Abstract: Aiming at the problems of energy waste and low transportation efficiency caused by continuous high-speed operation of scraper conveyor in fully mechanized mining face, this paper combined with two-way coal cutting technology, systematically analyzed the operation stage of scraper conveyor, established the energy consumption model of scraper conveyor, and proposed a speed control method based on load torque prediction and energy consumption optimization. Firstly, a coal quantity model is established to describe the dynamic characteristics of coal quantity changing with operating conditions. Then, according to the running resistance characteristics of the scraper conveyor, the relationship between coal volume, driving force and running resistance is defined, and the energy consumption model of the scraper conveyor is built. In order to cope with the complex and variable operating conditions of fully mechanized mining face, a rough radial basis neural network (RRBFNN) is introduced in this paper to accurately predict the load torque of the scraper conveyor and generate the key input variables required for the optimization model. On this basis, the improved particle swarm optimization algorithm (PSO) is used to optimize the running speed of the scraper conveyor with the goal of minimizing energy consumption. The improved algorithm balances the global search and local search capabilities while introducing dynamic inertia factor, thus improving the optimization accuracy and convergence efficiency. Finally, the proposed method is validated with the actual data of Yujialiang 43101 fully mechanized mining face. The results show that the speed control method can effectively reduce the energy consumption of the scraper conveyor by 10.42% in one production cycle.