Method for controlling flue gas oxygen content in urban solid waste incineration process
Through random configuration of network and adaptive update strategies, the flue gas oxygen content prediction model is established, combined with PID control and event triggering technology, the accuracy and stability of flue gas oxygen content control during urban solid waste incineration is solved, and the improvement is achieved High prediction accuracy and response speed reduce the computational burden.
Patent Information
- Application Number
- CN202411475761.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-22
- Publication Date
- 2024-12-03
AI Technical Summary
It is difficult to achieve accurate and stable control of flue gas oxygen content during urban solid waste incineration. Traditional methods lack control accuracy when facing complex nonlinear and time-varying characteristics, and model parameters are difficult to fix and predict.
A random configuration network is used to establish a prediction model of flue gas oxygen content, and the model parameters are adjusted through adaptive update strategies. Combined with PID control and event triggering technology, an event triggering model prediction controller based on PID is designed.
It improves the prediction accuracy and control accuracy of the oxygen content of the flue gas, enhances the robustness and response speed of the system, reduces the computing burden, and improves the real-time and operating efficiency of the system.
Smart Images

Figure CN119063006A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of industrial control, and in particular to a method for controlling the oxygen content of flue gas in a municipal solid waste incineration process. Background Art
[0002] Incineration of municipal solid waste is one of the important means of modern waste treatment. High-temperature combustion can effectively reduce the volume of waste, destroy harmful components and generate treatable ash. However, the flue gas generated during the incineration process contains a large amount of pollutants and harmful gases, which pose a threat to the environment and human health. In order to ensure the efficient and environmentally friendly operation of the incineration process, the control of the oxygen content in the flue gas is crucial.
[0003] The composition of urban solid waste is complex and varied. Different types of waste will show different combustion characteristics during the incineration process, which makes it extremely difficult to establish an accurate mathematical model. Traditional modeling methods are usually unable to fully capture the dynamic changes of all influencing factors during the incineration process, resulting in large deviations in the model in practical applications. Secondly, the incineration process itself has strong nonlinear characteristics. Parameters such as temperature, pressure, and air flow rate are constantly changing during the combustion process. These factors affect each other, further increasing the complexity of modeling. In addition, equipment aging and changes in the external environment will also have a significant impact on the incineration process, making it difficult to fix and predict model parameters.
[0004] In addition, the commonly used flue gas oxygen content control methods currently use traditional PID control strategies or experience-based adjustment methods. These methods often show problems of insufficient control accuracy and slow response speed when dealing with complex incineration processes with nonlinear and time-varying characteristics. In addition, various interference factors in the incineration process, such as changes in fuel composition, equipment aging, and changes in the external environment, also pose challenges to the stable control of flue gas oxygen content.
[0005] Therefore, a more advanced modeling method and intelligent control method are needed to adjust the flue gas oxygen content in real time during the complex incineration process and improve the stability and control accuracy of the system. Based on this background, the present invention proposes a control method for flue gas oxygen content in the process of municipal solid waste incineration, which realizes accurate prediction and control of flue gas oxygen content by combining model predictive control technology, PID control technology and event triggering technology. Summary of the invention
[0006] The present invention provides a method for controlling the oxygen content of flue gas in the process of urban solid waste incineration, which is used to solve the above technical problems. First, a prediction model for the oxygen content of flue gas in the process of urban solid waste incineration is established according to a random configuration network; second, the parameters of the prediction model are adjusted online through an adaptive update strategy; then, according to the prediction model, a PID-based model prediction controller is designed; finally, according to the PID-based model prediction controller, an event triggering strategy is introduced to obtain a PID-based event triggering model prediction controller.
[0007] A method for controlling the oxygen content of flue gas during the incineration of municipal solid waste, characterized in that it comprises the following steps:
[0008] S1. Establish a prediction model for flue gas oxygen content in the process of municipal solid waste incineration based on a random configuration network;
[0009] S2. adjusting the parameters of the prediction model online through an adaptive update strategy;
[0010] S3. Designing a PID-based model predictive controller according to the prediction model;
[0011] S4. According to the PID-based model predictive controller, an event triggering strategy is introduced to obtain a PID-based event triggering model predictive controller.
