Incineration parameter recommendation method, system, device and medium based on digital twinning

Abstract

This application discloses a method, system, equipment, and medium for recommending incineration parameters based on digital twins, relating to the technical field of waste treatment and digital twins. The method includes: acquiring historical operating data from a digital twin of a waste incineration system; constructing an incineration condition rating model based on the historical incineration process characteristic parameters and historical incineration evaluation level data, and constructing a process characteristic prediction model based on the historical incinerated material characteristic data, historical incineration parameters, and historical incineration process characteristic parameters; setting process characteristic parameter constraints, calling the incineration condition rating model, and searching for a set of target process characteristic parameters using a genetic algorithm; setting incineration parameter constraints, calling the process characteristic prediction model, and searching for a set of recommended incineration parameters using a particle swarm optimization algorithm. This application achieves automated and intelligent recommendation of incineration parameters, thereby improving the precise control of incineration conditions.

Description

Technical Field

[0001] This application relates to the technical field of waste treatment and digital twins, and in particular to methods, systems, equipment and media for recommending incineration parameters based on digital twins. Background Technology

[0002] Industrial combustible waste, such as discarded work clothes, gloves, plastic products, organic waste liquids, and resins, is one of the main wastes generated during the production processes of industries such as chemical, pharmaceutical, medical, and high-end manufacturing. Incineration is widely considered one of the effective technologies for treating this type of waste due to its significant volume reduction effect and the advantage of completely inorganicizing organic matter.

[0003] However, traditional industrial waste incineration process control still faces numerous bottlenecks. Firstly, its operation heavily relies on manual experience. The incinerator is a high-temperature, highly corrosive "black box" environment, making it impossible for operators to directly observe the actual distribution of the temperature and flow fields, or assess the completeness of combustion. When waste characteristics (such as calorific value, moisture content, and morphology) change, engineers primarily rely on personal experience and limited experimental data to manually adjust key parameters such as the primary and secondary air ratio, grate speed, and combustion temperature. This process is time-consuming, labor-intensive, highly subjective, and difficult to quickly and accurately achieve optimal operating conditions, easily leading to low combustion efficiency, excessive pollutant emissions, or accelerated equipment wear and tear.

[0004] In recent years, digital twin technology has offered a new solution to the aforementioned problems. This technology achieves real-time mapping and visualization of the incineration process by constructing high-fidelity virtual models of physical entities. However, most existing digital twin-based applications remain at the level of condition monitoring and process simulation, failing to deeply address the core optimization decision-making problem. Specifically: firstly, the model is disconnected from optimization control; the digital twin fails to be deeply coupled with advanced intelligent optimization algorithms to form a closed-loop decision-making process; secondly, there is a lack of scientific methods to translate multi-dimensional, multi-objective operational requirements into specific process parameters; and thirdly, it cannot achieve automatic optimization of adaptive parameters for specific waste characteristics and operational objectives.

[0005] Therefore, there is an urgent need for a method that can deeply integrate digital twin models and intelligent optimization algorithms to achieve automated and intelligent recommendation of incineration parameters, so as to fundamentally improve the safety, economy and environmental protection of industrial waste incineration treatment. Summary of the Invention

[0006] To achieve automated and intelligent recommendation of incineration parameters and thereby improve the precise control of incineration conditions, this application provides a method, system, equipment, and medium for recommending incineration parameters based on digital twins.

[0007] Firstly, this application provides a method for recommending incineration parameters based on digital twins, employing the following technical solution:

[0008] Digital twin-based incineration parameter recommendation methods include:

[0009] Historical operating data is obtained from the digital twin of the waste incineration system, including historical incinerated material characteristic data, historical incineration parameters, historical incineration process characteristic parameters, and historical incineration evaluation level data.

[0010] Based on the historical incineration process characteristic parameters and the historical incineration evaluation level data, an incineration condition rating model is constructed, and based on the historical incinerated material characteristic data, the historical incineration parameters, and the historical incineration process characteristic parameters, a process characteristic prediction model is constructed.

[0011] Set process characteristic parameter constraints, call the incineration condition rating model, and search for a set of target process characteristic parameters according to the genetic algorithm so that the corresponding prediction level is consistent with the user's current demand level, wherein the set of target process characteristic parameters satisfies the process characteristic parameter constraints.

[0012] Set incineration parameter constraints, call the process feature prediction model, and search for a set of recommended incineration parameters using the particle swarm optimization algorithm, so that the current incinerated material characteristic data and the predicted feature parameters of the set of recommended incineration parameters are closest to the set of target process feature parameters, wherein the set of recommended incineration parameters satisfies the incineration parameter constraints.

[0013] By adopting the above technical solution, historical operational data is obtained from the digital twin of the waste incineration system. This historical operational data includes historical incinerated material characteristic data, historical incineration parameters, historical incineration process characteristic parameters, and historical incineration evaluation level data. Then, based on the historical incineration process characteristic parameters and historical incineration evaluation level data, an incineration condition rating model is constructed. Next, based on the historical incinerated material characteristic data, historical incineration parameters, and historical incineration process characteristic parameters, a process characteristic prediction model is constructed. Then, process characteristic parameter constraints are set, and the incineration condition rating model is invoked. A set of target process characteristic parameters is searched using a genetic algorithm to ensure that the corresponding prediction level matches the user's current requirement level. This set of target process characteristic parameters satisfies the process characteristic parameter constraints. Then, incineration parameter constraints are set, and the process characteristic prediction model is invoked. A set of recommended incineration parameters is searched using a particle swarm optimization algorithm to ensure that the current incinerated material characteristic data and the recommended incineration parameters match the current incinerated material characteristic data. The predicted characteristic parameters of the number are closest to a set of target process characteristic parameters, where a set of recommended incineration parameters satisfies the incineration parameter constraints. The incineration parameter recommendation method based on digital twins proposed in this invention achieves intelligent and automated incineration parameter recommendation by constructing a dual-model and combining a two-layer optimization architecture of genetic algorithm and particle swarm optimization. This method can quickly and accurately automatically generate optimal operating parameters based on the dynamic changes of waste characteristics and user-specific operational goals, completely changing the outdated mode of relying on manual experience and trial-and-error adjustments. Through digitalization and automation, it reduces the dependence on operator experience, reduces human error, thereby improving the stability and processing efficiency of the incineration process, ensuring complete combustion and minimum pollutant emissions, and enhancing the system's adaptability to complex and variable raw materials. It also improves the flexibility and robustness of the production line, effectively ensuring operational safety and economy, and providing core key technical support for the intelligent upgrading of industrial incineration equipment.

[0014] Optionally, the step of constructing an incineration condition rating model based on the historical incineration process characteristic parameters and the historical incineration evaluation level data includes:

[0015] Preprocessing the historical incineration process characteristic parameters and the historical incineration evaluation level data yields the first model training data.

[0016] The training data of the first model is divided into a first training set and a second test set according to a preset ratio;

[0017] The hyperparameters of the predefined decision tree model are set according to the random network search algorithm, and the corresponding root mean square error (RMSE) and coefficient of determination (R²) are used as the first evaluation index.

[0018] The decision tree model is trained based on the first training set to obtain a trained decision tree model;

[0019] The trained decision tree model is tested according to the first test set, and the error is judged according to the first evaluation index to determine whether it is within the preset range. If so, the trained decision tree model is used as the incineration condition rating model.

[0020] By adopting the above technical solution, in order to construct the incineration condition rating model, the characteristic parameters of historical incineration processes and historical incineration evaluation level data are preprocessed to obtain the first model training data. Then, the first model training data is divided into a first training set and a second test set according to a preset ratio. Then, the hyperparameters of the predefined decision tree model are set according to the random network search algorithm, and the corresponding root mean square error (RMSE) and coefficient of determination (R²) are used as the first evaluation index. Then, the decision tree model is trained according to the first training set to obtain a trained decision tree model. Then, the trained decision tree model is tested according to the first test set, and the error is judged according to the first evaluation index to determine whether it is within the preset range. If the error is within the preset range, the trained decision tree model is used as the incineration condition rating model.

