Multi-objective Optimization Of Ozone Generation System Based On Deep Reinforcement Learning

Ozone has strong oxidative properties and disinfection properties,and is widely used in industrial sewage,food processing and other fields.Therefore,the preparation of ozone is in great demand in these fields.At present,scholars at home and abroad have conducted in-depth research on its characteristics and preparation methods,and have achieved a series of theoretical results,which have laid a strong foundation for our actual preparation of ozone.The method widely used in the field of industrial ozone production today is the dielectric barrier discharge method,which can produce highconcentration ozone.But there are still some problems with this method.The structure of the discharge chamber is relatively complex,and there are many internal parameters affecting ozone generation in the ozone generation system,so it cannot be modeled by mechanism.However,most of the existing ozone generators use open-loop control,which is not controlled according to the actual output and concentration requirements,so there are problems of high energy consumption and low efficiency.Therefore,it is necessary to conduct more in-depth research on system control.In order to improve the problems of low ozone generation efficiency and high energy consumption in the ozone generation system,it is necessary to improve the controllability and intelligence of the system.In this paper,the MFAC algorithm is deployed in the ozone generation system control,and a multi-objective optimization method for the ozone generation system based on deep reinforcement learning is proposed.Ozone has specific requirements for its concentration,response time and fluctuation in different application scenarios,so controlling these three goals is the key.Adjust the parameters of the MFAC algorithm under the framework of the DDPG algorithm,set the reward function of the DDPG algorithm according to the three key target values,and then adjust the MFAC parameters through the training of the neural network,and finally adjust the parameters to meet the three target values.In order to reflect the feasibility and superiority of the DDPG algorithm and the MFAC algorithm,this paper designs a comparison experiment between the MFAC algorithm and the traditional PID algorithm control,and a comparison experiment for the parameter adjustment between the DDPG algorithm and the PSO algorithm.The experimental results show that the control effect of the MFAC algorithm is better than that of the traditional PID algorithm for the ozone generation system.The parameter adjustment method of MFAC based on DDPG is better than the adjustment of PSO algorithm.Therefore,the multi-objective optimization method of ozone generation system based on deep reinforcement learning can ensure the stable operation of ozone generation system.And it can ensure that for different application scenarios,only need to change the setting of the reward function according to the demand,and after a period of learning,the ozone generation system can stably generate the corresponding ozone concentration.