Disturbance observer (DOB)-based robust control for a variable speed refrigeration system (VSRS) is described in this paper for accurate tracking of the reference temperature under model uncertainty and disturbances including external noise. In particular, a Q-filter, which is the core of a DOB, is designed by H∞ (H-infinity) norm optimization to estimate the disturbance and to compensate the model uncertainty of VSRS. Simultaneous control of the oil outlet temperature and superheat is carried out using a dual single-input single-output proportional–integral (PI) control scheme with the designed observer in the inner loop. The effectiveness of the observer is evaluated by means of computer simulations and real experiments. The results showed that settling time and steady-state error were significantly reduced when DOB was applied, compared to when DOB was not applied even in the presence of model uncertainties and disturbances. These control results show that PI control with the designed DOB can facilitate robust control even when the VSRS model contains uncertainties and unforeseen disturbances are added to the system.