[0012] Furthermore, a prediction model for flue gas oxygen content in the process of municipal solid waste incineration is established based on the random configuration network, and the prediction model is expressed as:
[0013] y p =f(u,y);
[0014] Where u is the input sequence of the random configuration network, y is the actual output value, and y p is the output value of the random configuration network, f(u,y) is the prediction model function, and when the random configuration network has L-1 nodes, f(u,y) is expressed as:
[0015]
[0016] Among them, L-1 is the number of nodes, β j is the output weight of the jth node in the hidden layer, g j (*) is the activation function of the jth node in the hidden layer, ω j is the input weight of the jth node in the hidden layer, b j is the bias of the jth node in the hidden layer.
[0017] Furthermore, the residual of the prediction model is defined as:
[0018] e L-1 =yyp ;
[0019] when ||e L-1 ||>e max , then add a new node as follows:
[0020]
[0021]
[0022] Among them, e max is the maximum allowable residual, h L is an intermediate variable, r is the randomly configured network parameter, and 0 <r<1,μ L is a sequence of non-negative real numbers that satisfies μ L <1-r,g L (*) is the activation function of the Lth node in the hidden layer, ω L is the input weight of the Lth node in the hidden layer, b L is the bias of the Lth node in the hidden layer, N represents the number of elements in the input sequence, u1 is the first value of the input sequence, u2 is the second value of the input sequence, and u N is the Nth value of the input sequence;
[0023] The random configuration network will L The candidate node corresponding to the maximum value is added to the hidden layer as the optimal node, and the hidden layer output weight is recalculated. The hidden layer output weight calculation formula is:
[0024]
[0025] Among them, β1 is the output weight of the first node in the hidden layer, β2 is the output weight of the second node in the hidden layer, and β L is the output weight of the Lth node in the hidden layer.
[0026] Furthermore, the parameters of the prediction model are adjusted online through an adaptive update strategy, and the cost function of the adaptive update of the parameters of the prediction model is defined as follows:
[0027]
[0028] Δθ(t k )=θ(t k )-θ(t k-1 );
[0029] Among them, t k is the current sampling time, L(t k ) is the cost function, y p (t k ) is tk The output value of the prediction model at time y(t k ) is t k The actual value at the moment, γ is the weight coefficient, i is the intermediate variable, n θ is the number of parameters of the prediction model, Δθ(t k ) is an intermediate variable, x(t k ) is t k The input value of the random configuration network at time θ(t k ) is t k The parameter value of the prediction model at time θ(t k-1 ) is t k-1 The parameter values of the prediction model at the time;
[0030] The optimization algorithm is used to update the parameters of the prediction model to obtain t k The cost function L(t k ) is the smallest parameter of the prediction model.
[0031] Further, the PID controller is designed as follows:
[0032]
[0033] e(t k )=y r (t k )-y(t k );
[0034] Among them, u(t k ) is t k The output value of the PID controller at time e(t k ) is t k The error at the moment, e(t k-1 ) is t k-1 The error of time, K p , K d and K i is the PID controller parameter, j is the intermediate variable, k is the current sampling number, and T is the predictive control time domain.
[0035] Furthermore, according to the prediction model, the model prediction controller is designed as follows:
[0036]
[0037] Δu(t k+j )=u(t k+j )-u(t k+j-1 );
[0038] Among them, J(Δu,t k) is the cost function, λ is the control parameter, t k is the current sampling time, j is the intermediate variable, N p is the number of prediction times, N u is the prediction control time domain number, y r (t k+j ) is t k+j Expected value at any time, t k+j The predicted value at the moment, Δu(t k+j ) is t k+j The change value of the control quantity is predicted at each moment, u(t k+j ) is t k+j The predicted control quantity at each moment, u(t k+j-1 ) is t k+j-1 Predict control quantity at all times;
[0039] y r (t k+j ) is calculated as follows:
[0040]
[0041] Among them, α is the adjustment factor, 0<α<1, y r (t k ) is t k Expected value at time, y r (t k+j-1 ) is t k+j-1 Expected value at time, y sp (t k ) is the reference setting value;
[0042] The calculation formula is as follows:
[0043]
[0044] Among them, e1(t k ) is t k The error between the output value of the prediction model and the actual output value at the moment, y p (t k+j ) is t k+j The output value of the moment prediction model, ε is the compensation coefficient.
[0045] Further, according to the PID controller and the model predictive controller, the PID-based model predictive controller is obtained as follows:
[0046] U(t k )=u(t k )+Δu(t k );
[0047] Among them, U(tk ) is the final control input, u(t k ) is the control input obtained by the PID controller, Δu(t k ) is the control input change obtained by the model predictive controller.
