Two controllers (single channel optimal controller (SCOC) and multivariable linear quadratic Gaussian controller (LQG)) are built to improve the performance of a transcritical CO2 refrigeration cycle. The SCOC is first proposed to search the optimal gas cooler pressure online by adjusting the nozzle throat area. Considering the uncontrollable cooling capacity by the SCOC

the second controller LQG is designed to achieve the controllable cooling capacity. Experimental results show that the optimal gas cooler pressure is actually achieved by the SCOC with a maximal COPh of about 3.15 under the given operating conditions. However

with the SCOC

the cooling capacity is changed passively and uncontrollably. Moreover

experiments with LQG indicate that the cooling capacity and gas cooler pressure are accurately tracked independently. However

LQG cannot ensure the maximal operating performance when the system becomes steady. This work points out that the two controllers have their own advantages and disadvantages. More work needs to be done to combine the two algorithms into one. This will achieve maximal efficiency under the desired cooling capacity.