A digital twin system for a waste incineration simulation model

Abstract

The present application relates to incineration control technical field, disclose a kind of digital twinborn system for garbage incineration simulation model, comprising: data input unit, for real-time continuous detection into the real-time working condition information of incinerator;Data regulation and control unit, for MPC predictive control analysis, generate the control instruction of different working conditions, and determine the regulation and control information parameters based on the control instruction of different working conditions, the regulation and control information includes feed quantity and air volume regulation instruction and environmental disturbance factor;Twin model construction unit, for based on Real2Sim2Real two-way interactive interface, DCS historical data is used as training sample to open simulation migration channel to generate garbage incineration process dynamic simulation model, as digital twinborn model;Result output unit, for the real-time working condition information and the regulation and control information parameters under corresponding working condition are input digital twinborn model, simulation output the combustion condition parameter of incinerator in prediction time period, and generate global control parameter based on combustion condition parameter.

Description

Technical Field

[0001] This invention relates to the field of incineration control technology, and more specifically to a digital twin system for waste incineration simulation models. Background Technology

[0002] Traditional waste incinerators rely heavily on the personal experience of DCS operators. Due to the complex and variable composition of waste (large fluctuations in calorific value), manual operation suffers from strong subjectivity and slow response, resulting in large fluctuations in combustion conditions and difficulty in maintaining stable operation. Furthermore, the high workload of operators can easily lead to fatigue and errors, affecting actual operating efficiency. Therefore, achieving automatic control and simulation optimization of the waste incineration process is a key challenge of existing technologies. Some existing enterprises have attempted to introduce intelligent control systems based on shallow learning algorithms such as random forests. However, these systems still suffer from significant deficiencies in generalization ability and control accuracy when facing complex combustion conditions, making it difficult to meet practical needs. Traditional intelligent control technologies, such as image recognition, are also limited by harsh environments such as high temperatures and dust, hindering their effectiveness. Therefore, this paper proposes a framework architecture that deeply integrates and coordinates Model Predictive Control (MPC) and reinforcement learning to ensure improved operational adaptability of the digital twin system of the waste incineration simulation model.

[0003] In reality, existing shallow algorithms such as random forests and backpropagation neural networks are difficult to accurately describe the strong coupling and nonlinear relationships in the incineration process. When encountering extreme conditions such as sudden changes in waste composition, the model prediction accuracy drops sharply, which can easily lead to control failure. Furthermore, existing reinforcement learning agents cannot be directly trained on real incinerators, and repeated trial and error can lead to a sharp increase in emergency costs. Summary of the Invention

[0004] The purpose of this invention is to provide a digital twin system for waste incineration simulation models, addressing the following technical problems: How to improve the coupling, optimization, and closed-loop control process of digital twin systems oriented towards waste incineration simulation models in waste incineration control, and improve the multi-objective optimization process of global feed information.

[0005] The objective of this invention can be achieved through the following technical solutions: A digital twin system for waste incineration simulation models includes: The data input unit is used to continuously monitor the real-time operating information entering the incinerator, including waste characteristic parameters and unit operating parameters. The data control unit is used to perform MPC predictive control analysis based on waste characteristic parameters and unit operating parameters, generate control commands for different operating conditions, and determine control information parameters based on the control commands for different operating conditions. The control information includes feed rate and air volume adjustment commands as well as environmental interference factors. The twin model building unit is used to generate a dynamic simulation model of the waste incineration process based on the Real2Sim2Real bidirectional interactive interface, using historical DCS data as training samples to open the simulation migration channel, thus serving as a digital twin model. The result output unit is used to input real-time operating condition information and corresponding control information parameters into the digital twin model, simulate and output the combustion condition parameters of the incinerator for the predicted time period, and generate global control parameters based on the combustion condition parameters.

[0006] Preferably, the waste characteristic parameters collected by the data input unit include waste composition and calorific value, the unit operating parameters include grate speed, pusher cycle, pressure of each air chamber, opening degree of primary and secondary air, multi-point temperature of furnace, oxygen content of flue gas, main steam pressure and flow rate, and environmental interference factors include ambient temperature and humidity and equipment aging status parameters.