[0048] Furthermore, according to the model predictive controller, an event triggering strategy is introduced as follows:
[0049] Fixed threshold trigger:
[0050] |y sp (t k )-y(t k )|≥κ;
[0051] Among them, κ is the set threshold;
[0052] Fixed time triggers are:
[0053] t k -t ke ≥t kmax ;
[0054] Among them, t ke is the last triggering time, t kmax The maximum time limit allowed.
[0055] Furthermore, according to the PID-based model predictive controller, an event triggering strategy is introduced to obtain a PID-based event triggered model predictive controller, that is, the model predictive controller performs calculations only when a predetermined triggering condition is met.
[0056] Compared with the prior art, the present invention has the following beneficial effects:
[0057] 1. A prediction model for flue gas oxygen content in the process of municipal solid waste incineration is established using a random configuration network, and the prediction model parameters are adjusted in real time through an adaptive update strategy, so that the prediction model can accurately reflect the complex dynamic changes in the incineration process. Compared with traditional methods, this prediction model has higher prediction accuracy. This adaptive ability enables the system to maintain high control performance in the face of external disturbances such as fuel composition changes, equipment aging, and environmental changes, significantly improving the robustness of the system.
[0058] 2. By combining PID control with model predictive control, the control system not only has the fast response capability of traditional PID control, but can also adjust future trends according to model predictions, improving control accuracy and response speed. The PID controller is responsible for quickly correcting instantaneous errors, while the model predictive controller is responsible for predicting future errors and correcting them in advance. This dual control strategy can more effectively cope with complex changes in the incineration process, reduce overshoot and steady-state errors, and thus achieve more accurate and stable flue gas oxygen content control.
[0059] 3. By designing an event-triggered model predictive controller, the controller calculation is performed only when specific trigger conditions are met. This event-triggered strategy greatly reduces the amount of controller calculation and reduces the system's calculation burden. Compared with the traditional continuous calculation method, the event-triggered control strategy can effectively reduce unnecessary calculations and improve the system's real-time performance and operating efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0060] Figure 1 It is a schematic diagram of the process of the present invention;
[0061] Figure 2 It is the control system block diagram;
[0062] Figure 3 It is the calculation flow chart of the event-triggered model predictive controller based on PID. DETAILED DESCRIPTION
[0063] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0064] Incineration of municipal solid waste is one of the important means of modern waste treatment. High-temperature combustion can effectively reduce the volume of waste, destroy harmful components and generate treatable ash. However, the control of flue gas oxygen content during the incineration process is one of the key factors to ensure efficient and environmentally friendly operation of the incineration process. The control methods commonly used in the prior art mostly adopt traditional PID control strategies or experience-based adjustment methods. These methods often show problems of insufficient control accuracy and slow response speed when dealing with complex, nonlinear and time-varying incineration processes. In addition, the accurate modeling of flue gas oxygen content in the process of municipal solid waste incineration also faces huge challenges. The existing models are difficult to accurately reflect the complex dynamic changes in the incineration process. In order to solve these problems, the present invention provides a method for controlling the oxygen content of flue gas in the process of municipal solid waste incineration. In order to illustrate the feasibility of the present invention, it will be described in detail in conjunction with specific embodiments below.
[0065] Embodiment 1
[0066] In the embodiment of the present application, a method for controlling the oxygen content of flue gas during the incineration of municipal solid waste is adopted. The specific implementation process is as follows Figure 1 As shown; first, a prediction model for flue gas oxygen content in the process of urban solid waste incineration is established based on a random configuration network; second, the parameters of the prediction model are adjusted online through an adaptive update strategy; then, according to the prediction model, a PID-based model prediction controller is designed; finally, according to the PID-based model prediction controller, an event triggering strategy is introduced to obtain a PID-based event triggering model prediction controller. The present invention can not only accurately control the flue gas oxygen content during the incineration process, improve the response speed and control accuracy of the system, but also effectively reduce the calculation burden, improve the real-time performance and calculation efficiency of the system, thereby ensuring the high efficiency and environmental protection of the incineration process.
[0067] A prediction model for flue gas oxygen content in the process of municipal solid waste incineration is established based on the random configuration network. The prediction model is expressed as:
[0068] y p =f(u,y);
[0069] Where u is the input sequence of the random configuration network, y is the actual output value, and y p is the output value of the random configuration network, f(u,y) is the prediction model function, and when the random configuration network has L-1 nodes, f(u,y) is expressed as:
[0070]
[0071] Among them, L-1 is the number of nodes, β j is the output weight of the jth node in the hidden layer, g j (*) is the activation function of the jth node in the hidden layer, ω j is the input weight of the jth node in the hidden layer, b j is the bias of the jth node in the hidden layer.
