Research On Optimal Control Method For Wastewater Treatment Process Based On ESN

Wastewater treatment process (WWTP) is a complex and dynamic process whichhas some characteristics of highly non-linear, large time-delay, multivariable, seriousdisturbance. Therefore, a precise mathematical model cannot be established. Theconventional control method cannot acquire satisfied control results. In order toimprove wastewater treatment capability, get satisfied effluent quality and reduceoperation cost, studying a new intelligent optimal control strategy for wastewatertreatment process is necessary. And it has importance theoretical significance as wellas practical value.Based on benchmark simulation model on.1(BSM1), a multivariable adaptivepredictive control strategy for WWTP is designed, and then an optimal controlmethod based on setpoints optimal model is proposed. Finally, an online predictiveoptimal control system is built. The main researches and innovation points of thispaper as follows:1. Deeply analyzed BSM1platform, which is provided by EuropeanCo-operation in the field of Science and Technology Research (COST) andInternational Water Association (IWA), is realized visualization in Matlabenvironment. Moreover, building BSM1lays a foundation for the research of newtype of control strategy.2. Due to the highly non-linear, large time-varying and time-delay characteristicsof wastewater treatment process, a multivariable adaptive predictive control strategy,based on echo state network (ESN), is proposed. First, a predictive model isestablished with ESN. Sencod, the ESN identifier is designed in order to calculate thedifferences between the identifier outputs and actual outputs to compensate theproduced errors when build the predictive model. The proposed multivariableadaptive control strategy is compared with PID control strategy and predictive controlbased on BP model on the BSM1platform. The experimental results show that thiscontrol method improves the control precision of dissolved oxygen concentration andnitrate nitrogen concentration.3. Analyzing the relationship between influent flow rate, influent componentsand control variable setpoints, setpoints optimal model (SOM) is established. Thesetpoints optimal model is the upper layer of the control system, meanwhile, the bottom layer adopt the ESN-based multivariate adaptive predictive control strategy, toachieve optimal control based on the SOM. The proposed method does not requirecomplicated calculation, and experimental results show that the SOM can produceoptimized setpoints according to the influent state, the SOM-based control methodcan make the effluent meet the requirement and reduce the energy consumption of theWWTP.4. In order to solve the problem of high energy consumption and adapt real-timefor WWTP, an online predictive optimal control system based on ESN for WWTP isproposed. The ESN optimal control algorithm is designed by studying various optimalalgorithm and is used to control the dissolved oxygen concentration and nitratenitrogen concentration. Experimental results indicate that the optimal control strategycan optimize the bottom controller’s setpoints dynamically and make the energylower.