Abstract: AC induction motors have been widely used in various industries, and their sensorless control system has the advantages of simple structure and low cost. To realize sensorless control of the induction motor, the rotor flux and speed must be observed. The sliding mode observer has strong robustness and high observation accuracy; therefore, it has attracted considerable attention. In addition, in the induction motor control system, the high switching frequency of the inverter causes high loss to the switching device and reduces the service life of the inverter. Therefore, the inverter switching frequency should be reduced to the maximum achievable level during the control process. However, in the traditional low switching frequency control method based on space vector pulse-width-modulation (SVPWM), a decrease in the switching frequency reduces the sampling frequency of the system. Simultaneously, it increases the sampling time delay, reduces the control bandwidth, and directly affects the observation accuracy of the rotor flux observer in the sensorless control method, thereby resulting in a poor sensorless control effect. Predictive control has the characteristics of rolling optimization, which realizes a low switching frequency and ensures a high sampling frequency of the system, thus ensuring the observation accuracy of the observer and system control effect in the sensorless application scenario. Therefore, this paper proposes a sensorless control method for an induction motor with low switching frequency to solve the problem that the traditional control method adversely affects noninductive flux observation. The sensorless predictive control system of the induction motor is constructed by combining a speed adaptive sliding mode observer under the boundary confined predictive control method. This sensorless control system can effectively reduce the switching frequency at a high sampling frequency and ensure a high dynamic characteristic of the system. In addition, the introduction of the sliding mode makes the observer robust. Experimental results reveal that the proposed low switching frequency sensorless control strategy exhibits high control accuracy and strong anti-interference performance. The adaptive sliding mode observer has good observation accuracy in the steady state and dynamic experiments, and the current loop response time can reach 1.52 ms. Moreover, it can achieve stable observation and operation when the speed is higher than 75 r·min^–1, and the highest average switching frequency of the system is maintained at about 500 Hz, which is a low level.