[0007] Preferably, the data control unit constructs MPC predictive control analysis in the following way: Construct a dynamic mathematical model to generate state-space equations:

[0008] in, For the current discrete time step, Let be the system state vector. To control the input, Let be the perturbation vector. For the output vector, , , , These are the vector system matrices identified based on the digital twin model; MPC calculates the output for future time steps using a prediction model. The control coefficients are represented by the prediction matrix and the initial state offset. The future prediction step size is obtained through the prediction equation. The predicted values ​​are output in the following way:

[0009] in, , indicating the future The output predicted sequence of the step, The state prediction matrix is ​​obtained through the matrix. and It is deduced that, It is a dynamic matrix, whose elements are determined by the matrix. and The decision indicates the strength of the influence of the control input on the output. , indicating the control increment sequence, To control the time domain, Represents a dynamic matrix. The perturbation prediction matrix is... For the perturbation prediction sequence, The basic bias vector represents the natural state evolution under the condition of no control increment.

[0010] Preferably, the data control unit further calculates the optimal control sequence by solving a quadratic programming problem, specifically: Generate a quadratic optimization objective function:

[0011] in, The reference trajectory is the desired combustion parameters. This is the output error weight matrix, used to adjust the priority of different objectives such as temperature and efficiency; The incremental weight matrix is ​​used to penalize drastic control actions. Construct hard constraints:

[0012] Solving the quadratic programming problem with the above hard constraints yields the optimal control increment sequence. And a rolling optimization strategy is used to calculate control commands:

[0013] in, To optimize the current time-independent control increment obtained from the solution, These are the control commands for different operating conditions that will be output in the final product.

[0014] Preferably, the calculation method for determining the control information parameters based on control commands under different operating conditions is as follows: The control commands output by the MPC are used as the reference control quantity, and feedforward compensation calculations are performed in combination with the real-time monitoring values ​​of environmental disturbance factors. The strong coupling relationship between the feed rate and the air volume is eliminated by a preset decoupling algorithm, and the final control information parameters are calculated. The control information parameters include the decoupled feeder speed setpoint, the grate segment speed setpoint, and the primary and secondary air volume distribution ratio instruction.

[0015] Preferably, the feedforward compensation calculation method is as follows:

[0016] in, For feedforward compensation control increment, The feedforward gain coefficients were obtained offline through digital twin model identification. Real-time monitoring values ​​of environmental disturbance factors; Increment of feedforward compensation control Superimposed on the baseline control increment of the MPC output:

[0017] That is, the control increment is transformed from the control coefficients obtained by solving quadratic programming.

[0018] Preferably, a preset decoupling algorithm is used to eliminate the strong coupling relationship between the feed rate and the air volume, specifically: A feedforward decoupling strategy is adopted, and the decoupling matrix is ​​set as follows:

[0019] After decoupling network processing, the actual control signal entering the actuator is calculated as follows:

[0020] Simplified to discrete control, the decoupled control information parameters are: Feeder speed setpoint :

[0021] Primary and secondary air volume distribution instructions :

[0022] in, For the decoupling strength coefficient of feed to air volume, The air volume decoupling strength coefficient for combustion. , The tracking errors are respectively for the feed circuit and the air volume circuit. , These are the reference control values ​​for the feed circuit and the air volume circuit after passing through the feedforward compensation field, respectively.

[0023] Preferably, the method by which the digital twin model of the result output unit simulates and calculates the combustion condition parameters is as follows: Receive real-time operating condition information and control information parameters as input vectors and input them into the constructed digital twin model; Based on a hybrid modeling method that combines mechanistic constraints and data-driven approaches, the simulation outputs the furnace temperature field distribution sequence, main steam flow prediction value, and flue gas pollutant emission trend within the predicted time period through time-series recursive calculation. The simulation process supports multi-step rolling prediction, and the prediction step size is aligned with the MPC prediction step size.

[0024] Preferably, the global control parameters are calculated as follows: Based on the combustion parameters output from the simulation, a multi-objective optimization function is constructed that includes constraints on maximizing combustion efficiency, minimizing pollutant emissions, and operational stability. A normalization process based on the weight coefficient method is adopted to transform the multi-objective optimization problem into a single-objective optimization problem; Solve for the optimal solution that satisfies all constraints, generate global control parameters including global feed rate, total air volume distribution and overall grate operating speed, and output them to the control execution layer.

