accurately identifying the dynamic characteristics of VSRSs is a very laborious and tedious task. Therefore, a servo control method that resists model uncertainties and disturbances based on a practical and simple as possible model is required. VSRS control systems generally belong to the class of a multi-input/multi-output (MIMO) system because they have two inputs and outputs to control the target temperature and superheat based on the rotation speed of the compressor and EEV opening angle. In terms of flexible access to multivariate MIMO systems and optimal control performance, applying the optimal control method based on the state space model is convenient. Therefore, in this paper, a more practical, simple, and high performance optimal robust servo controller design is proposed; the proposed approach solves the problems of the conventional optimal servo controller design for VSRSs. First, the state space model used for designing the controller is derived from the transfer function, which is easily obtained through dynamic characteristic experiments near the operating point of the control object; fewer parameters that affect the model uncertainty must be identified, which facilitates modeling. Second, the controller is designed as a servo-type controller to follow strictly the target value on the step. In particular, the weighting matrix for the evaluation function, which is a crucial parameter for the design of an optimal controller, can be easily selected with a genetic algorithm without trial and error procedures. Finally, a disturbance observer (DOB) with a simple Q-filter compensates for disturbances, which makes the system resistant to disturbances and model uncertainties. The proposed controller was verified through MATLAB-based simulations and experiments with a VSRS-based oil cooler system as the control object. Moreover, the robustness against disturbances and model uncertainties was assessed by comparing the results of the cases with and without DOB in detail.