Research On Control Method Of Natural Gas Desulfurization Process Based On Adaptive Dynamic Programming

The absorption tower is an important equipment for natural gas desulfurization,and its desulfurization effect directly affects the quality of natural gas products.Due to the complexity of natural gas desulfurization process,strong non-linearity,uncertainty,coupling and other characteristics,it is difficult to establish an accurate mathematical model of absorption tower desulfurization process,making it difficult to control.The existing control schemes for desulfurization process of absorption tower mostly adopt proportional-integral control with simple loop,without establishing the correlation among control parameters and relying too much on expert experience.Under the circumstances of large variation of feed gas parameters and environmental changes,the cushioning and stability of control system are poor,and the adjustment time of disturbance is longer,which will lead to poor quality of purification gas.And even cause major safety accidents.Taking the desulfurization process of methyl diethanolamine(MDEA)in Puguang gas field of China as an example,On the basis of domestic and foreign scholars' research,the model of desulfurization process in absorption tower is established by back propagation(BP)neural network.With this model as the control object,the control method of desulfurization process of absorber based on adaptive dynamic programming(ADP)is studied.The main contents of this paper are as follows:(1)Establishing the model of absorption tower desulfurization process.By analyzing the desulfurization process of absorption tower,the key parameters affecting the desulfurization effect are selected.Using the strong mapping ability of BP neural network,the actual production data of absorption tower desulfurization process are directly trained,and the relationship between input and output is established.Thus,the model of absorption tower desulfurization process is obtained,which overcomes the difficulty of modeling absorption tower in extremely complex industrial environment.The analysis of the training process and the test results of the model shows that the model can accurately reflect the relationship between input and output in the desulfurization process of absorption tower,and has good applicability and generalization ability.(2)Design and improvement of desulfurization process control strategy for absorption tower based on action-dependent heuristic dynamic programming(ADHDP).The traditional ADHDP algorithm is used to design the desulfurization process controller of the absorption tower.The basic structure and the realization scheme using neural network are given.Aiming at the problems of long training time and slow convergence speed of this method,Unscented Kalman Filter(UKF)algorithm is used to update the weights of critic network and action network.The specific process of updating optimization weights by UKF algorithm is given,and the control strategy based on UKF-ADHDP is designed.The simulation results show that the UKFADHDP control method has better control performance than the traditional ADHDP control method.(3)Desulfurization process control of absorber based on event-triggered heuristic dynamic programming(HDP).Combining with event triggering mechanism,an event triggering HDP controller is designed,and its basic structure and neural network implementation are given.A trigger threshold is designed for event triggering HDP controller,and a novel trigger condition for aperiodic sampling events is formulated.The stability of the system under this sampling rule is analyzed.The simulation results show that compared with the traditional HDP control method,the event triggered HDP control method not only ensures the stability of the control system,but also greatly saves the computational cost.