[0021] Optionally, the step of constructing a process characteristic prediction model based on the historical incinerated material characteristic data, the historical incineration parameters, and the historical incineration process characteristic parameters includes:

[0022] The historical incinerated material characteristic data, the historical incineration parameters, and the historical incineration process characteristic parameters are preprocessed to obtain the second model training data.

[0023] The training data of the second model is divided into a second training set and a second test set according to a preset ratio;

[0024] The hyperparameters of the predefined SVM model are set according to the random network search algorithm, and the corresponding root mean square error (RMSE) and coefficient of determination (R²) are used as the second evaluation index.

[0025] The SVM model is trained using the second training set to obtain a trained SVM model.

[0026] The trained SVM model is tested according to the second test set, and the error is judged according to the second evaluation index to determine whether it is within the preset range. If it is, the trained SVM model is used as the process feature prediction model.

[0027] By adopting the above technical solution, in order to construct a process feature prediction model, historical incineration characteristic data, historical incineration parameters, and historical incineration process characteristic parameters are preprocessed to obtain second model training data. Then, the second model training data is divided into a second training set and a second test set according to a preset ratio. Then, the hyperparameters of the predefined SVM model are set according to the random network search algorithm, and the corresponding root mean square error (RMSE) and coefficient of determination (R²) are used as the second evaluation index. Then, the SVM model is trained according to the second training set to obtain a trained SVM model. Then, the trained SVM model is tested according to the second test set, and the error is judged according to the second evaluation index to determine whether it is within the preset range. If the error is within the preset range, the trained SVM model is used as the process feature prediction model.

[0028] Optionally, the step of calling the incineration condition rating model and searching for a set of target process characteristic parameters using a genetic algorithm to ensure that the corresponding predicted level matches the user's current demand level includes:

[0029] The feasible region of each process characteristic parameter is determined according to the constraints of the process characteristic parameters, and an initial population containing M chromosomes is randomly generated within the feasible region.

[0030] For each chromosome of the initial population, the chromosome is decoded into a set of process feature parameters, and the set of process feature parameters is input into the incineration condition rating model to obtain the corresponding prediction level and the probability value corresponding to the prediction level. The corresponding first current fitness value is determined according to the probability value, wherein the first current fitness value is used to represent the difference or consistency between the prediction level and the current demand level.

[0031] Based on roulette wheel selection or tournament selection, superior individuals are selected from the current population to form the parent population according to the first current fitness value of the chromosome, wherein chromosomes with higher first current fitness values ​​are more likely to be selected.

[0032] Chromosomes are randomly selected from the parent population and crossover is performed to obtain crossover chromosomes. The crossover chromosomes are then mutated to obtain mutated chromosomes.

[0033] A new population is constructed based on the crossover chromosome and the variant chromosome, and the iteration is repeated based on the new population until a first iteration termination condition is met, wherein the first iteration termination condition includes a first maximum number of iterations, convergence of the first current fitness value, or the first current fitness value has reached a first preset value.

[0034] When the iteration terminates, the target chromosome with the highest fitness in the current population is output, and the target chromosome is decoded to obtain a set of target process feature parameters.

[0035] By adopting the above technical solution, in order to search for a set of target process feature parameters, the feasible region of each process feature parameter is determined according to the constraints of the process feature parameter. An initial population containing M chromosomes is randomly generated within the feasible region. Then, for each chromosome in the initial population, the chromosome is decoded into a set of process feature parameters, and this set of process feature parameters is input into the incineration condition rating model to obtain the corresponding prediction level and the probability value corresponding to the prediction level. The corresponding first current fitness value is determined based on the probability value, where the first current fitness value is used to represent the difference or consistency between the prediction level and the current demand level. Then, based on a roulette wheel selection or tournament selection method, the first current fitness value of the chromosome is used to determine the... Select superior individuals from the current population to form a parent population. Chromosomes with higher current fitness values ​​are more likely to be selected. Then, randomly select chromosomes from the parent population and perform crossover operations to obtain crossover chromosomes. Perform mutation operations on the crossover chromosomes to obtain mutated chromosomes. Then, construct a new population based on the crossover chromosomes and mutated chromosomes, and iterate again based on the new population until the first iteration termination condition is met. The first iteration termination condition includes the first maximum number of iterations, convergence of the first current fitness value, or the first current fitness value reaching a first preset value. When the iteration terminates, output the target chromosome with the highest fitness in the current population, and decode the target chromosome to obtain a set of target process feature parameters.

[0036] Optionally, the step of calling the process feature prediction model and searching for a set of recommended incineration parameters using the particle swarm optimization algorithm, so that the predicted feature parameters of the current incinerated material characteristic data and the set of recommended incineration parameters are closest to the set of target process feature parameters, includes:

[0037] The search space boundary for particles is set according to the incineration parameter constraints, and a group of particles is randomly initialized within the corresponding search space. Each particle represents a set of candidate combinations of incineration parameters, and each particle is assigned a random initial position x. 0 and initial velocity v 0 ;

[0038] Update formula based on location and speed update formula

[0039] Iterate for each particle separately, where x k+1 For the (k+1)th iteration point, x k For the k-th iteration point, v Kv represents the velocity in the k-th iteration. k+1 denoted as , where is the velocity in the (k+1)th iteration, w is the inertia weight, c1 is the cognitive factor, r1 is a random number between 0 and 1, pBest is the best position found during the search process, c2 is the social learning factor, r2 is a random number between 0 and 1, gBest is the best position found globally, rand is the random number function, and δ is the perturbation amplitude.

[0040] During the iteration process, for each particle in each iteration, the current position x of the current particle is... k The current incinerator characteristics data are input into the process feature prediction model to predict intermediate process characteristic parameters. ;

[0041] Calculate the intermediate process characteristic parameters and the set of target process characteristic parameters y target The corresponding second current fitness value F(x) k ), and based on the second current fitness value F(x) k The individual's historical best fitness value F(pBest) and the global historical best fitness value F(gBest) are updated, wherein the second current fitness value F(x) is updated. k The mean square error or Euclidean distance is negative.

[0042] Determine whether the second iteration termination condition is met. If so, the combination of incineration parameters represented by gBest is used as a set of recommended incineration parameters. The second iteration termination condition includes the second maximum number of iterations, the convergence of the global historical optimal fitness value F(gBest) corresponding to the global historical optimal solution gBest, or the global historical optimal fitness value F(gBest) has reached the second preset value.

[0043] By adopting the above technical solution, in order to search for a set of recommended incineration parameters, the search space boundary of the particles is set according to the constraints of the incineration parameters, and a group of particles is randomly initialized in the corresponding search space. Each particle represents a set of candidate combinations of incineration parameters, and each particle is assigned a random initial position x. 0 and initial velocity v 0 Then update the formula based on the position. and speed update formula

[0044] Iterate for each particle separately, where x k+1 For the (k+1)th iteration point, x k For the k-th iteration point, v K v represents the velocity in the k-th iteration. k+1Let x be the velocity in the (k+1)th iteration, w be the inertia weight, c1 be the cognitive factor, r1 be a random number between 0 and 1, pBest be the best position found during the search process, c2 be the social learning factor, r2 be a random number between 0 and 1, gBest be the best position found globally, rand be the random number function, and δ be the perturbation amplitude. Then, during the iteration process, for each particle in each iteration, the current position x of the current particle is... k The current incinerator characteristics data are input into the process characteristic prediction model to predict intermediate process characteristic parameters. Then calculate the intermediate process characteristic parameters. and a set of target process characteristic parameters y target The corresponding second current fitness value F(x) k ), and based on the second current fitness value F(x) k The individual's historical best fitness value F(pBest) and the global historical best fitness value F(gBest) are updated, where the second current fitness value F(x) is updated. k If the mean square error or Euclidean distance is negative, then it is determined whether the second iteration termination condition is met. If the second iteration termination condition is met, the combination of incineration parameters represented by gBest is used as a set of recommended incineration parameters. The second iteration termination condition includes the second maximum number of iterations, the convergence of the global historical optimal fitness value F(gBest) corresponding to the global historical optimal solution gBest, or the global historical optimal fitness value F(gBest) has reached the second preset value.

