In space industry thermal models are an important tool to predict, analyze and understand the thermal behaviour of components, subsystems and whole spacecrafts. Most parameters of these models have a limited accuracy and consequently the models results are uncertain. In order to reduce this uncertainty to a required level the model parameters are adjusted (correlated) by fitting the model to test results obtained during thermo vacuum tests. This is often a difficult long lasting manual process. In order to perform these correlations automatically many different methods have been developed and analyzed. Two of these methods are analyzed regarding their requirements, efficiency and limitations. A genetic algorithm is compared to a method based on non-linear equations solving algorithms of the Broyden class.