[0072] In an embodiment of the present application, a prediction model for flue gas oxygen content in the process of municipal solid waste incineration is established based on the random configuration network, which can effectively capture complex nonlinear relationships. The random configuration network uses its unique network structure and adaptive capabilities to effectively handle uncertainties in multivariable, nonlinear and time-varying systems. The prediction model uses the relationship between the input sequence and the actual output value and the output value of the random configuration network for estimation. The input sequence and output value of the random configuration network are nonlinearly mapped through activation functions and weights, thereby achieving high-precision prediction of flue gas oxygen content.
[0073] Furthermore, the residual of the prediction model is defined as:
[0074] e L-1 =yy p ;
[0075] when ||e L-1 ||>e max , then add a new node as follows:
[0076]
[0077]
[0078] Among them, e max is the maximum allowable residual, h L is an intermediate variable, r is the randomly configured network parameter, and 0 <r<1,μ L is a sequence of non-negative real numbers that satisfies μ L <1-t,g L (*) is the activation function of the Lth node in the hidden layer, ω L is the input weight of the Lth node in the hidden layer, b L is the bias of the Lth node in the hidden layer, N represents the number of elements in the input sequence, u1 is the first value of the input sequence, u2 is the second value of the input sequence, and u N is the Nth value of the input sequence;
[0079] The random configuration network will L The candidate node corresponding to the maximum value is added to the hidden layer as the optimal node, and the hidden layer output weight is recalculated. The hidden layer output weight calculation formula is:
[0080]
[0081] Among them, β1 is the output weight of the first node in the hidden layer, β2 is the output weight of the second node in the hidden layer, and β L is the output weight of the Lth node in the hidden layer.
[0082] In the embodiment of the present application, the residual of the prediction model is defined, and new nodes are added based on this. When the residual of the prediction model exceeds the set threshold, the model is corrected by adding new nodes, thereby improving the prediction accuracy and adaptability. The random configuration network calculates the candidate nodes corresponding to the maximum value and recalculates the output weights of the hidden layer, thereby optimizing the network structure, enabling it to better fit complex nonlinear relationships and time-varying characteristics, and improving the accuracy and robustness of the prediction model.
[0083] Furthermore, the parameters of the prediction model are adjusted online through an adaptive update strategy, and the second cost function of the adaptive update of the parameters of the prediction model is defined as follows:
[0084]
[0085] Δθ(t k )=θ(t k )-θ(t k -1);
[0086] Among them, t k is the current sampling time, L(t k ) is the second cost function, y p (t k ) is t k The output value of the prediction model at time y(t k ) is t k The actual value at the moment, γ is the weight coefficient, i is the intermediate variable, n θ is the number of parameters of the prediction model, Δθ(t k ) is an intermediate variable, x(t k ) is t k The input value of the random configuration network at time θ(t k ) is t k The parameter value of the prediction model at time θ(t k -1) is t k -1 The parameter value of the prediction model at time;
[0087] The optimization algorithm is used to update the parameters of the prediction model to obtain t k The cost function L(t k ) is the smallest parameter of the prediction model. In the embodiment of the present application, the parameters of the prediction model are updated using the gradient descent method. The parameter update formula of the prediction model is:
[0088]
[0089] Among them, θ(t k +1) is t k +1 is the parameter value of the prediction model at time ζ, and ζ is the learning rate.
[0090] In the embodiment of the present application, the parameters of the prediction model are adjusted online through the adaptive update strategy, and the present invention can respond to various changes in the incineration process in real time and adjust the control strategy in time. This adaptive capability enables the system to maintain high control performance in the face of external disturbances such as fuel composition changes, equipment aging and environmental changes, and significantly improves the robustness of the system.
[0091] Further, the PID controller is designed as follows:
[0092]
[0093] e(t k )=y r (t k )-y(t k );
[0094] Among them, u(t k ) is t k The output value of the PID controller at time e(t k ) is t k The error at the moment, e(t k-1 ) is t k-1 The error of time, K p , K d and K i is the PID controller parameter, j is the intermediate variable, k is the current sampling number, and T is the predictive control time domain.
[0095] In the embodiment of the present application, a PID controller is established, which can calculate the influence of the current error and the historical error on the control quantity in real time, thereby adjusting the output. This design can ensure that the system can respond quickly, improve the control accuracy of the flue gas oxygen content in the municipal solid waste incineration process, and maintain a stable control effect in complex environments and variable working conditions, thereby improving the reliability and operation efficiency of the system.
