The hypothermia phenomenons on body surface and hypothermia therapy control platform are introduced. MSP430 microprocessor is taken as a key control part, upper computer and thin film transistor (TFT) LCD touch screen as operation display interface and variable parameter prediction proportion integral derivative (PID) based on ant colony algorithm as key algorithm. Neural network is taken as prediction models and the control parameter of PID is optimized on line using the global optimization ability of ant colony algorithm to monitor the temperature of human body in real time. Experimental results show that unit step response overshoot and adjustment time are drastically reduced, dynamics and stability has been significantly improved and the surface temperature of human body can be stabilized accurately and rapidly in the range of requirements processed by the algorithm.