Abstract: To improve the temperature prediction accuracy of liver tumor radiofrequency ablation (RFA), a temperature distribution model with a proportional integral (PI) feedback controller was established. Furthermore, the influences of different heat transfer equations (Pennes and Hyperbolic equations) on temperature distribution in simulation models were studied. First, a single-electrode RFA ablator was used to perform RFA experiments on ex vivo pig livers. The steady-state temperature of the electrode tip was set to 90℃, the ablation time was 600 s, and the ablation data of the temperature measurement points was obtained by using thermometers. Second, a finite element model was constructed in combination with a PI controller. The PI controller dynamically adjusted the voltage source to maintain the tip temperature consistent with the experimental data. Hyperbolic equation based on the non-Fourier heat transfer law and Pennes equation based on the Fourier heat transfer law were investigated. Temperature-dependent functions for the electrical conductivity and thermal conductivity of the biological tissue were employed. Finally, the simulation data was verified by comparison with the measured temperatures. Results show that compared with Hyperbolic equation, the simulation model using Pennes equation achieves higher prediction accuracy. The mean values of the standard deviation, the average error and the maximum error between the simulations and the measurements are 0.78, 1.46 and 2.90℃, respectively. Therefore, the finite element model based on PI controller and Pennes bio-heat transfer equation can effectively predict RFA temperatures.