Studies On Intelligent Prediction And Optimal Control Methods Of Oxygen Content In Flue Gas For Municipal Solid Waste Incineration Process

Incineration is an important municipal solid waste(MSW)treatment method with a significant reduction in volume,recovery of energy,and complete disinfection,representing mainstream technology.The oxygen content in flue gas is an important process parameter,closely related to combustion efficiency and nitrogen oxide(NOx)emission.There are a variety of physical and chemical reactions in the municipal solid waste incineration(MSWI)process.The efficient and stable operation of the MSWI plant is inseparable from the intelligent prediction and optimal control methods of oxygen content in flue gas.However,there are several challenges in designing the intelligent prediction and optimal control scheme of oxygen content in flue gas.First,the mechanism of the MSWI process is complex and the working condition fluctuates frequently.It is difficult to describe or predict the dynamic behavior of the oxygen content in flue gas accurately.Second,in the actual incineration process,the oxygen content in flue gas is difficult to be effectively controlled by conventional control methods,along with some problems such as model mismatch,environmental interference,and frequent operation of actuators.Third,combustion efficiency and NOx emission are conflicting operation indicators,which cannot be optimized at the same time.It is difficult to solve the accurate optimal set points of oxygen content in flue gas to realize the optimal operation of the incineration process.Therefore,based on the analysis of the mechanism in the MSWI process and exploration of the internal relationship between the manipulated variables and the oxygen content in flue gas,it is necessary to study the quantitative description and intelligent prediction method of the dynamic characteristics of oxygen content in flue gas,design the data-driven predictive control and event-based adaptive predictive control approach,and exploit the optimal setting method and intelligent optimal control strategy.Besides,the collaborative improvement of combustion efficiency and denitrification efficiency is achieved through the stable control of oxygen content in flue gas.The main research work and innovation of this thesis are as follows:(1)Design on intelligent modeling of MSWI process oriented to the control of oxygen content in flue gasIn order to accurately describe the dynamic characteristics of oxygen content in flue gas for the MSWI process,an intelligent modeling method of the MSWI process oriented to the control of oxygen content in flue gas is proposed.First,the dynamic characteristics of oxygen content in flue gas under specific working conditions are studied,and manipulated variables related to the oxygen content in flue gas are preliminary selected based on the incineration mechanism and empirical knowledge.Second,the correlation information contained in the data is fully mined,and the input variables of the process model are determined by the Pearson correlation coefficient feature selection method.Finally,an intelligent model based on an adaptive fuzzy neural network is constructed,and Levenberg-Marquardt(LM)algorithm is used to update the model parameters to accurately describe the dynamic characteristics of the oxygen content in flue gas.(2)Research on intelligent prediction of oxygen content in flue gas based on self-organizing LSTM neural networkTo accurately predict the oxygen content in flue gas for the MSWI process,a data-driven prediction model based on self-organizing long short-term memory(SOLSTM)neural network is proposed.First,a prediction model based on the self-organizing long and short-term memory network was designed to dynamically adjust the hidden layer structure combined with neuron activity and significance,improving the prediction accuracy of oxygen content in flue gas.Second,the weight and bias are updated by the time back propagation algorithm to ensure the convergence of estimation errors.Then,based on the Lyapunov theory,the convergence of the designed model is analyzed to ensure its feasibility in practical application.Finally,the validity of the intelligent prediction method is verified based on the experiment of butane concentration prediction in debutane tower and the experiment of oxygen content in flue gas prediction.(3)Design on data-driven model predictive control of oxygen content in flue gasIn order to accurate control the oxygen content in flue gas for the MSWI process,a data-driven model predictive control method is exploited.First,a model predictive control scheme based on the SOLSTM neural network is designed,and a prediction model of oxygen content in flue gas based on the SOLSTM neural network is established.Second,the gradient descent method is utilized to solve the control law to control the oxygen content in flue gas.Then,the stability of the control system is analyzed based on the Lyapunov stability theory,which ensures the reliability of the control system in practical application.Finally,the effectiveness of the proposed control method is verified by control experiments of a nonlinear dynamic system and oxygen content in flue gas.(4)Research on event-based adaptive predictive control of oxygen content in flue gasTo accurately and efficiently control the oxygen content in flue gas for the MSWI process,an event-based adaptive predictive control strategy is developed.First,an event-triggering mechanism is proposed to synthesize the prediction accuracy and control accuracy,which reduces the computational burden of online optimization and improves the efficiency of control law optimization.Secondly,an error-triggered online learning mechanism is designed to update model parameters adaptively based on prediction errors,solving the problem of model mismatch caused by uncertain interference and equipment aging.In addition,the stability of the proposed control strategy is analyzed to ensure the safe and stable operation of the controller in practical application.Finally,simulation comparisons are provided to show control accuracy and control efficiency of the proposed event-based adaptive predictive control strategy via control experiments of a nonlinear dynamic system and oxygen content in flue gas.(5)Design on the optimal setting method of oxygen content in flue gas for the MSWI processIn order to achieve the combustion optimization for the MSWI process,an optimal setting method of oxygen content in flue gas is proposed.First,an online adaptive fuzzy neural network is designed to construct the objective functions to estimate the combustion efficiency and NOx emission concentration.Second,the multi-objective particle swarm optimization algorithm based on space partition and hybrid distance(SPHD-MOPSO)is developed to obtain the optimal set points of oxygen content in flue gas.Third,the 1-norm distance between all Pareto solutions and utopian points in the external archive is calculated,and the one with the smallest distance is taken as the execution solution,which can automatically obtain the optimal setting value of the oxygen content in flue gas.The experimental results demonstrate that the optimal setting method can achieve the dynamic acquisition of the set points of the oxygen content in flue gas,improve combustion efficiency,and reduce NOx emission concentration.(6)Research on intelligent optimal control of oxygen content in flue gas for the MSWI processTo achieve the practical requirements of operation optimization and oxygen content control in the MSWI process,a multiobjective optimal control strategy with utopia-decision is developed.First,the intelligent optimal control scheme of oxygen content in flue gas for the MSWI process is proposed to accurately describe the multi-objective optimal control problem.Second,the optimal set points of oxygen content in flue gas are obtained based on the SPHD-MOPSO algorithm and the utopian decision method.Third,a double LSTM neural networks-based model predictive control strategy is exploited.The prediction model and error feedback correction model are designed based on LSTM neural networks to improve the tracking control accuracy of optimal set points.The experimental results show that the proposed intelligent optimal control strategy can achieve better dynamic optimization and tracking control performance of oxygen content in flue gas,which can realize the optimal operation of the MSWI process.