[0096] Furthermore, according to the prediction model, the model prediction controller is designed as follows:
[0097]
[0098] Δu(t k+j )=u(t k+j )-u(t k+j-1 );
[0099] Among them, J(Δu,t k ) is the cost function, λ is the control parameter, t k is the current sampling time, j is the intermediate variable, N p is the number of prediction times, N u is the prediction control time domain number, y r (t k+j ) is t k+j Expected value at any time, t k+j The predicted value at the moment, Δu(t k+j ) is t k+j The change value of the control quantity is predicted at each moment, u(t k+j ) is tk+j The predicted control quantity at each moment, u(t k+j-1 ) is t k+j-1 Predict control quantity at all times;
[0100] y r (t k+j ) is calculated as follows:
[0101]
[0102] Among them, α is the adjustment factor, 0<α<1, y r (t k ) is t k Expected value at time, y r (t k+j-1 ) is t k+j-1 Expected value at time, y sp (t k ) is the reference setting value;
[0103] The calculation formula is as follows:
[0104]
[0105] Among them, e1(t k ) is t k The error between the output value of the prediction model and the actual output value at the moment, y p (t k+j ) is t k+j The output value of the moment prediction model, ε is the compensation coefficient.
[0106] In an embodiment of the present application, the model predictive controller designed according to the prediction model can effectively achieve accurate prediction and real-time adjustment of system behavior by dynamically adjusting control parameters, combining time domain prediction and real-time feedback. Through the optimization of the cost function, the controller can respond and adjust quickly within the current sampling time, thereby achieving accurate tracking and control of the expected value at the moment. In the design of the model predictive controller, the introduced adjustment factor and compensation coefficient effectively improve the stability and robustness of the control system. The adjustment factor dynamically adjusts the control parameters according to the comparison between the prediction error and the reference set value to optimize the response speed and accuracy of the controller. The compensation coefficient is corrected in real time according to the error between the output value of the prediction model at the moment and the actual output value to ensure the stability and accuracy of the control quantity.
[0107] Further, according to the PID controller and the model predictive controller, the PID-based model predictive controller is obtained as follows:
[0108] U(t k )=u(t k )+Δu(t k);
[0109] Among them, U(t k ) is the final control input, u(t k ) is the control input obtained by the PID controller, Δu(t k ) is the control input change obtained by the model predictive controller.
[0110] In the embodiment of the present application, by combining PID control with model predictive control, the control system not only has the fast response capability of traditional PID control, but also can adjust future trends according to model prediction, thereby improving control accuracy and response speed. The PID controller is responsible for quickly correcting instantaneous errors, while the model predictive controller is responsible for predicting future errors and correcting them in advance. This dual control strategy can more effectively cope with complex changes in the incineration process, reduce overshoot and steady-state errors, and thus achieve more accurate and stable flue gas oxygen content control.
[0111] Furthermore, according to the model predictive controller, an event triggering strategy is introduced as follows:
[0112] Fixed threshold trigger:
[0113]
[0114] in, is the set threshold;
[0115] Fixed time triggers are:
[0116] t k -t ke ≥t kmax ;
[0117] Among them, t ke is the last triggering time, t kmax The maximum time limit allowed.
[0118] In an embodiment of the present application, an event triggering strategy is introduced according to the model predictive controller, which effectively optimizes the operating efficiency and resource utilization of the control system. The strategy combines two mechanisms, fixed threshold triggering and fixed time triggering, and aims to intelligently determine when to perform prediction and control operations according to the current state of the system and control requirements, thereby minimizing unnecessary calculations and resource consumption. The fixed threshold triggering strategy sets a threshold for the prediction error, and triggers the controller to perform prediction and adjustment operations when the prediction error exceeds the set threshold to ensure that the control system can respond and adjust in a timely manner. This strategy effectively avoids frequent calculations when the system state changes insignificantly, thereby improving the operating efficiency and energy efficiency of the system. At the same time, the fixed time triggering strategy sets a maximum allowable time limit, and forces the control operation to be triggered after exceeding the set time, ensuring that the control system can maintain stability and reliability during long-term operation. This strategy is suitable for scenarios where control operations need to be performed regularly, and effectively avoids the performance degradation or failure of the control system due to time delays.
[0119] Furthermore, according to the PID-based model predictive controller, an event triggering strategy is introduced to obtain an event-triggered model predictive controller based on PID, that is, the model predictive controller performs calculation only when a predetermined triggering condition is met. The specific process is as follows: Figure 2 shown.
