Abstract: Thermostat has been widely applied to drugs, textiles, food processing and other fields. However, because there is a relatively large time lag of the temperature measurement and coupled with a time-varying and large time constant for temperature control in thermostat, the temperature precise regulation in thermostat is a challenge problem. Regular control methods such as IMC-PID can not perform very well. Till now, most of the modern control strategies are model-based and system identification methods are widely applied to get the plant model. However, in view of the time delay and large time constant features of thermostat regulation, the stability time of the step response is too long to carry out the step response identification. Because of the limits of safe operation, it is also difficult to apply the identification methods based on the pseudo random signal to the real operation. To identify the process model rapidly and satisfy the limits of safe operation, the relay feedback identification method is applied in this work. By applying the adaptive control, large inertia and time-variant can be overcome. The initial value of the adaptive controller parameter affects the overall control effect. It is necessary to design the initial value of the adaptive control parameter. In this work, a model identification method for temperature control of the thermostat with relay feedback identification is proposed. Based on the identified model, an adaptive controller is designed and the initial parameter of adaptive controller is optimized via solving linear martix inequality (LMI) with robust control method. The proposed method is applied to develop a temperature regulator based on C2B8SNFA MCU (Micro Controller Unit) for thermostat. The real application results verify the feasibility and effectiveness of the proposed method and the developed temperature regulator for the thermostat.