[0025] The beneficial effects of this invention are: (1) This invention realizes the transformation from the shallow level of multi-source data perception to the execution of global control parameters, realizes the replacement of the traditional decision-making mode that relies on human experience, and significantly reduces the workload of DCS operators; it also ensures that the system can always make decisions based on the latest operating conditions by collecting waste characteristics and unit operating parameters in real time, thereby solving the problem that the traditional offline simulation model is lagging behind actual production.

[0026] (2) This invention combines real-time MPC predictive control with feedforward compensation and decoupling algorithm, so that by using state space equation and quadratic programming to solve, the optimal control command is generated under the hard constraints of strictly meeting certain furnace temperature and other conditions, and the main steam flow fluctuation is reduced to the expected index. Furthermore, the feedforward compensation calculation is used to offset the influence of interference factors such as waste calorific value fluctuation and environmental temperature and humidity changes in advance, avoiding the lag defect of traditional feedback control after the fact. In addition, by using a preset decoupling matrix and feedforward strategy, the strong nonlinear coupling problem between feed rate and air volume is successfully solved, so that adjusting the feed rate no longer causes drastic fluctuations in air volume, and adjusting the air volume no longer affects the stability of the material layer, ensuring the smooth operation of the air volume reduction process.

[0027] (3) This invention achieves the integration of heat and mass transfer mechanisms with multi-source real-time data by constructing a digital twin model driven by both mechanism and data and a Real2Sim2Real interface. Combined with extended Kalman filter (EKF) for online identification, the model error is strictly controlled within a small range. Compared with traditional shallow learning algorithms, the prediction accuracy no longer drops sharply under extreme conditions such as sudden changes in waste composition. At the same time, through the Real2Sim2Real bidirectional interactive interface, historical data of DCS is transformed into virtual training samples, realizing the "virtual pre-training" of reinforcement learning agents. Simultaneously, the simulation optimization strategy is seamlessly transferred to the real system, completely avoiding the risks of equipment damage and pollutant exceedance caused by direct debugging on the real incinerator, and the cost of a single debugging session is drastically reduced. Furthermore, this system also supports online parameter updates, which can better adapt to the aging changes of various equipment and extend the effective cycle of the control system.

[0028] (4) Based on the furnace temperature field, steam flow and pollutant emission trend output by simulation, this invention constructs a multi-objective function including "maximizing combustion efficiency, minimizing pollutant emissions and operational stability". Through the weighting coefficient method and normalization processing, the multi-objective conflict problem is scientifically solved, the comprehensive benefits are optimized, and the system is guaranteed to continuously predict future operating conditions and dynamically adjust global control parameters by supporting multi-step rolling prediction aligned with the MPC prediction step size, thus ensuring the long-term stable compliance of pollutant emissions.

[0029] Of course, any product implementing this invention does not necessarily need to achieve all the advantages described above at the same time. Attached Figure Description

[0030] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1 This is a unit diagram of a digital twin system for waste incineration simulation models according to the present invention; Figure 2 This is a flowchart illustrating the calculation method for determining control information parameters based on control commands under different operating conditions according to the present invention. Figure 3 This is a flowchart illustrating the steps involved in simulating and calculating combustion parameters using the digital twin model of this invention. Figure 4 This is a diagram illustrating the steps involved in calculating the global control parameters of this invention. Detailed Implementation

[0032] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] Please see Figure 1 As shown, this invention is a digital twin system for waste incineration simulation models, comprising: The data input unit is used to continuously monitor the real-time operating information entering the incinerator, including waste characteristic parameters and unit operating parameters. The data control unit is used to perform MPC predictive control analysis based on waste characteristic parameters and unit operating parameters, generate control commands for different operating conditions, and determine control information parameters based on the control commands for different operating conditions. The control information includes feed rate and air volume adjustment commands as well as environmental interference factors. The twin model building unit is used to generate a dynamic simulation model of the waste incineration process based on the Real2Sim2Real bidirectional interactive interface, using historical DCS data as training samples to open the simulation migration channel, thus serving as a digital twin model. The result output unit is used to input real-time operating condition information and corresponding control information parameters into the digital twin model, simulate and output the combustion condition parameters of the incinerator for the predicted time period, and generate global control parameters based on the combustion condition parameters.