[0120] In the embodiment of the present application, by designing a PID-based event-triggered model predictive controller, the controller calculation is performed only when specific trigger conditions are met. This event-triggered strategy greatly reduces the amount of calculation of the controller and reduces the calculation burden of the system. Compared with the traditional continuous calculation method, the event-triggered control strategy can effectively reduce unnecessary calculations and improve the real-time performance and operation efficiency of the system.
[0121] Embodiment 2
[0122] Figure 3The control system architecture of flue gas oxygen content in the process of municipal solid waste incineration is presented. The system includes components such as PID controller, model predictive controller, feedback correction module and prediction model. The operation process of the whole system is as follows: First, the set value is input into the PID controller, and the PID controller outputs a control signal according to the deviation between the set value and the actual value, and the signal controls the actual municipal solid waste incineration process. The actual municipal solid waste incineration process produces an actual flue gas oxygen content, which is used to update the prediction model to predict the future flue gas oxygen content. At the same time, the adaptive parameter update module also adjusts the parameters of the prediction model according to the deviation between the actual value and the predicted value, thereby improving the prediction accuracy. The feedback correction module compares the predicted value with the actual value, and adjusts the control signal according to the difference to keep the actual value close to the set value. In addition, the event trigger module also plays a role in monitoring the deviation between the actual value and the set value. When the deviation exceeds the threshold, the corresponding event can be triggered, such as an alarm or adjustment of the control strategy. Through the coordinated action of these components, the system can stably and accurately control the flue gas oxygen content in the process of municipal solid waste incineration.
[0123] In the control system, a prediction model for flue gas oxygen content in the municipal solid waste incineration process is established according to the random configuration network, and the prediction model is expressed as:
[0124] y p =f(u,y);
[0125] Where u is the input sequence of the random configuration network, y is the actual output value, and y p is the output value of the random configuration network, f(u,y) is the prediction model function, and when the random configuration network has L-1 nodes, f(u,y) is expressed as:
[0126]
[0127] Among them, L-1 is the number of nodes, β j is the output weight of the jth node in the hidden layer, g j (*) is the activation function of the jth node in the hidden layer, ω j is the input weight of the jth node in the hidden layer, b j is the bias of the jth node in the hidden layer.
[0128] Furthermore, the residual of the prediction model is defined as:
[0129] e L-1 =yy p ;
[0130] when ||e L-1 ||>e max , then add a new node as follows:
[0131]
[0132]
[0133] Among them, e max is the maximum allowable residual, h L is an intermediate variable, r is the randomly configured network parameter, and 0 <r<1,μ L is a sequence of non-negative real numbers that satisfies μ L <1-r,g L (*) is the activation function of the Lth node in the hidden layer, ω L is the input weight of the Lth node in the hidden layer, b L is the bias of the Lth node in the hidden layer, N represents the number of elements in the input sequence, u1 is the first value of the input sequence, u2 is the second value of the input sequence, and u N is the Nth value of the input sequence;
[0134] The random configuration network will L The candidate node corresponding to the maximum value is added to the hidden layer as the optimal node, and the hidden layer output weight is recalculated. The hidden layer output weight calculation formula is:
[0135]
[0136] Among them, β1 is the output weight of the first node in the hidden layer, β2 is the output weight of the second node in the hidden layer, and β L is the output weight of the Lth node in the hidden layer.
[0137] Furthermore, the parameters of the prediction model are adjusted online through an adaptive update strategy, and the second cost function of the adaptive update of the parameters of the prediction model is defined as follows:
[0138]
[0139] Δθ(t k )=θ(t k )-θ(t k -1);
[0140] Among them, t k is the current sampling time, L(t k ) is the second cost function, y p (t k ) is t k The output value of the prediction model at time y(t k ) is t k The actual value at the moment, γ is the weight coefficient, i is the intermediate variable, n θis the number of parameters of the prediction model, Δθ(t k ) is an intermediate variable, x(t k ) is t k The input value of the radial basis neural network at the moment, θ(t k ) is t k The parameter value of the prediction model at time θ(t k -1) is t k -1 The parameter value of the prediction model at time;
[0141] The optimization algorithm is used to update the parameters of the prediction model to obtain t k The cost function L(t k ) is the smallest parameter of the prediction model. In this embodiment of the present application, a genetic algorithm is used to update the parameters of the prediction model, and the genetic algorithm will not be described in detail here.
