Magnetic nanofluid hyperthermia (MNH) damages malignant cells by the heat generated by magnetic nanoparticles (MNPs) exposed to an alternating magnetic field during therapy. The key point for magnetic hyperthermia is to maintain the treatment temperature to a safe range for bio-tissue, in which malignant tissue will be damaged due to its higher heat sensitivity but not the healthy tissue. However, the therapeutic system for MNH should be a nonlinear one since the treatment temperature is generally determined by many certain and uncertain factors, which result in the difficulty to modulate the temperature to a specific range in a practical application. This study proposes a control method for an MNH system by introducing a proportional–integral–derivative (PID) control algorithm and dynamically optimizes the PID coefficients for the proposed system by considering simulated annealing (SA) algorithm during therapy. The treatment temperature distribution for bio-tissue is obtained by solving an improved Pennes bio-heat transfer equation using finite element method (FEM). The simulation results demonstrate that the proposed control system can effectively modulate the power dissipation for MNPs and further exactly regulate the transient treatment temperature to an expected value. In addition, this proposed system can also automatically adapt to different cases during therapy, in which the nanofluid concentration distribution inside tumor region changes with nanofluid injection strategy.