[0034] In the above technical solution, a digital twin system for waste incineration simulation models is constructed by sequentially connecting a data input unit, a data control unit, a twin model construction unit, and a result output unit. The system ensures that the control system always makes decisions based on the latest operating conditions by collecting waste characteristics and unit operating parameters in real time, thus solving the problem of traditional offline simulation models lagging behind actual production. The system's workflow is as follows: The data input unit continuously collects real-time operating information at a frequency of seconds through a sensor network (such as thermocouples, pressure transmitters, oxygen analyzers, etc.) installed at the incinerator site and the DCS system interface. The real-time operating information includes at least: waste characteristic parameters (such as waste composition and calorific value detected by near-infrared spectrometer), unit operating parameters (including grate speed, pusher cycle, pressure of each air chamber, primary / secondary air opening, furnace multi-point temperature, flue gas oxygen content, main steam pressure and flow rate), and in practice, environmental interference factors (such as external temperature and humidity obtained from weather stations, and equipment aging status parameters estimated based on equipment operating time). In this embodiment, only waste characteristic parameters and unit operating parameters are considered as the data basis for subsequent Fangzheng control. The data control unit implements the control decision-making process. It receives the aforementioned multi-source data and first constructs a dynamic mathematical model of the waste incineration process based on a digital twin model. Then, it performs MPC predictive control analysis to generate control commands for different operating conditions (such as cold start, full-load operation, and sudden changes in waste calorific value). Subsequently, the system performs feedforward compensation and decoupling calculations on these control commands to determine the final control information parameters. The twin model building unit completes the twin simulation stage. Specifically, through the Real2Sim2Real bidirectional interactive interface, it retrieves massive amounts of operational data from the DCS historical database as training samples to build and continuously update the dynamic simulation model of the waste incineration process, which serves as a digital twin model for further prediction and optimization. The specific prediction and optimization process is completed through the result output unit. The result output unit inputs real-time operating condition information and control information parameters into the digital twin model, simulates and calculates and outputs combustion operating condition parameters for the future prediction period. Based on these combustion operating condition parameters, the system further performs multi-objective optimization calculations, generates global control parameters and sends them to the field control execution layer (such as PLC / DCS), thereby completing a complete "perception-decision-simulation-execution" closed-loop control process.

[0035] As one embodiment of the present invention, the waste characteristic parameters collected by the data input unit include waste composition and calorific value, the unit operating parameters include grate speed, pusher cycle, pressure of each air chamber, opening degree of primary and secondary air, multi-point temperature of furnace, oxygen content of flue gas, pressure and flow rate of main steam, and environmental interference factors include ambient temperature and humidity and equipment aging status parameters.

[0036] As one embodiment of the present invention, the data control unit constructs MPC predictive control analysis in the following way: Construct a dynamic mathematical model to generate state-space equations:

[0037] in, For the current discrete time step, Let be the system state vector. To control the input, Let be the perturbation vector. For the output vector, , , , These are the vector system matrices identified based on the digital twin model; MPC calculates the output for future time steps using a prediction model. The control coefficients are represented by the prediction matrix and the initial state offset. The future prediction step size is obtained through the prediction equation. The predicted values ​​are output in the following way:

[0038] in, , indicating the future The output predicted sequence of the step, The state prediction matrix is ​​obtained through the matrix. and It is deduced that, It is a dynamic matrix, whose elements are determined by the matrix. and The decision indicates the strength of the influence of the control input on the output. , indicating the control increment sequence, To control the time domain, Represents a dynamic matrix. The perturbation prediction matrix is... For the perturbation prediction sequence, The basic bias vector represents the natural state evolution under the condition of no control increment.

[0039] In the above technical solution, to achieve the combination of MPC and decoupling algorithm, a specific implementation logic process for the data control unit is provided: First, a discrete state-space equation describing the dynamic characteristics of the incinerator is constructed. The system matrix in this equation , , , It is obtained offline based on the digital twin model through subspace identification methods (such as the N4SID algorithm), and the state vector Implicit factors include material layer thickness, combustion temperature field distribution, etc., and control inputs. Including feeder speed, air volume, valve opening; disturbance vector This indicates the fluctuation of the calorific value of waste, and the output vector is... The main steam flow rate and furnace temperature are measured; then, based on the discrete state-space equation model, the MPC controller calculates the output prediction sequence for future time lengths using the prediction equation, and determines the prediction step size. This indicates that, typically set to 30 steps, the output predicted sequence is: According to the predicted sequence The calculation formula provides the state prediction matrix. Dynamic matrix Control Increment Sequence The perturbation prediction term includes the perturbation prediction matrix. and perturbation prediction sequence Through the predicted bias vector Describe the natural state evolution under uncontrolled increments.