[0142] Further, the PID controller is designed as follows:
[0143]
[0144] e(t k )=y r (t k )-y(t k );
[0145] Among them, u(t k ) is t k The output value of the PID controller at time e(t k ) is t k The error at the moment, e(t k-1 ) is t k-1 The error of time, K p , K d and K i is the PID controller parameter, j is the intermediate variable, k is the current sampling number, and T is the predictive control time domain.
[0146] Furthermore, according to the prediction model, the model prediction controller is designed as follows:
[0147]
[0148] Δu(t k+j )=u(t k+j )-u(t k+j-1 );
[0149] Among them, J(Δu,t k ) is the cost function, λ is the control parameter, t k is the current sampling time, j is the intermediate variable, Np is the number of prediction times, N u is the prediction control time domain number, y r (t k+j ) is t k+j Expected value at any time, t k+j The predicted value at the moment, Δu(t k+j ) is t k+j The change value of the control quantity is predicted at each moment, u(t k+j ) is t k+j The predicted control quantity at each moment, u(t k+j-1 ) is t k+j-1 Predict control quantity at all times;
[0150] y r (t k+j ) is calculated as follows:
[0151]
[0152] Among them, α is the adjustment factor, 0<α<1, y r (t k ) is t k Expected value at time, y r (t k+j-1 ) is t k+j-1 Expected value at time, y sp (t k ) is the reference setting value;
[0153] The calculation formula is as follows:
[0154]
[0155] Among them, e1(t k ) is t k The error between the output value of the prediction model and the actual output value at the moment, y p (t k+j ) is t k+j The output value of the moment prediction model, ε is the compensation coefficient.
[0156] Further, according to the PID controller and the model predictive controller, the PID-based model predictive controller is obtained as follows:
[0157] U(t k )=u(t k )+Δu(t k );
[0158] Among them, U(t k ) is the final control input, u(t k) is the control input obtained by the PID controller, Δu(t k ) is the control input change obtained by the model predictive controller.
[0159] Furthermore, according to the model predictive controller, an event triggering strategy is introduced as follows:
[0160] Fixed threshold trigger:
[0161]
[0162] in, is the set threshold;
[0163] Fixed time triggers are:
[0164] t k -t ke ≥t kmax ;
[0165] Among them, t ke is the last triggering time, t kmax The maximum time limit allowed.
[0166] Furthermore, according to the PID-based model predictive controller, an event triggering strategy is introduced to obtain a PID-based event triggered model predictive controller, that is, the model predictive controller performs calculations only when a predetermined triggering condition is met.
[0167] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.
Claims
1. A method for controlling the oxygen content of flue gas during the incineration of municipal solid waste. It is characterized in that The following steps are involved: S1. Establish a prediction model for flue gas oxygen content in the process of municipal solid waste incineration based on a random configuration network; S2. adjusting the parameters of the prediction model online through an adaptive update strategy; S3. Designing a PID-based model predictive controller according to the prediction model; S4. According to the PID-based model predictive controller, an event triggering strategy is introduced to obtain a PID-based event triggering model predictive controller.
2. A method for controlling the oxygen content of flue gas during the incineration of municipal solid waste according to claim 1, It is characterized in that A prediction model for flue gas oxygen content in the process of municipal solid waste incineration is established based on the random configuration network. The prediction model is expressed as: y p =f(u,y); Where u is the input sequence of the random configuration network, y is the actual output value, and y p is the output value of the random configuration network, f(u,y) is the prediction model function, and when the random configuration network has L-1 nodes, f(u,y) is expressed as: Among them, L-1 is the number of nodes, β j is the output weight of the jth node in the hidden layer, g j (*) is the activation function of the jth node in the hidden layer, ω j is the input weight of the jth node in the hidden layer, b j is the bias of the jth node in the hidden layer.
3. A method for controlling the oxygen content of flue gas during the incineration of urban solid waste according to claim 2, It is characterized in that The residual of the prediction model is defined as: and L-1 =yy p ; when ||e L-1 ||>e max , then add a new node as follows: Among them, e max is the maximum allowable residual, h L is an intermediate variable, r is the randomly configured network parameter, and 0 <r<1,μ L is a sequence of non-negative real numbers that satisfies μ L <1-r,g L (*) is the activation function of the Lth node in the hidden layer, ω L is the input weight of the Lth node in the hidden layer, b L is the bias of the Lth node in the hidden layer, N represents the number of elements in the input sequence, and u 1 is the first value of the input sequence, u 2 is the second value of the input sequence, u N is the Nth value of the input sequence; The random configuration network will L The candidate node corresponding to the maximum value is added to the hidden layer as the optimal node, and the hidden layer output weight is recalculated. The hidden layer output weight calculation formula is: Among them, β 1 is the output weight of the first node in the hidden layer, β 2 is the output weight of the second node in the hidden layer, β L is the output weight of the Lth node in the hidden layer.
