Research On Soft Measurement Method And Policy Gradient Control In Wastewater Treatment Process

With the continuous development of science and technology,people’s quality of life has also improved.Modern biotechnology represented by cell engineering is widely used in many fields such as food,medicine,and marine environment.In recent years,as the process of wastewater biological treatment has become more and more complex,higher and higher requirements have been put forward for the control technology in the wastewater biological reaction process.Related research has also become a research hotspot in the field of biological reaction process of wastewater.This article starts with the complex and non-linear biological reaction process of wastewater,and studies the soft measurement and control strategy of the wastewater biological reaction process.The specific work includes the following two aspects.In the soft measurement of wastewater biological reaction process,this paper proposes a design method of soft sensor in the reaction process.This article focuses on key parameters that are difficult to measure in real time during the reaction process,such as biochemical indicators with a long detection time like five-day biochemical oxygen demand.In this paper,the entry point is the wastewater influent that has the greatest impact on the key variables of the reaction process,and the influent mode discriminator(IMD)is designed to discriminate and classify the wastewater influent mode.The discriminator is used to label different conditions(such as weather).Input the labeled influent data into a neural network with a specific structure for training and prediction,and finally realize real-time and effective monitoring of key parameters in the reaction process.Regarding the control strategy of the wastewater biological reaction process,this paper proposes a nonlinear prediction-based deep deterministic policy gradient(NPDDPG)control method,and combined with constraint conditions to optimize the reaction process.Aiming at the control task of the highly nonlinear wastewater biological reaction process,this paper regards the entire reaction process as an environment,introduces and improves the deep deterministic policy gradient method under the topic of reinforcement learning.On the one hand,it integrates the nonlinear prediction with specified time step,and set a threshold for the predicted output of the algorithm exploration value to reduce the original algorithm’s exploration of the invalid space.On the other hand,after the algorithm implements an exploration,the calculated result and the loss value in the prediction process are incorporated into the gradient of the deterministic policy network.Through the above two improvements,the proposed method is more directional in the training process,has a faster convergence speed,and also ensures the accuracy of the prediction process.Subsequently,according to the international standard wastewater treatment biochemical reaction process equation,a simulation platform was built to verify the monitoring and control method proposed in this paper.Combined with the soft measurement method of key variables,under the constraints of water quality,with the help of optimization algorithms,the controlled variable of the wastewater biological treatment process is optimized,and then the controlled variable setting value that makes the operating cost of the whole process the lowest is found.The final experimental results show that the biological reaction process monitoring and control method proposed in this paper is effective on the established international simulation platform.Compared with other methods,it has the advantages of fast response speed,higher accuracy,and stable performance.