[0040] As one embodiment of the present invention, the data control unit further calculates the optimal control sequence by solving a quadratic programming problem, specifically: Generate a quadratic optimization objective function:

[0041] in, The reference trajectory is the desired combustion parameters. This is the output error weight matrix, used to adjust the priority of different objectives such as temperature and efficiency; The incremental weight matrix is ​​used to penalize drastic control actions. Construct hard constraints:

[0042] Solving the quadratic programming problem with the above hard constraints yields the optimal control increment sequence. And a rolling optimization strategy is used to calculate control commands:

[0043] in, To optimize the current time-independent control increment obtained from the solution, These are the control commands for different operating conditions that will be output in the final product.

[0044] In the above technical solution, in order to generate the optimal control command, this system constructs a quadratic optimization objective function. By introducing hard constraints, the optimal control increment sequence can be obtained by solving the above quadratic programming (QP) problem using the interior point method or the effective set method under the defined constraints. Among them, hard constraints In the formula, " The phrase "" indicates ensuring the furnace temperature remains within a preset range; generally, the furnace temperature should not be lower than 850℃. In the formula, " "Used to limit the rate of change of grate speed; where " "Used to limit maximum and minimum airflow; by employing a rolling optimization strategy, the control increment at the current moment is applied to this system to determine..." In this formula It serves as the base control command output by the MPC.

[0045] In summary, by implementing MPC predictive control combined with feedforward compensation and decoupling algorithms, the optimal control command is generated by solving the state-space equation and quadratic programming under strict constraints such as furnace temperature and equipment physical limits, which significantly reduces the fluctuation of main steam flow.

[0046] Please see Figure 2 As shown, in one embodiment of the present invention, the calculation method for determining the control information parameters according to different operating condition control commands is as follows: The control commands output by the MPC are used as the reference control quantity, and feedforward compensation calculations are performed in combination with the real-time monitoring values ​​of environmental disturbance factors. The strong coupling relationship between the feed rate and the air volume is eliminated by a preset decoupling algorithm, and the final control information parameters are calculated. The control information parameters include the decoupled feeder speed setpoint, the grate segment speed setpoint, and the primary and secondary air volume distribution ratio instruction.

[0047] In the above technical solution, the control information parameters corresponding to the control commands under different operating conditions are further determined. For example, in the rainy season, the garbage has high moisture content and low calorific value. In order to overcome environmental interference and internal coupling, the above-mentioned baseline control commands are further processed. The processing adopts feedforward compensation calculation and decoupling calculation: by using the control commands output by MPC as the baseline control quantity, feedforward compensation calculation is performed in combination with the real-time monitoring value of environmental interference factors; and by using a preset decoupling algorithm to eliminate the strong coupling relationship between the feed rate and the air volume, the final control information parameters are calculated. The control information parameters include the decoupled feeder speed setpoint, grate segment speed setpoint, and primary and secondary air volume distribution ratio command.

[0048] As one embodiment of the present invention, the feedforward compensation calculation method is as follows:

[0049] in, For feedforward compensation control increment, The feedforward gain coefficients were obtained offline through digital twin model identification. Real-time monitoring values ​​of environmental disturbance factors; Increment of feedforward compensation control Superimposed on the baseline control increment of the MPC output:

[0050] That is, the control increment is transformed from the control coefficients obtained by solving quadratic programming.

[0051] In the above technical solution, feedforward compensation calculation is used. It offsets the effects of interference factors such as fluctuations in waste calorific value and changes in ambient temperature and humidity, utilizing the feedforward gain coefficient. The step response was obtained through offline testing using a digital twin model. Subsequently, the compensation amount was superimposed on the MPC output. This avoids the lag in "post-event remedies" inherent in traditional feedback control.