4. The method for controlling the oxygen content of flue gas during the incineration of municipal solid waste according to claim 1, It is characterized in that The parameters of the prediction model are adjusted online through an adaptive update strategy. The cost function of the adaptive update of the parameters of the prediction model is defined as follows: Δθ(t k )=θ(t k )-θ(t k-1 ); Among them, t k is the current sampling time, L(t k ) is the cost function, y p (t k ) is t k The output value of the prediction model at time y(t k ) is t k The actual value at the moment, γ is the weight coefficient, i is the intermediate variable, n θ is the number of parameters of the prediction model, Δθ(t k ) is an intermediate variable, x(t k ) is t k The input value of the random configuration network at time θ(t k ) is t k The parameter value of the prediction model at time θ(t k-1 ) is t k-1 The parameter values of the prediction model at the time; The optimization algorithm is used to update the parameters of the prediction model to obtain t k The cost function L(t k ) is the smallest parameter of the prediction model.
5. The method for controlling the oxygen content of flue gas during the incineration of municipal solid waste according to claim 1, It is characterized in that The PID controller is designed as follows: e(t k )=y r (t k )-y(t k ); Among them, u(t k ) is t k The output value of the PID controller at time e(t k ) is t k The error at the moment, e(t k-1 ) is t k-1 The error of time, K p , K d and K i is the PID controller parameter, j is the intermediate variable, k is the current sampling number, and T is the predictive control time domain.
6. A method for controlling the oxygen content of flue gas during the incineration of municipal solid waste according to claim 4, It is characterized in that According to the prediction model, the model prediction controller is designed as follows: Δu(t k+j )=u(t k+j )-u(t k+j-1 ); Among them, J(Δu,t k ) is the cost function, λ is the control parameter, t k is the current sampling time, j is the intermediate variable, N p is the number of prediction times, N u is the prediction control time domain number, y r (t k+j ) is t k+j Expected value at any time, t k+j The predicted value at the moment, Δu(t k+j ) is t k+j The change value of the control quantity is predicted at each moment, u(t k+j ) is t k+j The predicted control quantity at each moment, u(t k+j-1 ) is t k+j-1 Predict control quantity at all times; y r (t k+j ) is calculated as follows: Among them, α is the adjustment factor, 0<α<1, y r (t k ) is t k Expected value at time, y r (t k+j-1 ) is t k+j-1 Expected value at time, y sp (t k ) is the reference setting value; The calculation formula is as follows: Among them, e 1 (t k ) is t k The error between the output value of the prediction model and the actual output value at the moment, y p (t k+j ) is t k+j The output value of the moment prediction model, ε is the compensation coefficient.
7. A method for controlling the oxygen content of flue gas during the incineration of municipal solid waste according to claim 6, It is characterized in that According to the PID controller and the model predictive controller, the PID-based model predictive controller is obtained as follows: U(t k )=u(t k )+Δu(t k ); Among them, U(t k ) is the final control input, u(t k ) is the control input obtained by the PID controller, Δu(t k ) is the control input change obtained by the model predictive controller.
8. A method for controlling the oxygen content of flue gas during the incineration of municipal solid waste according to claim 7, It is characterized in that According to the model predictive controller, the event triggering strategy is introduced as follows: Fixed threshold trigger: |y sp (t k )-y(t k )|≥κ; Among them, κ is the set threshold; Fixed time triggers are: t k -t ke ≥t kmax ; Among them, t ke is the last triggering time, t kmax is the maximum time limit allowed.
9. A method for controlling the oxygen content of flue gas during the incineration of municipal solid waste according to claim 8, It is characterized in that According to the PID-based model predictive controller, an event triggering strategy is introduced to obtain a PID-based event triggered model predictive controller, that is, the model predictive controller performs calculation only when a predetermined triggering condition is met.
Citation Information
Patent Citations
Boiler combustion system control method based on multivariable PID-PFC
CN105700357A
Plasma incinerator temperature control system and method based on model predictive control
CN117270604A
Air and coal adjusting method based on deep peak regulation working condition
CN117803945A
Real-time control method and system for flue gas oxygen content in solid waste incineration process
CN117847538A
Incinerator control method
JP1994337705A
