A boiler combustion optimization embodied intelligent prediction and parameter setting method and system

Abstract

This invention discloses an embodied intelligent prediction and parameter tuning method and system for boiler combustion optimization, relating to the field of boiler combustion optimization control technology in thermal power plants. The method includes: acquiring and preprocessing the operating condition data of the target boiler to construct state information; based on the state information, establishing a first prediction model for predicting nitrogen oxide emission concentration and a second prediction model for predicting boiler thermal efficiency; constructing an objective function for optimizing the target boiler combustion; iteratively optimizing the boiler operating parameters using an improved embodied intelligent gray wolf optimization algorithm to obtain optimal operating parameters that satisfy constraints and optimize the objective function; applying the optimal operating parameters to the target boiler operation, and updating the prediction model and / or optimization algorithm based on feedback data after operation. This invention enables the perception, prediction, and parameter optimization of boiler combustion state within an embodied intelligent framework, thereby improving boiler thermal efficiency and reducing pollutant emissions.

Description

Technical Field

[0001] This invention relates to the field of boiler combustion optimization control technology in thermal power plants, and in particular to a method and system for intelligent prediction and parameter tuning of boiler combustion optimization. Background Technology

[0002] Currently, thermal power generation remains the primary mode of electricity supply in my country, with coal being consumed in large quantities as the main heat source for power plant boilers. The combustion process in these boilers produces polluting gases, including nitrogen oxides, which have adverse environmental impacts. Therefore, improving the combustion thermal efficiency of circulating fluidized bed boilers in power plants and reducing their pollutant emission concentrations, while ensuring the safe and stable operation of the units, has become an urgent problem for thermal power generation companies to solve.

[0003] To improve the combustion thermal efficiency of circulating fluidized bed boilers in power plants and reduce their nitrogen oxide emission concentration, it is necessary to predict the boiler thermal efficiency and nitrogen oxide emission levels, and to adjust and optimize the key parameters during boiler operation based on the prediction results. This will provide a basis for thermal power plant managers to formulate and adjust boiler operation plans, thereby ensuring the safe and efficient operation of circulating fluidized bed boilers.

[0004] However, existing methods for combustion prediction and parameter tuning in circulating fluidized bed boilers often fail to accurately reflect the actual combustion state of the boiler under different operating conditions, and parameter adjustments mainly rely on experience. Existing technologies suffer from a disconnect between prediction and control, and insufficient adaptability to changes in operating conditions, thus failing to meet the needs of users such as thermal power plants. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention provides a method and system for embodied intelligent prediction and parameter tuning of boiler combustion optimization. Under the framework of embodied intelligence, it can realize the perception, prediction and parameter optimization of boiler combustion status, thereby improving boiler thermal efficiency and reducing pollutant emissions.

[0006] To achieve the above objectives, the present invention provides a method for intelligent prediction and parameter tuning of boiler combustion optimization, comprising: Acquire operating data of the target boiler during its operation phase; the operating data includes boiler load data, nitrogen oxide emission concentration, and boiler thermal efficiency. The operating data is preprocessed to construct state information characterizing the boiler combustion conditions; Based on the state information, a first prediction model for predicting nitrogen oxide emission concentration and a second prediction model for predicting boiler thermal efficiency are established using a conditional generative neural network. Construct an objective function for combustion optimization of the target boiler; the objective function takes the prediction result of the first prediction model as the optimization objective and the prediction result of the second prediction model as the constraint, or takes the prediction result of the second prediction model as the optimization objective and the prediction result of the first prediction model as the constraint. An improved embodied intelligent gray wolf optimization algorithm is used to iteratively optimize the boiler operating parameters of the target boiler to obtain the optimal operating parameters that satisfy the constraints and make the objective function optimal. The improved embodied intelligent gray wolf optimization algorithm endows each search individual with environmental perception capabilities to adaptively adjust the position update strategy according to changes in the search environment, and integrates teaching and learning mechanisms during the iteration process to maintain population diversity. The optimal operating parameters are applied to the operation of the target boiler, and the first prediction model, the second prediction model, and / or the improved embodied intelligent gray wolf optimization algorithm are updated based on the feedback data after operation.

[0007] Optionally, the conditional generative neural network is a conditional generative adversarial network, comprising a generator and a discriminator; The generator outputs predicted values ​​of nitrogen oxide emission concentration or boiler thermal efficiency based on the state information; the discriminator is used to distinguish the predicted values ​​from the corresponding actual values.

