Research on the Multiobjective and Efficient Ore-Blending Scheduling of Open-Pit Mines Based on Multiagent Deep Reinforcement Learning

In order to solve the problems of a slow solving speed and easily falling into local optimization an ore-blending process model (of polymetallic multiobjective open-pit mines), efficient scheduling method based on multiagent deep reinforcement learning is proposed. Firstly, according actual production situation mine, optimal control for ore blending was established with goal minimizing deviations in grade lithology. Secondly, ore-matching problem transformed partially observable Markov decision process, supply strategy continuously optimized feedback environmental indicators obtain decision-making sequence. Thirdly, algorithm introduced, which trained modeled environment strategy. Finally, taking large metal mine as example, depth verified via experiments, training displayed graphical interface. The experimental results show that constructed more line requirements mine. When compared traditional algorithm, efficiency accuracy solution have been greatly improved, calculation can be obtained real-time.