Reinforcement learning control for gas collector pressure of coke ovens

Due to nonlinear and strong coupling of coke ovens, a distributed reinforcement learning control algorithm based on a radial basis function (RBF) for gas collector pressure control was proposed, in which the discrete space was dealt with by RBF and the critic-actor method was used to update the parameters of search network and evaluation network of the controller. Moreover, the control strategies of the controller were optimized by trial and error in the strong interference environment. The simulation of gas collector pressure control of coke ovens has shown that the algorithm can effectively solve the coupling problem, and has better stability.