Abstract: As an important demand response resource, the air conditioning load can adjust the active load in real time and participate in the frequency regulation of power systems. The traditional temperature set-point control method assumes that the temperature set-point of the air conditioner is continuously adjustable, which ignores the minimum stepsize of temperature adjustment（MSoTA）of the air conditioner. Since the MSoTA of most air conditioners is 1 ℃, a control strategy of the air conditioning load based on temperature set-point adjustment probability（TSPAP）is proposed, which makes the aggregate power of the air conditioning load cluster continuously adjustable, so as to better participate in the frequency regulation of power systems.The proposed method uses the framework of two-layer broadcasting control to divide the air conditioning loads into six independent states, and the power regulation demand is modulated into TSPAP according to different states. TSPAP is sent to each decentralized controller as a global control signal, and is demodulated to the temperature set-point adjustment signal of ±1 ℃,which is sent to each air conditioner. The simulation results show that the proposed method can continuously adjust the aggregate power of a large number of air conditioning loads under the condition that the MSoTA of a single air conditioner is 1 ℃, so as to obtain better results in the primary and secondary frequency regulation.