[0045] Optionally, the method further includes:

[0046] The target incineration process characteristic parameters are obtained, wherein the target incineration process characteristic parameters include at least one of thermal parameters, flue gas composition parameters, and combustion efficiency parameters; the thermal parameters include at least one of furnace temperature parameters, secondary combustion chamber outlet temperature, and loss on ignition; and the flue gas composition parameters include at least one of oxygen concentration, carbon monoxide concentration, nitrogen oxide concentration, and sulfur oxide concentration.

[0047] The characteristic parameters of the target incineration process are input into the incineration condition rating model to obtain the corresponding predicted operation evaluation level.

[0048] By adopting the above technical solution, in order to predict the operation evaluation level based on the incineration condition rating model, the characteristic parameters of the target incineration process are obtained. The characteristic parameters of the target incineration process include at least one of thermal parameters, flue gas composition parameters, and combustion efficiency parameters. The thermal parameters include at least one of furnace temperature parameters, secondary combustion chamber outlet temperature, and loss on ignition. The flue gas composition parameters include at least one of oxygen concentration, carbon monoxide concentration, nitrogen oxide concentration, and sulfur oxide concentration. Then, the characteristic parameters of the target incineration process are input into the incineration condition rating model to obtain the corresponding predicted operation evaluation level.

[0049] Optionally, the method further includes:

[0050] Acquire target incinerator characteristic data and target incineration parameters, wherein the target incinerator characteristic data includes at least one of physical morphology, element content, and calorific value; the target incineration parameters include at least one of feed parameters, combustion air parameters, furnace operation parameters, auxiliary fuel parameters, and exhaust gas treatment unit operation parameters; the feed parameters include feed rate; the combustion air parameters include at least one of primary air volume, secondary air volume, primary air temperature, secondary air temperature, and total combustion air volume; the furnace operation parameters include at least one of grate speed and furnace negative pressure; the auxiliary fuel parameters include at least one of auxiliary burner power and fuel addition rate; and the exhaust gas treatment unit operation parameters include at least one of quench alkali dosage, quench tower outlet target temperature, adsorbent injection rate, bag filter cleaning frequency, and differential pressure setting.

[0051] The target incinerator characteristic data and target incineration parameters are input into the process feature prediction model to obtain the corresponding predicted process feature parameters.

[0052] By adopting the above technical solution, in order to predict process characteristic parameters based on the process characteristic prediction model, target incinerator characteristic data and target combustion parameters are obtained. The target incinerator characteristic data includes at least one of physical morphology, elemental content, and calorific value. The target combustion parameters include at least one of feed parameters, combustion air parameters, furnace operating parameters, auxiliary fuel parameters, and tail gas treatment unit operating parameters. Feed parameters include feed rate; combustion air parameters include at least one of primary air volume, secondary air volume, primary air temperature, secondary air temperature, and total combustion air volume; furnace operating parameters include at least one of grate speed and furnace negative pressure; auxiliary fuel parameters include at least one of auxiliary burner power and fuel addition rate; and tail gas treatment unit operating parameters include at least one of quench alkali dosage, quench tower outlet target temperature, adsorbent injection rate, bag filter cleaning frequency, and differential pressure setting. Then, the target incinerator characteristic data and target combustion parameters are input into the process characteristic prediction model to obtain the corresponding predicted process characteristic parameters.

[0053] Secondly, this application also provides a digital twin-based incineration parameter recommendation system, which adopts the following technical solution:

[0054] A digital twin-based incineration parameter recommendation system includes:

[0055] The data acquisition module is used to acquire historical operating data from the digital twin of the waste incineration system. The historical operating data includes historical incinerated material characteristic data, historical incineration parameters, historical incineration process characteristic parameters, and historical incineration evaluation level data.

[0056] The model building module is used to build an incineration condition rating model based on the historical incineration process characteristic parameters and the historical incineration evaluation level data, and to build a process characteristic prediction model based on the historical incinerated material characteristic data, the historical incineration parameters and the historical incineration process characteristic parameters.

[0057] The process characteristic parameter search module is used to set process characteristic parameter constraints, call the incineration condition rating model, and search for a set of target process characteristic parameters according to the genetic algorithm so that the corresponding prediction level is consistent with the user's current demand level, wherein the set of target process characteristic parameters satisfies the process characteristic parameter constraints.

[0058] The recommended incineration parameter search module is used to set incineration parameter constraints, call the process feature prediction model, and search for a set of recommended incineration parameters based on the particle swarm optimization algorithm, so that the current incinerated material characteristic data and the predicted feature parameters of the set of recommended incineration parameters are closest to the set of target process feature parameters, wherein the set of recommended incineration parameters satisfies the incineration parameter constraints.

[0059] Thirdly, this application also provides a computer device, which adopts the following technical solution:

[0060] A computer device includes a memory and a processor, the memory storing a computer program executable on the processor, the processor executing the computer program to implement the method described in the first aspect.

[0061] Fourthly, this application also provides a computer-readable storage medium, which adopts the following technical solution:

[0062] A computer-readable storage medium storing a computer program capable of being loaded by a processor and executing the method described in the first aspect.

[0063] In summary, this application includes at least the following beneficial technical effects: Historical operating data is obtained from the digital twin of the waste incineration system, including historical incinerated material characteristic data, historical incineration parameters, historical incineration process characteristic parameters, and historical incineration evaluation level data. Then, based on the historical incineration process characteristic parameters and historical incineration evaluation level data, an incineration condition rating model is constructed. Furthermore, based on the historical incinerated material characteristic data, historical incineration parameters, and historical incineration process characteristic parameters, a process characteristic prediction model is constructed. Then, process characteristic parameter constraints are set, the incineration condition rating model is invoked, and a set of target process characteristic parameters is searched using a genetic algorithm to ensure that the corresponding prediction level matches the user's current demand level. This set of target process characteristic parameters satisfies the process characteristic parameter constraints. Finally, incineration parameter constraints are set, the process characteristic prediction model is invoked, and a set of recommended incineration parameters is searched using a particle swarm optimization algorithm to ensure that the current incinerated material characteristic data and a set of recommended incineration parameters match the user's current demand level. The predicted characteristic parameters of the recommended incineration parameters are closest to a set of target process characteristic parameters, where the set of recommended incineration parameters satisfies the incineration parameter constraints. The incineration parameter recommendation method based on digital twins proposed in this invention achieves intelligent and automated incineration parameter recommendation by constructing a dual-model and combining a two-layer optimization architecture of genetic algorithm and particle swarm optimization. This method can quickly and accurately automatically generate optimal operating parameters based on the dynamic changes in waste characteristics and user-specific operational goals, completely changing the outdated mode of relying on manual experience and trial-and-error adjustments. Through digitalization and automation, it reduces the dependence on operator experience, reduces human error, thereby improving the stability and processing efficiency of the incineration process, ensuring complete combustion and minimum pollutant emissions, and enhancing the system's adaptability to complex and variable raw materials. It also improves the flexibility and robustness of the production line, effectively ensuring operational safety and economy, and providing core key technical support for the intelligent upgrading of industrial incineration equipment. Attached Figure Description

[0064] Figure 1 This is a schematic diagram of the overall process of an embodiment of this application.

[0065] Figure 2 This is a schematic diagram of the system structure of this application.

[0066] Figure 3 This is a structural block diagram of the computer device described in this application. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0068] This application discloses a method for recommending incineration parameters based on digital twins.

[0069] Reference Figure 1 A digital twin-based method for recommending incineration parameters includes:

[0070] Step S11: Obtain historical operating data from the digital twin of the waste incineration system.