[0052] As one embodiment of the present invention, a preset decoupling algorithm is used to eliminate the strong coupling relationship between the feed rate and the air volume, specifically as follows: A feedforward decoupling strategy is adopted, and the decoupling matrix is ​​set as follows:

[0053] After decoupling network processing, the actual control signal entering the actuator is calculated as follows:

[0054] Simplified to discrete control, the decoupled control information parameters are: Feeder speed setpoint :

[0055] Primary and secondary air volume distribution instructions :

[0056] in, For the decoupling strength coefficient of feed to air volume, The air volume decoupling strength coefficient for combustion. , The tracking errors are respectively for the feed circuit and the air volume circuit. , These are the reference control values ​​for the feed circuit and the air volume circuit after passing through the feedforward compensation field, respectively.

[0057] In the above technical solution, considering that changes in feed rate directly affect the oxygen consumption in the furnace, leading to changes in air volume demand, and vice versa, the system adopts a feedforward decoupling strategy and sets a decoupling matrix: Furthermore, the calculation of decoupled control information parameters, including the feeder speed setpoint, is implemented in discrete control. Primary / Secondary Air Volume Distribution Instructions Based on the above calculations, the system finally outputs the decoupled feeder speed setpoint, grate segment speed setpoint, and primary and secondary air volume distribution ratio instructions. In summary, by using a pre-set decoupling matrix and a feedforward decoupling strategy, the strong nonlinear coupling problem between "feed rate and air volume" was successfully solved, so that adjusting the feed rate no longer causes drastic fluctuations in air volume, and adjusting the air volume no longer affects the stability of the material bed, thus ensuring the smooth operation of the incineration process.

[0058] In addition, it should be noted that the construction process of the digital twin model in the twin model construction unit adopts a hybrid modeling method of "mechanism constraint + data-driven": Model structure: Construct a hybrid sub-model library of "linear + nonlinear" models; call the linear sub-model under steady-state or small fluctuation conditions, and activate the nonlinear sub-model under large disturbance conditions such as sudden changes in garbage composition; Online parameter identification: The extended Kalman filter (EKF) algorithm is used to update the model parameters in real time; the EKF algorithm dynamically adjusts the state vector based on the latest real-time data (such as the deviation between the measured furnace temperature and the model prediction). The estimated values ​​are obtained, and sensor noise is filtered out through multi-source data fusion technology to ensure that the model output closely follows the changes in actual working conditions; Real2Sim2Real interface implementation: Real2Sim: Develops an OPC UA / MTConnect interface to read historical and real-time data streams from the field DCS system. After data cleaning and timing alignment, the data is mapped to the input parameters of the simulation model, completing the initialization and training of the simulation model.

[0059] Sim2Real encapsulates the control strategy (such as the optimal grate speed curve) obtained from simulation optimization into standardized Modbus TCP or Profinet command packets, which are then transmitted to the field DCS via the network, realizing the migration of virtual strategies to the physical world. Through the above mechanism, this system strictly controls the simulation model error to a low range, such as 5%.

[0060] Please see Figure 3 As shown, in one embodiment of the present invention, the method by which the digital twin model of the result output unit simulates and calculates the combustion condition parameters is as follows: Receive real-time operating condition information and control information parameters as input vectors and input them into the constructed digital twin model; Based on a hybrid modeling method that combines mechanistic constraints and data-driven approaches, the simulation outputs the furnace temperature field distribution sequence, main steam flow prediction value, and flue gas pollutant emission trend within the predicted time period through time-series recursive calculation. The simulation process supports multi-step rolling prediction, and the prediction step size is aligned with the MPC prediction step size.

[0061] In the above technical solution, the result output unit receives real-time operating condition information (current state) from the data input unit and control information parameters (control actions) from the data control unit. These data are then injected into the digital twin model as initial and boundary conditions.

[0062] Based on the laws of mass conservation, energy conservation, and chemical reaction kinetics, the system employs the fourth-order Runge-Kutta method for time-series recursive calculations to simulate and extrapolate future events. The system status within a time step includes: furnace temperature field distribution sequence: three-dimensional temperature cloud map data, used to determine whether there is a local low temperature zone (risk of dioxin formation); main steam flow prediction value: used to assess power generation efficiency; flue gas pollutant emission trend: including predicted curves of NOx, SOx and dioxin equivalent concentrations.

