가변속 냉동시스템의 강인제어를 위한 H∞ 놈 최적화 기반의 외란관측기 설계

Abstract

The application of a variable speed refrigeration system(VSRS) is expanding to the various industrial fields because the system copes with partial loads and has energy saving performance. VSRS is composed of a variable speed compressor, an electronic expansion valve(EEV), and heat exchangers, with strong inherent nonlinear characteristics. Even a sophisticated mathematical model used for VSRS, namely the high order linear state space model, has many drawbacks when designing a control system because of model uncertainity and disturbances including noise. Moreover, accurate identification of the dynamic characteristics of VSRS through modeling is very laborious and tedious. Therefore, a control method with strong robustness against model uncertainty and disturbance is inevitably required for VSRS. So far, many approaches such as PID controller logic, LQR(Linear Quadratic Regulator), LQG(Linear Quadratic Gaussian), and fuzzy logic control have been applied to VSRS to obtain desirable control performance. Fuzzy logic control has excellent robustness against model uncertainty and disturbance because it can be applied without the mathematical dynamic model of VSRS. However, precise control in economical manner is difficult to achieve with this method in addition to analysis of the dynamic behaviors of the system. LQR and LQG are suitable for multiple-input and multiple-output(MIMO) control. These two methods can provide the optimum control performance between input energy and control accuracy. However, they are insufficient for control robustness due to modeling errors in the linearized state space model of VSRS. With VSRS, it is not easy to construct an analytical linearized state space model because of high-order terms and difficulties with parameter identification. In contrast, PID logic, which is based on a transfer function model, is widely used in industrial applications because it is very simple to design, implement, and use, in addition to its effectiveness on the premise that the model accurate. The transfer function model for PID control can be easily obtained from simple dynamic experiments. Unfortunately, the model will inevitably include uncertainties due to linear approximation and characteristic parameter identification. Furthermore, in the actual operating environment for VSRS such as ambient temperature, thermal load will be different from modeling conditions and unexpected noises may be added to the system. Therefore, it is difficult for the conventional PID control to achieve robust control performance on VSRS. Hence, in this paper, a disturbance observer(DOB) based on Q-filter is designed and incorporated into a conventional PI controller to provide strong robustness against model uncertainty and disturbance. First, a linear first-order transfer function model of VSRS is obtained through dynamic characteristic experiments near the operating point and its validity is confirmed over the entire operating range of VSRS. Next, factors affecting model uncertainty are fully investigated through several experiments. To be more specific, the biggest influential factors on the model uncertainty and the variation range of the characteristic parameters of the model are clarified. Finally, Q-filter, which is the core of DOB, is formalized through (H-infinity) norm optimization. The effectiveness of the designed DOB-based PI control is evaluated by means of Matlab-based computer simulations and through real experiments comparing the control performance of PI with and without DOB. The control results of simulations and experiments show that PI combined with the designed DOB significantly improves robust control performance even when the VSRS model contains model uncertainties and unforeseen disturbances are added to the system.