[0071] The historical operational data includes historical incinerated material characteristic data, historical incineration parameters, historical incineration process characteristic parameters, and historical incineration evaluation level data.

[0072] It should be noted that in step S11, by connecting to the digital twin of the waste incineration system, historical operating data is comprehensively collected and integrated. This data covers four dimensions: incinerator characteristic data (such as calorific value, moisture, and composition), incineration operating parameters (such as air volume, temperature, and velocity), process characteristic parameters (such as furnace temperature distribution and pollutant concentration), and expert evaluation data on historical operating conditions.

[0073] Step S12: Based on historical incineration process characteristic parameters and historical incineration evaluation level data, construct an incineration condition rating model, and based on historical incinerated material characteristic data, historical incineration parameters, and historical incineration process characteristic parameters, construct a process characteristic prediction model.

[0074] It should be noted that in step S12, based on the data obtained in step S11, two key models are constructed in parallel: the incineration condition rating model and the process characteristic prediction model. The incineration condition rating model is a classification or scoring model that learns the mapping relationship between process characteristic parameters and operating evaluation levels. For example, it can learn what combinations of temperature, pressure, and emission concentration correspond to "excellent," "good," or "poor" operating conditions. The process characteristic prediction model is a regression prediction model that learns how different incineration operating parameters affect the process output (i.e., process characteristic parameters) under given waste characteristics. It is essentially a "virtual incinerator" used to predict combustion results under different parameters.

[0075] Step S13: Set process characteristic parameter constraints, call the incineration operating condition rating model, and search for a set of target process characteristic parameters according to the genetic algorithm so that the corresponding prediction level is consistent with the user's current demand level. Among them, a set of target process characteristic parameters satisfies the process characteristic parameter constraints.

[0076] It should be noted that step S13 defines the "ideal target" that the incineration process needs to achieve, thereby realizing the reverse search of the target operating conditions. First, constraints on process characteristic parameters are set according to process and safety requirements (such as the temperature must not exceed the upper limit of refractory materials and the pollutant concentration must be lower than the national standard limit). Then, taking the user's current demand level (such as requiring "excellent" level operation) as the optimization target, the advantages of global search by genetic algorithm are used to reverse search for a set of optimal target process characteristic parameters within the constraints (e.g., temperature = 1150°C ± 20°C, CO < 50 mg / m³, efficiency > 99%).

[0077] Step S14: Set incineration parameter constraints, call the process feature prediction model, and search for a set of recommended incineration parameters based on the particle swarm optimization algorithm, so that the current incinerated material characteristic data and the predicted feature parameters of a set of recommended incineration parameters are closest to a set of target process feature parameters, wherein a set of recommended incineration parameters satisfies the incineration parameter constraints.

[0078] It should be noted that in step S14, in order to solve the problem of "how to operate to achieve the ideal goal set in step S13", constraints on the incineration parameters (such as the adjustment range of air volume and rotation speed) are set for the specific waste to be processed. Then, the process feature prediction model is called and the particle swarm optimization algorithm is used to quickly search for a set of recommended incineration parameters that are most suitable for the current waste (such as: the primary and secondary air ratio is set to 7:3 and the grate speed is set to 50%), so as to guide the actual production operation.

[0079] In the above implementation, historical operating data is obtained from the digital twin of the waste incineration system. This historical operating data includes historical incinerated material characteristic data, historical incineration parameters, historical incineration process characteristic parameters, and historical incineration evaluation level data. Then, based on the historical incineration process characteristic parameters and historical incineration evaluation level data, an incineration condition rating model is constructed. Next, based on the historical incinerated material characteristic data, historical incineration parameters, and historical incineration process characteristic parameters, a process characteristic prediction model is constructed. Then, process characteristic parameter constraints are set, and the incineration condition rating model is invoked. A set of target process characteristic parameters is searched using a genetic algorithm to ensure that the corresponding prediction level matches the user's current requirement level. This set of target process characteristic parameters satisfies the process characteristic parameter constraints. Then, incineration parameter constraints are set, and the process characteristic prediction model is invoked. A set of recommended incineration parameters is searched using a particle swarm optimization algorithm to ensure that the current incinerated material characteristic data and the recommended incineration parameters match the current incinerated material characteristic data. The predicted characteristic parameters are closest to a set of target process characteristic parameters, where a set of recommended incineration parameters satisfies the incineration parameter constraints. The incineration parameter recommendation method based on digital twin proposed in this invention achieves intelligent and automated incineration parameter recommendation by constructing a dual-model and combining a two-layer optimization architecture of genetic algorithm and particle swarm optimization. This method can quickly and accurately automatically generate optimal operating parameters based on the dynamic changes of waste characteristics and user-specific operational goals, completely changing the outdated mode of relying on manual experience and trial-and-error adjustments. Through digitalization and automation, it reduces the dependence on operator experience, reduces human error, thereby improving the stability and processing efficiency of the incineration process, ensuring complete combustion and minimum pollutant emissions, and enhancing the system's adaptability to complex and variable raw materials. It also improves the flexibility and robustness of the production line, effectively ensuring operational safety and economy, and providing core key technical support for the intelligent upgrading of industrial incineration equipment.

[0080] As a further implementation of the method, the step of constructing an incineration condition rating model based on historical incineration process characteristic parameters and historical incineration evaluation level data includes:

[0081] Step S21: Preprocess the characteristic parameters of the historical incineration process and the historical incineration evaluation level data to obtain the first model training data.

[0082] It should be noted that step S21 transforms the raw, messy field data into a high-quality, well-organized dataset. Specific tasks include handling missing values ​​and obvious outliers in the data to prevent them from interfering with model learning; standardizing or normalizing feature parameters with different dimensions (such as temperature, pressure, and concentration) to eliminate the impact of numerical range differences on the model; and converting textual evaluation ratings (such as "Excellent" and "Good") into numerical labels to prepare for subsequent supervised learning.

[0083] Step S22: Divide the training data of the first model into a first training set and a second test set according to a preset ratio.

[0084] It should be noted that, through step S22, the preprocessed complete dataset is randomly divided into a first training set and a second test set according to a preset ratio (e.g., 8:2), thus artificially constructing an ideal verification environment. The training set is used to simulate the inherent patterns of the model data, while the test set is used to objectively and fairly evaluate the model's learning effect.

[0085] Step S23: Set the hyperparameters of the predefined decision tree model according to the random network search algorithm, and use the corresponding root mean square error (RMSE) and coefficient of determination (R²) as the first evaluation index.

[0086] It should be noted that step S22 configures the optimal "parameter tuning scheme" for the selected decision tree model. The performance of the decision tree model is highly dependent on the setting of hyperparameters (such as the maximum depth of the tree and the minimum number of samples in the leaf nodes). This step uses a random network search algorithm, which is an efficient hyperparameter optimization strategy. It finds the best configuration by randomly sampling and combining and verifying parameters in a defined parameter space. At the same time, the root mean square error (RMSE) and the coefficient of determination (R²) are explicitly used as the primary evaluation indicators to provide a quantitative and objective evaluation standard for the model performance.

[0087] Step S24: Train the decision tree model based on the first training set to obtain a trained decision tree model.

[0088] It should be noted that in step S24, the first training set (feature data and their corresponding level labels) is input into a decision tree algorithm with a pre-defined architecture and hyperparameters. The algorithm iteratively searches for the optimal feature split point, gradually constructing a decision tree that can most accurately classify data from different operating conditions. After training, the output is a decision tree model that has learned the mapping relationship from process features to operating levels and possesses preliminary predictive capabilities.

[0089] Step S25: Test the trained decision tree model according to the first test set, and determine whether the error is within the preset range according to the first evaluation index. If so, use the trained decision tree model as the incineration condition rating model.

[0090] It should be noted that in step S25, a second test set that was not involved in training is input into the trained decision tree model to obtain its prediction results, and the RMSE and R² values ​​are calculated. By comparing these two key indicators with preset performance thresholds, it is scientifically determined whether the model error is within an acceptable range; if the verification passes, the incineration condition rating model is obtained; if it fails, the aforementioned steps need to be repeated.