[0008] Optionally, the first prediction model or the second prediction model is trained in the following manner: Based on the state information, a conditional input vector and the nitrogen oxide emission concentration or boiler thermal efficiency corresponding to the conditional input vector are constructed as the actual output, forming a conditional sample pair. The conditional input vector is input into the generator to generate a prediction output; The conditional input vector is combined with the true output, and the conditional input vector is combined with the predicted output. Both are then input into the discriminator for adversarial training. By alternately optimizing the objective functions of the generator and the discriminator, the generator learns the mapping relationship between the state information and the optimization objective. After the training reaches the preset convergence condition, the parameters of the generator are fixed and used as the corresponding first prediction model or second prediction model.

[0009] Optionally, each search entity can be endowed with environmental awareness, including: For each search individual, the corresponding environmental perception quantity is determined by an exponentially weighted moving average based on the change in the objective function between the current iteration and the previous iteration; the environmental perception quantity is used to characterize the intensity of dynamic changes in the environment in which the corresponding search individual is located.

[0010] Optionally, the location update strategy can be adaptively adjusted according to changes in the search environment, including: For each individual being searched, the current search stage is determined based on the rate of change of the corresponding environmental perception quantity, and a location update strategy corresponding to the current search stage is executed to update the location; the location update strategy includes at least one of a circling strategy, a pursuit strategy, and a hovering strategy.

[0011] Optionally, teaching and learning mechanisms can be integrated into the iterative process to maintain population diversity, including: When the population diversity falls below a preset threshold, a teaching and learning mechanism is triggered; the teaching and learning mechanism includes a teaching phase and a learning phase. During the teaching phase, the search individual in the current iterative population that satisfies the constraints and has the optimal objective function value is identified as the teacher individual, and the teacher individual guides other search individuals to update their positions. During the learning phase, several learning operations are performed: each learning operation randomly selects two search individuals for pairing, and the inferior search individual in the pair learns from the superior search individual to update its position.

[0012] Optionally, when the prediction result of the second prediction model is used as the optimization objective and the prediction result of the first prediction model is used as the constraint, the objective function is to maximize the predicted value of boiler thermal efficiency, and the constraint is that the nitrogen oxide emission concentration does not exceed the preset emission limit.

[0013] Optionally, when the prediction result of the first prediction model is used as the optimization objective and the prediction result of the second prediction model is used as the constraint, the objective function is to minimize the predicted value of nitrogen oxide emission concentration, and the constraint is that the boiler thermal efficiency is not lower than the preset thermal efficiency value.

[0014] Optionally, when updating the first prediction model, the second prediction model, and / or the improved embodied intelligent gray wolf optimization algorithm based on the feedback data after the operation, the update is triggered after the feedback data accumulates to a preset amount, or when the error between the model prediction value and the actual value corresponding to the feedback data exceeds a preset error threshold.

[0015] This invention also provides a boiler combustion optimization-integrated intelligent prediction and parameter tuning system, comprising: The sensing module is used for: Acquire operating data of the target boiler during its operation phase; the operating data includes boiler load data, nitrogen oxide emission concentration, and boiler thermal efficiency. The operating data is preprocessed to construct state information characterizing the boiler combustion conditions; The learning module is used to establish a first prediction model for predicting nitrogen oxide emission concentration and a second prediction model for predicting boiler thermal efficiency based on the state information using a conditional generative neural network. The decision optimization module is used for: Construct an objective function for combustion optimization of the target boiler; the objective function takes the prediction result of the first prediction model as the optimization objective and the prediction result of the second prediction model as the constraint, or takes the prediction result of the second prediction model as the optimization objective and the prediction result of the first prediction model as the constraint. An improved embodied intelligent gray wolf optimization algorithm is used to iteratively optimize the boiler operating parameters of the target boiler to obtain the optimal operating parameters that satisfy the constraints and make the objective function optimal. The improved embodied intelligent gray wolf optimization algorithm endows each search individual with environmental perception capabilities to adaptively adjust the position update strategy according to changes in the search environment, and integrates teaching and learning mechanisms during the iteration process to maintain population diversity. The feedback update module is used to apply the optimal operating parameters to the operation of the target boiler, and to update the first prediction model, the second prediction model, and / or the improved embodied intelligent gray wolf optimization algorithm based on the feedback data after operation.

[0016] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects: The embodied intelligent prediction and parameter tuning method for boiler combustion optimization provided by this invention standardizes and extracts features from the operating data of the target boiler during its operating phase, constructing state information that accurately and completely characterizes the actual combustion conditions of the boiler. Based on this, two independent prediction models are established, corresponding to nitrogen oxide emission concentration and boiler thermal efficiency, respectively. This accurately depicts the mapping relationship between boiler operating parameters and combustion optimization targets, changing the existing technology's reliance on manual experience to judge the boiler combustion state and adjust operating parameters. It effectively solves the problem that existing methods cannot accurately reflect the actual combustion state of the boiler under different operating conditions, providing accurate and reliable quantitative basis for boiler combustion optimization.