[0063] Furthermore, the simulation process supports multi-step rolling prediction, meaning that every time a new real-time data point is obtained, a full-time prediction is restarted, and its prediction step size is strictly aligned with the MPC prediction step size to ensure data consistency.

[0064] Please see Figure 4 As shown, in one embodiment of the present invention, the global control parameters are calculated as follows: Based on the combustion parameters output from the simulation, a multi-objective optimization function is constructed that includes constraints on maximizing combustion efficiency, minimizing pollutant emissions, and operational stability. A normalization process based on the weight coefficient method is adopted to transform the multi-objective optimization problem into a single-objective optimization problem; Solve for the optimal solution that satisfies all constraints, generate global control parameters including global feed rate, total air volume distribution and overall grate operating speed, and output them to the control execution layer.

[0065] In the above technical solution, based on the multi-dimensional operating parameters output by the simulation, the system performs multi-objective optimization: Constructing the objective function: Define an objective function with three dimensions:

[0066] in, To synthesize the objective function value, The weight representing combustion efficiency, As a weight for pollutant emissions, The weights are for operational stability; and satisfy the following conditions: ; For the combustion efficiency evaluation function, Main steam mass flow rate, For pollutant emission evaluation functions, For pollutant emission concentration, To run the stability evaluation function, This indicates the rate of change of furnace temperature; Next, normalization was performed: due to efficiency (percentage), emissions (mg / m³) 3 Due to the different dimensions, the system uses the range method to normalize each objective:

[0067] in, For the first The normalized objective function value For the first One original target value, For the first The minimum allowable value for a target represents the ideal lower limit of that indicator. For efficiency, It may be the lowest acceptable efficiency and the lowest emission concentration; For the first The maximum allowable value for a target is usually derived from a digital twin model or historical data statistics, representing the ideal upper limit of the indicator; Finding the optimal solution: Transform the normalized multi-objective function into a single-objective optimization problem. Within the constraint space defined by the digital twin model (such as temperature boundaries and device physical limits), the optimal solution is obtained by calling the genetic algorithm (GA) or the particle swarm optimization algorithm (PSO). Output control parameters: The optimal solution obtained is converted into global control parameters, including global feed rate (t / h) and total air volume distribution (Nm³). 3 The parameters include the grate speed (m / min) and the overall grate operating speed. These parameters are sent to the control and execution layer to guide the precise operation of the equipment at the next level.

[0068] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, devices, and non-volatile computer storage media are basically similar to the method embodiments, and therefore described more simply; relevant parts can be referred to the descriptions of the method embodiments.

[0069] The above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined in this application, they should all fall within the protection scope of the present invention.

Claims

1. A digital twin system for waste incineration simulation models, characterized in that, include: The data input unit is used to continuously detect real-time operating information entering the incinerator, including waste characteristic parameters and unit operating parameters. The data control unit is used to perform MPC predictive control analysis based on waste characteristic parameters and unit operating parameters, generate control commands for different operating conditions, and determine control information parameters based on the control commands for different operating conditions. The control information includes feed rate and air volume adjustment commands as well as environmental interference factors. The twin model building unit is used to generate a dynamic simulation model of the waste incineration process based on the Real2Sim2Real bidirectional interactive interface, using historical DCS data as training samples to open the simulation migration channel, thus serving as a digital twin model. The result output unit is used to input real-time operating condition information and corresponding control information parameters into the digital twin model, simulate and output the combustion condition parameters of the incinerator for the predicted time period, and generate global control parameters based on the combustion condition parameters.

2. The digital twin system for waste incineration simulation models according to claim 1, characterized in that, The data input unit collects waste characteristic parameters including waste composition and calorific value. The unit operating parameters include grate speed, pusher cycle, pressure of each air chamber, opening degree of primary and secondary air, multi-point temperature of furnace, oxygen content of flue gas, pressure and flow rate of main steam. The environmental interference factors include ambient temperature and humidity and equipment aging status parameters.