[0091] In the above implementation, in order to construct the incineration condition rating model, the characteristic parameters of the historical incineration process and the historical incineration evaluation level data are preprocessed to obtain the first model training data. Then, the first model training data is divided into a first training set and a second test set according to a preset ratio. Then, the hyperparameters of the predefined decision tree model are set according to the random network search algorithm, and the corresponding root mean square error (RMSE) and coefficient of determination (R²) are used as the first evaluation index. Then, the decision tree model is trained according to the first training set to obtain a trained decision tree model. Then, the trained decision tree model is tested according to the first test set, and the error is judged according to the first evaluation index to determine whether it is within the preset range. If the error is within the preset range, the trained decision tree model is used as the incineration condition rating model.

[0092] As a further implementation of the method, the step of constructing a process characteristic prediction model based on historical incinerated material characteristic data, historical incineration parameters, and historical incineration process characteristic parameters includes:

[0093] Step S31: Preprocess the historical incineration material characteristic data, historical incineration parameters, and historical incineration process characteristic parameters to obtain the second model training data.

[0094] Step S32: Divide the training data of the second model into a second training set and a second test set according to a preset ratio.

[0095] Step S33: Set the hyperparameters of the predefined SVM model according to the random network search algorithm, and use the corresponding root mean square error (RMSE) and coefficient of determination (R²) as the second evaluation index.

[0096] Step S34: Train the SVM model based on the second training set to obtain the trained SVM model.

[0097] Step S35: Test the trained SVM model according to the second test set, and determine whether the error is within the preset range according to the second evaluation index. If so, use the trained SVM model as the process feature prediction model.

[0098] It should be noted that the technical principles of steps S31 to S35 are basically the same as those of steps S21 to S25, and the implementation process of steps S31 to S35 can be referred to steps S21 to S25.

[0099] In the above implementation, in order to construct the process feature prediction model, the historical incinerator characteristic data, historical incineration parameters, and historical incineration process characteristic parameters are preprocessed to obtain the second model training data. Then, the second model training data is divided into a second training set and a second test set according to a preset ratio. Then, the hyperparameters of the predefined SVM model are set according to the random network search algorithm, and the corresponding root mean square error (RMSE) and coefficient of determination (R²) are used as the second evaluation index. Then, the SVM model is trained according to the second training set to obtain the trained SVM model. Then, the trained SVM model is tested according to the second test set, and the error is judged according to the second evaluation index to determine whether it is within the preset range. If the error is within the preset range, the trained SVM model is used as the process feature prediction model.

[0100] As a further implementation of the method, the step of calling the incineration condition rating model and searching for a set of target process characteristic parameters using a genetic algorithm to ensure that the corresponding predicted level matches the user's current demand level includes:

[0101] Step S41: Determine the feasible region of each process characteristic parameter according to the process characteristic parameter constraints, and randomly generate an initial population containing M chromosomes within the feasible region.

[0102] It should be noted that step S41 is the starting point of the genetic algorithm optimization, which aims to generate a set of random feasible solutions. First, the feasible range of each parameter is determined according to the process characteristic parameter constraints defined by process and safety requirements (such as upper temperature limit and pollutant concentration limit), which together constitute the search space (feasible region) of the algorithm. Then, an initial population containing M chromosomes is randomly generated within this feasible region, with each chromosome representing a possible combination of a set of process characteristic parameters.

[0103] Step S42: For each chromosome in the initial population, the chromosome is decoded into a set of process feature parameters, and the set of process feature parameters is input into the incineration condition rating model to obtain the corresponding prediction level and the probability value corresponding to the prediction level. The corresponding first current fitness value is determined according to the probability value, wherein the first current fitness value is used to represent the difference or consistency between the prediction level and the current demand level.

[0104] It should be noted that in step S42, for each chromosome in the current population, it is first decoded into a set of specific process feature parameters. These parameters are then input into the trained incineration condition rating model to obtain its predicted operating level and the probability value corresponding to that predicted level. The core purpose of defining the first current fitness value here is to quantify the consistency between the prediction result and the user's expectations. Its calculation logic closely revolves around the user-set current demand level. That is, if the model's predicted level exactly matches the current demand level, the fitness value is the probability value of that predicted level, thereby encouraging the algorithm to prioritize parameter combinations that can achieve the target operating condition with high confidence. Conversely, if the predicted level does not match the demand level, the fitness value will be assigned a penalty value of zero or a very low value to ensure that the algorithm effectively eliminates solutions that deviate from the target during the evolutionary process. In this way, the fitness function directly guides the search process to optimize in the direction of meeting the user's specific needs.

[0105] Step S43: Based on roulette wheel selection or tournament selection, select superior individuals from the current population to form the parent population according to the first current fitness value of the chromosome, wherein chromosomes with higher first current fitness values ​​have a higher probability of being selected.

[0106] It should be noted that in step S43, superior individuals are selected from the current population to serve as parents for reproduction. Based on strategies such as roulette wheel selection or tournament selection, selection is made according to the fitness value of each chromosome calculated in the previous step. Chromosomes with higher fitness values ​​have a greater probability of being selected into the parent population, thus ensuring that superior genes have a higher chance of being preserved and passed on to offspring, driving the population as a whole towards a better solution.

[0107] Step S44: Randomly select chromosomes from the parent population and perform crossover operations to obtain crossover chromosomes. Then, perform mutation operations on the crossover chromosomes to obtain mutated chromosomes.

[0108] It should be noted that in step S44, the crossover operation involves randomly selecting two chromosomes from the parent population and exchanging some of their "genes" (parameter values) with a certain crossover probability to generate a new crossover chromosome, thereby combining the superior traits of the parents; the mutation operation involves randomly changing the values ​​of one or more gene positions of the chromosome generated after crossover with a small mutation probability to obtain a mutated chromosome, thereby introducing new genes and helping the algorithm to escape local optima and explore new possible regions.

[0109] Step S45: Construct a new population based on the crossover chromosome and the mutated chromosome, and iterate again based on the new population until the first iteration termination condition is met. The first iteration termination condition includes the first maximum number of iterations, the convergence of the first current fitness value, or the first current fitness value reaching the first preset value.

[0110] It should be noted that in step S45, this step completes one generation of evolution and determines whether to stop the search. First, a completely new offspring population is constructed using the new chromosomes generated after crossover and mutation operations, replacing the original parent population. Then, it is checked whether the algorithm meets the first iteration termination condition. These conditions typically include: the number of generations has reached the preset first maximum number of iterations, the first current fitness value of the best chromosome in the population has converged (i.e., it no longer significantly improves over multiple generations), or the first current fitness value of the best chromosome has reached the first preset value (e.g., 0.99). If any of these conditions are met, the iteration stops; otherwise, the process returns to step S42 to continue evolution.

[0111] Step S46: When the iteration terminates, output the target chromosome with the highest fitness in the current population, and decode the target chromosome to obtain a set of target process feature parameters.

[0112] It should be noted that in step S46, when the iteration process ends due to the fulfillment of the termination condition, the algorithm selects the target chromosome with the highest fitness from the current final population. This chromosome represents the best solution found in the entire evolutionary search process. After decoding it, the final recommended set of target process feature parameters (e.g., optimal temperature, ideal pollutant concentration, etc.) can be obtained. This set of parameters is the process state target that has the highest probability of enabling the user's current demand level (e.g., "excellent") to achieve.

