Abstract: This paper takes the heavy-duty scraper conveyor with asynchronous variable frequency motor as the research object. Aiming at the problems of multi-motor drive power imbalance and system response lag of heavy-duty scraper conveyors under complex working conditions, an adaptive control method for multi-motor power balance of scraper conveyors based on the combination of long short-term memory network and proportion-al-integral control (LSTM-PI) is proposed. This method establishes a power balance control system model by analyzing the transmission characteristics of the scraper conveyor and the power coupling mechanism of mul-tiple motors. Then, the temporal learning ability of the LSTM neural network is integrated with the real-time feedback characteristics of PI control to construct the LSTM-PI adaptive control model, achieving dynamic prediction and parameter adaptive adjustment for the power balance control of multiple motors of the scraper conveyor. Finally, a MATLAB/Simulink simulation model was established, and the control algorithm proposed in this paper was experimentally verified under three working conditions: start-up, sudden load change, and uneven load. The results show that the LSTM-PI control method can achieve smooth acceleration in the start-up stage, the chain speed error is controlled within 1.2%, and the system impact load is reduced. Under the con-dition of sudden load changes, the peak power deviation of the motor decreases by 48%, and the recovery time is shortened by 32%. Under uneven load conditions, the power deviation of the three motors remains stable within 1%, enabling real-time power adaptive distribution among multiple motors. In addition, the LSTM-PI model also demonstrates a significant advantage in response speed, greatly enhancing the operational stability of the scraper conveyor under complex working conditions.