3. The digital twin system for waste incineration simulation models according to claim 1, characterized in that, The data control unit constructs MPC predictive control analysis in the following way: Construct a dynamic mathematical model to generate state-space equations: in, For the current discrete time step, Let be the system state vector. To control the input, Let be the perturbation vector. For the output vector, , , , These are the vector system matrices identified based on the digital twin model; MPC calculates the output at future time steps using a prediction model. The control coefficients are represented by the prediction matrix and the initial state offset. The future prediction step size is obtained through the prediction equation. The predicted values ​​are output in the following way: in, , indicating the future The output predicted sequence of the step, The state prediction matrix is ​​obtained through the matrix. and It is deduced that, It is a dynamic matrix, whose elements are determined by the matrix. and The decision indicates the strength of the influence of the control input on the output. , indicating the control increment sequence, To control the time domain, Represents a dynamic matrix. The perturbation prediction matrix is... For the perturbation prediction sequence, The basic bias vector represents the natural state evolution under the condition of no control increment.

4. A digital twin system for waste incineration simulation models according to claim 3, characterized in that, The data control unit also calculates the optimal control sequence by solving a quadratic programming problem, specifically: Generate a quadratic optimization objective function: in, The reference trajectory is the desired combustion parameters. This is the output error weight matrix, used to adjust the priority of different objectives such as temperature and efficiency; The incremental weight matrix is ​​used to penalize drastic control actions. Construct hard constraints: Solving the quadratic programming problem with the above hard constraints yields the optimal control increment sequence. And a rolling optimization strategy is used to calculate control commands: in, To optimize the current time-independent control increment obtained from the solution, These are the control commands for different operating conditions that will be output in the final product.

5. A digital twin system for waste incineration simulation models according to claim 1, characterized in that, The calculation method for determining the control information parameters based on different operating condition control commands is as follows: The control commands output by the MPC are used as the reference control quantity, and feedforward compensation calculations are performed in combination with the real-time monitoring values ​​of environmental disturbance factors. The strong coupling relationship between the feed rate and the air volume is eliminated by a preset decoupling algorithm, and the final control information parameters are calculated. The control information parameters include the decoupled feeder speed setting value, the grate segment speed setting value, and the primary and secondary air volume distribution ratio instruction.

6. A digital twin system for waste incineration simulation models according to claim 5, characterized in that, The feedforward compensation calculation method is as follows: in, For feedforward compensation control increment, The feedforward gain coefficients were obtained offline through digital twin model identification. Real-time monitoring values ​​of environmental disturbance factors; Increment of feedforward compensation control Superimposed on the baseline control increment of the MPC output: That is, the control increment is transformed from the control coefficients obtained by solving quadratic programming.

7. A digital twin system for waste incineration simulation models according to claim 6, characterized in that, The process of eliminating the strong coupling between feed rate and air volume using a preset decoupling algorithm specifically involves: A feedforward decoupling strategy is adopted, and the decoupling matrix is ​​set as follows: After decoupling network processing, the actual control signal entering the actuator is calculated as follows: Simplified to discrete control, the decoupled control information parameters are: Feeder speed setpoint : Primary / Secondary Air Volume Distribution Instructions : in, For the decoupling strength coefficient of feed to air volume, The air volume decoupling strength coefficient for combustion. , The tracking errors are respectively for the feed circuit and the air volume circuit. , These are the reference control values ​​for the feed circuit and the air volume circuit after passing through the feedforward compensation field, respectively.

8. A digital twin system for waste incineration simulation models according to claim 1, characterized in that, The method by which the digital twin model of the result output unit simulates and calculates the combustion condition parameters is as follows: Receive real-time operating condition information and control information parameters as input vectors and input them into the constructed digital twin model; Based on a hybrid modeling method that combines mechanistic constraints and data-driven approaches, the simulation outputs the furnace temperature field distribution sequence, main steam flow prediction value, and flue gas pollutant emission trend within the predicted time period through time-series recursive calculation. The simulation process supports multi-step rolling prediction, and the prediction step size is aligned with the MPC prediction step size.

9. A digital twin system for waste incineration simulation models according to claim 8, characterized in that, The global control parameters are calculated as follows: Based on the combustion parameters output from the simulation, a multi-objective optimization function is constructed that includes constraints on maximizing combustion efficiency, minimizing pollutant emissions, and operational stability. A normalization process based on the weight coefficient method is adopted to transform the multi-objective optimization problem into a single-objective optimization problem; Solve for the optimal solution that satisfies all constraints, generate global control parameters including global feed rate, total air volume distribution and overall grate operating speed, and output them to the control execution layer.

Images