[0113] In the above implementation, in order to search for a set of target process feature parameters, the feasible region of each process feature parameter is determined according to the constraints of the process feature parameters. An initial population containing M chromosomes is randomly generated within the feasible region. Then, for each chromosome in the initial population, the chromosome is decoded into a set of process feature parameters, and this set of process feature parameters is input into the incineration condition rating model to obtain the corresponding prediction level and the probability value corresponding to the prediction level. The corresponding first current fitness value is determined based on the probability value, where the first current fitness value represents the difference or consistency between the prediction level and the current demand level. Then, based on a roulette wheel selection or tournament selection method, the chromosomes are selected from the first current fitness value... In the current population, superior individuals are selected to form a parent population. Chromosomes with higher current fitness values ​​are more likely to be selected. Then, chromosomes are randomly selected from the parent population and crossover is performed to obtain crossover chromosomes. These crossover chromosomes are then mutated to obtain mutated chromosomes. A new population is then constructed based on the crossover and mutated chromosomes, and the iteration is repeated until the first iteration termination condition is met. The first iteration termination condition includes the first maximum number of iterations, convergence of the first current fitness value, or the first current fitness value reaching a first preset value. When the iteration terminates, the target chromosome with the highest fitness in the current population is output, and the target chromosome is decoded to obtain a set of target process feature parameters.

[0114] As a further implementation of the method, the step of calling the process feature prediction model and searching for a set of recommended incineration parameters using the particle swarm optimization algorithm, so that the predicted feature parameters of the current incinerated material characteristic data and the set of recommended incineration parameters are closest to a set of target process feature parameters, includes:

[0115] Step S51: Set the search space boundary for particles according to the incineration parameter constraints, and randomly initialize a group of particles within the corresponding search space. Each particle represents a set of candidate incineration parameter combinations, and each particle is assigned a random initial position x. 0 and initial velocity v 0 .

[0116] It should be noted that step S51 sets up the search framework and generates an initial solution for the Particle Swarm Optimization (PSO) algorithm. First, the value range of each parameter to be optimized is clearly defined based on the physical limits and operating procedures of the incineration system (i.e., incineration parameter constraints, such as airflow range and velocity limits), forming the algorithm's search space. Then, a swarm of particles is randomly initialized within this search space. Each particle represents a candidate combination of incineration parameters (such as a specific set of primary and secondary airflow and grate velocity values), and is randomly assigned an initial position (x). 0 (representing parameter values) and initial velocity (v) 0(representing the trend and magnitude of parameter changes), preparing for iterative search.

[0117] Step S52, update the formula based on position and speed update formula

[0118] Iterate for each particle separately, where x k+1 For the (k+1)th iteration point, x k For the k-th iteration point, v K v represents the velocity in the k-th iteration. k+1 denoted as ρ, where ρ is the velocity in the (k+1)th iteration, w is the inertia weight, c1 is the cognitive factor, r1 is a random number between 0 and 1, pBest is the best position found during the search process, c2 is the social learning factor, r2 is a random number between 0 and 1, gBest is the best position found globally, rand is the random number function, and δ is the perturbation amplitude.

[0119] Step S53: During the iteration process, for each particle in each iteration, the current position x of the current particle is... k The current incinerator characteristics data are input into the process characteristic prediction model to predict intermediate process characteristic parameters. .

[0120] It should be noted that, for step S51, in each iteration, for each particle's current candidate parameters (i.e., position x) k The data, along with the characteristic data of the waste to be treated, is input into a pre-trained process feature prediction model. This model, acting as a "virtual incinerator," simulates the intermediate process characteristic parameters that can be achieved under this set of parameters. (e.g., predicted temperature, emission concentration, etc.), which is equivalent to a rapid "digital twin test burn" of countless combinations of parameters.

[0121] Step S54: Calculate the intermediate process characteristic parameters and a set of target process characteristic parameters y target The corresponding second current fitness value F(x) k ), and based on the second current fitness value F(x) k The individual's historical best fitness value F(pBest) and the global historical best fitness value F(gBest) are updated, where the second current fitness value F(x) is updated. k The mean square error or Euclidean distance is negative.

[0122] Step S55: Determine whether the second iteration termination condition is met. If so, the combination of incineration parameters represented by gBest is used as a set of recommended incineration parameters. The second iteration termination condition includes the second maximum number of iterations, the convergence of the global historical optimal fitness value F(gBest) corresponding to the global historical optimal solution gBest, or the global historical optimal fitness value F(gBest) has reached the second preset value.

[0123] In the above implementation, in order to search for a set of recommended incineration parameters, the search space boundary of the particles is set according to the constraints of the incineration parameters, and a group of particles is randomly initialized in the corresponding search space. Each particle represents a set of candidate combinations of incineration parameters, and each particle is assigned a random initial position x. 0 and initial velocity v 0 Then update the formula based on the position. and speed update formula

[0124] Iterate for each particle separately, where x k+1 For the (k+1)th iteration point, x k For the k-th iteration point, v K v represents the velocity in the k-th iteration. k+1 Let x be the velocity in the (k+1)th iteration, w be the inertia weight, c1 be the cognitive factor, r1 be a random number between 0 and 1, pBest be the best position found during the search process, c2 be the social learning factor, r2 be a random number between 0 and 1, gBest be the best position found globally, rand be the random number function, and δ be the perturbation amplitude. Then, during the iteration process, for each particle in each iteration, the current position x of the current particle is... k The current incinerator characteristics data are input into the process characteristic prediction model to predict intermediate process characteristic parameters. Then calculate the intermediate process characteristic parameters. and a set of target process characteristic parameters y target The corresponding second current fitness value F(x) k ), and based on the second current fitness value F(x) k The individual's historical best fitness value F(pBest) and the global historical best fitness value F(gBest) are updated, where the second current fitness value F(x) is updated. kIf the mean square error or Euclidean distance is negative, then it is determined whether the second iteration termination condition is met. If the second iteration termination condition is met, the combination of incineration parameters represented by gBest is used as a set of recommended incineration parameters. The second iteration termination condition includes the second maximum number of iterations, the convergence of the global historical optimal fitness value F(gBest) corresponding to the global historical optimal solution gBest, or the global historical optimal fitness value F(gBest) has reached the second preset value.

[0125] As a further implementation of the method, the method also includes:

[0126] Step S61: Obtain the characteristic parameters of the target incineration process, wherein the characteristic parameters of the target incineration process include at least one of thermal parameters, flue gas composition parameters, and combustion efficiency parameters; the thermal parameters include at least one of furnace temperature parameters, secondary combustion chamber outlet temperature, and loss on ignition; and the flue gas composition parameters include at least one of oxygen concentration, carbon monoxide concentration, nitrogen oxide concentration, and sulfur oxide concentration.

[0127] Step S62: Input the characteristic parameters of the target incineration process into the incineration condition rating model to obtain the corresponding predicted operation evaluation level.

[0128] In the above embodiments, in order to predict the operation evaluation level based on the incineration condition rating model, target incineration process characteristic parameters are obtained. The target incineration process characteristic parameters include at least one of thermal parameters, flue gas composition parameters, and combustion efficiency parameters. The thermal parameters include at least one of furnace temperature parameters, secondary combustion chamber outlet temperature, and loss on ignition. The flue gas composition parameters include at least one of oxygen concentration, carbon monoxide concentration, nitrogen oxide concentration, and sulfur oxide concentration. Then, the target incineration process characteristic parameters are input into the incineration condition rating model to obtain the corresponding predicted operation evaluation level.

[0129] As a further implementation of the method, the method also includes:

[0130] Step S71: Obtain target incinerator characteristic data and target combustion parameters. The target incinerator characteristic data includes at least one of physical morphology, element content, and calorific value. The target combustion parameters include at least one of feed parameters, combustion air parameters, furnace operation parameters, auxiliary fuel parameters, and tail gas treatment unit operation parameters. The feed parameters include feed rate. The combustion air parameters include at least one of primary air volume, secondary air volume, primary air temperature, secondary air temperature, and total combustion air volume. The furnace operation parameters include at least one of grate speed and furnace negative pressure. The auxiliary fuel parameters include at least one of auxiliary burner power and fuel addition rate. The tail gas treatment unit operation parameters include at least one of quench alkali dosage, quench tower outlet target temperature, adsorbent injection rate, bag filter cleaning frequency, and differential pressure setting.

[0131] Step S72: Input the target incinerator characteristic data and target incineration parameters into the process characteristic prediction model to obtain the corresponding predicted process characteristic parameters.

