Industrial process pi control loop performance optimization method, apparatus, medium, and product

Abstract

This application discloses a method, device, medium, and product for optimizing the performance of a PI control loop in an industrial process, relating to the field of industrial automatic control technology. The method includes initializing the initial parameters of a deep reinforcement learning agent and a PI controller; generating setpoints for the industrial process based on a continuous smooth curve; acquiring the process variables of the current control cycle; calculating the tracking error and error rate of change based on the setpoints; outputting a reference adjustment amount for the PI controller parameters through a policy network; trimming the reference adjustment amount; updating the proportional gain and integral gain of the PI controller based on the trimmed reference adjustment amount; calculating the immediate reward of a composite reward function based on the tracking error and error rate of change; dynamically and adaptively adjusting the weight coefficients in the composite reward function; updating the evaluation network and policy network of the deep reinforcement learning agent; and outputting the optimized PI controller parameters. This application improves the smoothness and safety of the control process.

Images