Magnetic Resonance Imaging (MRI) is an attractive method of temperature monitoring in vivo due to its non-invasive nature. The natural extension of this temperature monitoring is to implement temperature control. This work outlines a method of MRI-based thermal modelling for multi-source phased array heating systems that can potentially be employed, in the future, for real time temperature prediction and control. This method is based on Pennes bioheat equation. It employs the superposition of an empirically acquired basis set of temperature distributions that define the heating system's temperature response. MR thermal images based on the proton resonance frequency shift (PRFS) technique are used to acquire this basis set. The feasibility of this approach is tested in phantom using a radiofrequency (RF) heating system. The results show that this method can accurately reproduce measured temperature distributions outside of the basis set.