To investigate and optimize a refrigeration system, the behavior at various operating conditions must be known or determined. The performance and improvement possibilities may then be inferred from measurement data and compared with corresponding performance key figures. These values are typically referred to normal conditions and it is usually unknown which ones represent an adequate operation. However, it is relevant for refrigeration plant operators to have reference values for a large range of operation conditions as a baseline for determining the obtainable improvements. The present work proposes the application of steady-state models of refrigeration machines for increasing the range of applicability of the exergy-based optimization potential index method. Four different modeling approaches are evaluated and discussed: equation-fit, physical lumped parameter, refrigeration cycle and artificial neural network based models. The practical usage of the improved evaluation method is shown for the subsystem refrigeration machine on a real field installation as a case study. With the introduced additional limits for the optimization potential index, the interpretability of the results is increased. The distinction between adequate (technical requirements exceeded), acceptable (technical requirements fulfilled) and inadequate (potential for improvement) operation according to the state of the art in technology is straightforward, which is important in practice.