[0132] In the above embodiments, in order to predict process characteristic parameters based on the process characteristic prediction model, target incinerator characteristic data and target combustion parameters are obtained. The target incinerator characteristic data includes at least one of physical morphology, element content, and calorific value. The target combustion parameters include at least one of feed parameters, combustion air parameters, furnace operation parameters, auxiliary fuel parameters, and tail gas treatment unit operation parameters. The feed parameters include feed rate. The combustion air parameters include at least one of primary air volume, secondary air volume, primary air temperature, secondary air temperature, and total combustion air volume. The furnace operation parameters include at least one of grate speed and furnace negative pressure. The auxiliary fuel parameters include at least one of auxiliary burner power and fuel addition rate. The tail gas treatment unit operation parameters include at least one of quench alkali dosage, quench tower outlet target temperature, adsorbent injection rate, bag filter cleaning frequency, and differential pressure setting. Then, the target incinerator characteristic data and target combustion parameters are input into the process characteristic prediction model to obtain the corresponding predicted process characteristic parameters.

[0133] This application also discloses a digital twin-based incineration parameter recommendation system.

[0134] refer to Figure 2 A digital twin-based incineration parameter recommendation system includes:

[0135] The data acquisition module is used to acquire historical operating data from the digital twin of the waste incineration system. The historical operating data includes historical incinerated material characteristic data, historical incineration parameters, historical incineration process characteristic parameters, and historical incineration evaluation level data.

[0136] The model building module is used to build an incineration condition rating model based on historical incineration process characteristic parameters and historical incineration evaluation level data, and to build a process characteristic prediction model based on historical incinerated material characteristic data, historical incineration parameters and historical incineration process characteristic parameters.

[0137] The process characteristic parameter search module is used to set process characteristic parameter constraints, call the incineration operating condition rating model, and search for a set of target process characteristic parameters according to the genetic algorithm so that the corresponding prediction level is consistent with the user's current demand level. Among them, a set of target process characteristic parameters satisfies the process characteristic parameter constraints.

[0138] The recommended incineration parameter search module is used to set incineration parameter constraints, call the process feature prediction model, and search for a set of recommended incineration parameters based on the particle swarm optimization algorithm. This ensures that the current incinerated material characteristic data and the predicted feature parameters of the recommended incineration parameters are as close as possible to a set of target process feature parameters. The recommended incineration parameters satisfy the incineration parameter constraints.

[0139] The digital twin-based incineration parameter recommendation system of the present invention can implement any of the digital twin-based incineration parameter recommendation methods, and the specific working process of the digital twin-based incineration parameter recommendation system of the present invention can refer to the corresponding process in the above-mentioned digital twin-based incineration parameter recommendation methods.

[0140] This application also discloses a computer device.

[0141] refer to Figure 3 A computer device includes a memory and a processor, the memory storing a computer program executable on the processor, the processor executing the computer program to implement any of the above-described methods for recommending incineration parameters based on digital twins.

[0142] This application also discloses a computer-readable storage medium.

[0143] A computer-readable storage medium storing a computer program that can be loaded by a processor and executed any of the above-described methods for recommending incineration parameters based on digital twins.

[0144] The computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device; the program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0145] The above are all preferred embodiments of this application and are not intended to limit the scope of protection of this application. Any feature disclosed in this specification (including the abstract and drawings) may be replaced by other equivalent or similar features unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is only one example of a series of equivalent or similar features.

Claims

1. A method for recommending incineration parameters based on digital twins, characterized in that, include: Historical operating data is obtained from the digital twin of the waste incineration system, including historical incinerated material characteristic data, historical incineration parameters, historical incineration process characteristic parameters, and historical incineration evaluation level data. Based on the historical incineration process characteristic parameters and the historical incineration evaluation level data, an incineration condition rating model is constructed, and based on the historical incinerated material characteristic data, the historical incineration parameters, and the historical incineration process characteristic parameters, a process characteristic prediction model is constructed. Set process characteristic parameter constraints, call the incineration condition rating model, and search for a set of target process characteristic parameters according to the genetic algorithm so that the corresponding prediction level is consistent with the user's current demand level, wherein the set of target process characteristic parameters satisfies the process characteristic parameter constraints. Set incineration parameter constraints, call the process feature prediction model, and search for a set of recommended incineration parameters according to the particle swarm optimization algorithm, so that the current incinerated material characteristic data and the predicted feature parameters of the set of recommended incineration parameters are closest to the set of target process feature parameters, wherein the set of recommended incineration parameters satisfies the incineration parameter constraints. The step of calling the incineration condition rating model and searching for a set of target process characteristic parameters using a genetic algorithm to ensure that the corresponding predicted level matches the user's current demand level includes: The feasible region of each process characteristic parameter is determined according to the constraints of the process characteristic parameters, and an initial population containing M chromosomes is randomly generated within the feasible region. For each chromosome in the initial population, the chromosome is decoded into a set of process feature parameters, and the set of process feature parameters is input into the incineration condition rating model to obtain the corresponding prediction level and the probability value corresponding to the prediction level. The corresponding first current fitness value is determined according to the probability value, wherein the first current fitness value is used to represent the difference or consistency between the prediction level and the current demand level. Based on roulette wheel selection or tournament selection, superior individuals are selected from the current population to form the parent population according to the first current fitness value of the chromosome, wherein chromosomes with higher first current fitness values ​​are more likely to be selected. Chromosomes are randomly selected from the parent population and crossover is performed to obtain crossover chromosomes. The crossover chromosomes are then mutated to obtain mutated chromosomes. A new population is constructed based on the crossover chromosome and the variant chromosome, and the iteration is repeated based on the new population until a first iteration termination condition is met, wherein the first iteration termination condition includes a first maximum number of iterations, convergence of the first current fitness value, or the first current fitness value has reached a first preset value. When the iteration terminates, the target chromosome with the highest fitness in the current population is output, and the target chromosome is decoded to obtain a set of target process feature parameters; The step of calling the process feature prediction model and searching for a set of recommended incineration parameters using the particle swarm optimization algorithm, so that the predicted feature parameters of the current incinerated material characteristics and the set of recommended incineration parameters are closest to the set of target process feature parameters, includes: The search space boundary for particles is set according to the incineration parameter constraints, and a group of particles is randomly initialized within the corresponding search space. Each particle represents a set of candidate combinations of incineration parameters, and each particle is assigned a random initial position x. 0 and initial velocity v 0 ; Update formula based on location and speed update formula Iterate for each particle separately, where x k+1 For the (k+1)th iteration point, x k For the k-th iteration point, v K v represents the velocity in the k-th iteration. k+1 denoted as , where is the velocity in the (k+1)th iteration, w is the inertia weight, c1 is the cognitive factor, r1 is a random number between 0 and 1, pBest is the best position found during the search process, c2 is the social learning factor, r2 is a random number between 0 and 1, gBest is the best position found globally, rand is the random number function, and δ is the perturbation amplitude. During the iteration process, for each particle in each iteration, the current position x of the current particle is... k The current incinerator characteristics data are input into the process feature prediction model to predict intermediate process characteristic parameters. ; Calculate the intermediate process characteristic parameters and the set of target process characteristic parameters y target The corresponding second current fitness value F(x) k ), and based on the second current fitness value F(x) k The individual's historical best fitness value F(pBest) and the global historical best fitness value F(gBest) are updated, wherein the second current fitness value F(x) is updated. k The mean square error or Euclidean distance is negative. Determine whether the second iteration termination condition is met. If so, the combination of incineration parameters represented by gBest is used as a set of recommended incineration parameters. The second iteration termination condition includes the second maximum number of iterations, the convergence of the global historical optimal fitness value F(gBest) corresponding to the global historical optimal solution gBest, or the global historical optimal fitness value F(gBest) has reached the second preset value.

