At present our country has paid more and more attention to the energy problems. The proportion of the energy consumption in buildings to energy consumption in total country has exceeded 25%, and the proportion of the energy consumption of HVAC (heating, ventilation and air-conditioning) to the total energy consumption in a building has been close to 50%. Improving the operation and control level in a HVAC system will reduce energy consumption in a building apparently, and then release the problem of energy absence. For the characteristics of delay time, time-variation and nonlinearity in an air-conditioning system,it makes traditional PI/PID control method can't get good control performance, and then the phenomena of over-heating and over-cooling often appear and more energy will be consumed. Some adaptive control strategies and relevant system identification methods fit for air-conditioning systems were put forward in this paper. The research object is the air-conditioning system in a test room of the testing platform of air enthalpy-difference method. The specialties of the air-conditioning system in the test room can represent the majority of air-conditioning systems. Above adaptive control strategies and relevant system identification methods were applied to the air-conditioning system in the test room. Also a remote control system of refrigeration & air-conditioning testing-platform based on Internet which adopts adaptive Fuzzy PI control algorithm was developed in this paper. The primary research and achievement are: At first the mathematical model of the air-conditioning system in a test room of the testing platform of air enthalpy-difference method was built by mechanism. Then the simulation environment of the air-conditioning system was set up by MATLAB/Simulink. The theoretical model was validated by experiment using step response.An air-conditioning system must be identified when applies self-tuning control to the air-conditioning system. For a common air-conditioning object can be considered as first-order-plus-delay-time model and the algorithm of recursive least squares can't identify the delay time, the paper put forward two identification methods: RLS with exponential forgetting combined with model of zero frequency method and gradient method. The two identification methods can estimate the parameters of the air-conditioning object including delay time, and they have the advantages of simple, fast and fit for real-time control. The identification methods were applied to the air-conditioning system in the test room by simulation research. A self-tuning PI control strategy with robustness was put forward in the paper; the self-tuning control strategy adopts the RLS with exponential forgetting combined with model of zero frequency method which is used to estimate the parameters of an air-conditioning model under close-loop. In addition for the time constant of an air-conditioning model is always large, the paper designed a PI tuning formula with robustness which can guarantee the real air-conditioning process with gain-margin Am=2.98 and phase-marginφm=59.8°.The self-tuning PI control strategy with robustness, a self-tuning PI control strategy based on H∞and a self-tuning PI control strategy based on the Z-N tuning formulas were applied to the air-conditioning system in the test room and their control performances were compared among each other. For the delay time of an air-conditioning model will reduce the control performance, self-tuning control strategy with Smith predictor was put forward in this paper. The self-tuning control method also uses above combination algorithm to estimate the parameters of an air-conditioning model including delay time, and a Smith predictor is adopted to realize accurate compensator for the an air-conditioning model's delay time. The control performance of the self-tuning control with predictor was compared to a well tuned PID controller; the simulation results show the self-tuning control algorithm has better control performance