2. The method for recommending incineration parameters based on digital twins according to claim 1, characterized in that, The step of constructing an incineration condition rating model based on the historical incineration process characteristic parameters and historical incineration evaluation level data includes: Preprocessing the historical incineration process characteristic parameters and historical incineration evaluation level data yields the first model training data. The training data of the first model is divided into a first training set and a first test set according to a preset ratio; The hyperparameters of the predefined decision tree model are set according to the random network search algorithm, and the corresponding root mean square error (RMSE) and coefficient of determination (R²) are used as the first evaluation index. The decision tree model is trained based on the first training set to obtain a trained decision tree model; The trained decision tree model is tested according to the first test set, and the error is judged according to the first evaluation index to determine whether it is within the preset range. If so, the trained decision tree model is used as the incineration condition rating model.

3. The method for recommending incineration parameters based on digital twins according to claim 1, characterized in that, The step of constructing a process characteristic prediction model based on the historical incinerated material characteristic data, the historical incineration parameters, and the historical incineration process characteristic parameters includes: The historical incinerated material characteristic data, the historical incineration parameters, and the historical incineration process characteristic parameters are preprocessed to obtain the second model training data. The training data of the second model is divided into a second training set and a second test set according to a preset ratio; The hyperparameters of the predefined SVM model are set according to the random network search algorithm, and the corresponding root mean square error (RMSE) and coefficient of determination (R²) are used as the second evaluation index. The SVM model is trained using the second training set to obtain a trained SVM model. The trained SVM model is tested according to the second test set, and the error is judged according to the second evaluation index to determine whether it is within the preset range. If it is, the trained SVM model is used as the process feature prediction model.

4. The method for recommending incineration parameters based on digital twins according to claim 1, characterized in that, The method further includes: The target incineration process characteristic parameters are obtained, wherein the target incineration process characteristic parameters include at least one of thermal parameters, flue gas composition parameters, and combustion efficiency parameters; the thermal parameters include at least one of furnace temperature parameters, secondary combustion chamber outlet temperature, and loss on ignition; and the flue gas composition parameters include at least one of oxygen concentration, carbon monoxide concentration, nitrogen oxide concentration, and sulfur oxide concentration. The characteristic parameters of the target incineration process are input into the incineration condition rating model to obtain the corresponding predicted operation evaluation level.

5. The method for recommending incineration parameters based on digital twins according to claim 1, characterized in that, The method further includes: Acquire target incinerator characteristic data and target incineration parameters, wherein the target incinerator characteristic data includes at least one of physical morphology, element content, and calorific value; the target incineration parameters include at least one of feed parameters, combustion air parameters, furnace operation parameters, auxiliary fuel parameters, and exhaust gas treatment unit operation parameters; the feed parameters include feed rate; the combustion air parameters include at least one of primary air volume, secondary air volume, primary air temperature, secondary air temperature, and total combustion air volume; the furnace operation parameters include at least one of grate speed and furnace negative pressure; the auxiliary fuel parameters include at least one of auxiliary burner power and fuel addition rate; and the exhaust gas treatment unit operation parameters include at least one of quench alkali dosage, quench tower outlet target temperature, adsorbent injection rate, bag filter cleaning frequency, and differential pressure setting. The target incinerator characteristic data and target incineration parameters are input into the process feature prediction model to obtain the corresponding predicted process feature parameters.

6. A digital twin-based incineration parameter recommendation system, characterized in that, include: The data acquisition module is used to acquire historical operating data from the digital twin of the waste incineration system. The historical operating data includes historical incinerated material characteristic data, historical incineration parameters, historical incineration process characteristic parameters, and historical incineration evaluation level data. The model building module is used to build an incineration condition rating model based on the historical incineration process characteristic parameters and the historical incineration evaluation level data, and to build a process characteristic prediction model based on the historical incinerated material characteristic data, the historical incineration parameters and the historical incineration process characteristic parameters. The process characteristic parameter search module is used to set process characteristic parameter constraints, call the incineration condition rating model, and search for a set of target process characteristic parameters according to the genetic algorithm so that the corresponding prediction level is consistent with the user's current demand level, wherein the set of target process characteristic parameters satisfies the process characteristic parameter constraints. The recommended incineration parameter search module is used to set incineration parameter constraints, call the process feature prediction model, and search for a set of recommended incineration parameters according to the particle swarm optimization algorithm, so that the current incinerated material characteristic data and the predicted feature parameters of the set of recommended incineration parameters are closest to the set of target process feature parameters, wherein the set of recommended incineration parameters satisfies the incineration parameter constraints. The step of calling the incineration condition rating model and searching for a set of target process characteristic parameters using a genetic algorithm to ensure that the corresponding predicted level matches the user's current demand level includes: The feasible region of each process characteristic parameter is determined according to the constraints of the process characteristic parameters, and an initial population containing M chromosomes is randomly generated within the feasible region. For each chromosome in the initial population, the chromosome is decoded into a set of process feature parameters, and the set of process feature parameters is input into the incineration condition rating model to obtain the corresponding prediction level and the probability value corresponding to the prediction level. The corresponding first current fitness value is determined according to the probability value, wherein the first current fitness value is used to represent the difference or consistency between the prediction level and the current demand level. Based on roulette wheel selection or tournament selection, superior individuals are selected from the current population to form the parent population according to the first current fitness value of the chromosome, wherein chromosomes with higher first current fitness values ​​are more likely to be selected. Chromosomes are randomly selected from the parent population and crossover is performed to obtain crossover chromosomes. The crossover chromosomes are then mutated to obtain mutated chromosomes. A new population is constructed based on the crossover chromosome and the variant chromosome, and the iteration is repeated based on the new population until a first iteration termination condition is met, wherein the first iteration termination condition includes a first maximum number of iterations, convergence of the first current fitness value, or the first current fitness value has reached a first preset value. When the iteration terminates, the target chromosome with the highest fitness in the current population is output, and the target chromosome is decoded to obtain a set of target process feature parameters; The step of calling the process feature prediction model and searching for a set of recommended incineration parameters using the particle swarm optimization algorithm, so that the predicted feature parameters of the current incinerated material characteristics and the set of recommended incineration parameters are closest to the set of target process feature parameters, includes: The search space boundary for particles is set according to the incineration parameter constraints, and a group of particles is randomly initialized within the corresponding search space. Each particle represents a set of candidate combinations of incineration parameters, and each particle is assigned a random initial position x. 0 and initial velocity v 0 ; Update formula based on location and speed update formula Iterate for each particle separately, where x k+1 For the (k+1)th iteration point, x k For the k-th iteration point, v K v represents the velocity in the k-th iteration. k+1 denoted as , where is the velocity in the (k+1)th iteration, w is the inertia weight, c1 is the cognitive factor, r1 is a random number between 0 and 1, pBest is the best position found during the search process, c2 is the social learning factor, r2 is a random number between 0 and 1, gBest is the best position found globally, rand is the random number function, and δ is the perturbation amplitude. During the iteration process, for each particle in each iteration, the current position x of the current particle is... k The current incinerator characteristics data are input into the process feature prediction model to predict intermediate process characteristic parameters. ; Calculate the intermediate process characteristic parameters and the set of target process characteristic parameters y target The corresponding second current fitness value F(x) k ), and based on the second current fitness value F(x) k The individual's historical best fitness value F(pBest) and the global historical best fitness value F(gBest) are updated, wherein the second current fitness value F(x) is updated. k The mean square error or Euclidean distance is negative. Determine whether the second iteration termination condition is met. If so, the combination of incineration parameters represented by gBest is used as a set of recommended incineration parameters. The second iteration termination condition includes the second maximum number of iterations, the convergence of the global historical optimal fitness value F(gBest) corresponding to the global historical optimal solution gBest, or the global historical optimal fitness value F(gBest) has reached the second preset value.

7. A computer device, characterized in that, The method includes a memory and a processor, wherein the memory stores a computer program that can run on the processor, and the processor executes the computer program to implement the method of any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer program is stored that can be loaded by a processor and execute the method of any one of claims 1 to 5.

Images