[0017] Furthermore, based on the outputs of the two prediction models, an optimization objective function that balances boiler combustion efficiency improvement and pollutant emission control was constructed. The output of one prediction model was used as the core optimization objective, and the output of the other prediction model was used as a rigid constraint. At the same time, an improved embodied intelligent gray wolf optimization algorithm was used to iteratively optimize the boiler operating parameters, allowing the prediction results of the combustion state to directly serve the tuning and optimization of the operating parameters. This achieved a deep integration of combustion state prediction and parameter optimization control, effectively solving the problem of the disconnect between prediction and control in the existing technology. It ensured that the optimized operating parameters not only conformed to the actual operating conditions of the boiler, but also met the dual engineering requirements of efficiency improvement and emission control.

[0018] This invention constructs a complete embodied intelligent closed-loop optimization framework of "perception-learning-decision-feedback". After applying the optimized operating parameters to the actual operation of the target boiler, the prediction model and optimization algorithm can be continuously iterated and updated based on the feedback data collected during operation. This enables the entire optimization system to adaptively adjust to changes in boiler operating conditions and fluctuations in equipment operating status, effectively solving the problem of insufficient adaptability to changes in operating conditions in existing technologies. It can achieve continuous optimization and stable control of the combustion process throughout the entire boiler operating cycle, ensuring that the boiler is in a safe, efficient, and low-emission operating state for a long time. Attached Figure Description

[0019] The above and other objects, features and advantages of the present invention will become more apparent from the more detailed description of exemplary embodiments of the invention in conjunction with the accompanying drawings, wherein the same reference numerals generally represent the same parts.

[0020] Figure 1 This is a schematic diagram of the method flow of the boiler combustion optimization embodied intelligent prediction and parameter tuning method shown in an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the framework of the boiler combustion optimization embodied intelligent prediction and parameter tuning method according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the fitting curve of the second prediction model shown in an embodiment of the present invention; Figure 4 This is a schematic diagram of the fitting curve of the first prediction model shown in an embodiment of the present invention; Figure 5 This is a schematic diagram showing the comparison of fitting curves before and after optimization of the thermal efficiency of a circulating fluidized bed boiler according to an embodiment of the present invention. Figure 6 This is a schematic diagram showing the comparison of nitrogen oxides before and after optimization in a circulating fluidized bed boiler according to an embodiment of the present invention. Figure 7This is a schematic diagram of the module structure of the boiler combustion optimization-embedded intelligent prediction and parameter tuning system according to an embodiment of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Please see Figure 1 and Figure 2 , Figure 1 A flowchart illustrating the method for intelligent prediction and parameter tuning for boiler combustion optimization. Figure 2 A schematic diagram of the framework for a personalized intelligent prediction and parameter tuning method for boiler combustion optimization.

[0023] Boiler combustion optimization is based on intelligent prediction and parameter tuning methods, including: S101: Obtain the operating condition data of the target boiler during the operation phase.

[0024] The operating data includes boiler load data, nitrogen oxide emission concentration, and boiler thermal efficiency.

[0025] In this application, the target boiler can be a commercially operational circulating fluidized bed boiler, whose operating data can accurately reflect the actual combustion status of the boiler. Operating data should include at least boiler load data, nitrogen oxide emission concentration, and boiler thermal efficiency. Among these, boiler load data is a key parameter reflecting the boiler's output level, nitrogen oxide emission concentration is a core indicator for environmental assessment, and boiler thermal efficiency directly reflects energy conversion efficiency.

[0026] In practice, the above data can be collected in real time through the boiler's distributed control system (DCS). The DCS system continuously records boiler operating parameters at a fixed sampling period, for example, it can be set to collect a set of data every 2 seconds, thereby accumulating a large number of raw samples over a long period of time. Taking a circulating fluidized bed boiler with a rated capacity of 330MW as an example, tens of thousands of data points can be obtained after continuous operation for a period of time, such as 28,800 samples. Considering that the boiler's operating parameters fluctuate little under stable operating conditions and the differences between adjacent samples are limited, directly using all samples to train the model may lead to insufficient model generalization ability. Therefore, the raw data can be resampled, for example, selecting one sample every 10 samples. This preserves the trend of operating condition changes and expands the differences between samples, ultimately obtaining approximately 2,880 valid data samples.

[0027] It should be noted that there are 26 characteristics that affect the thermal efficiency and nitrogen oxide emission concentration of circulating fluidized bed boilers, including boiler load, coal feed rate, primary air velocity, secondary air velocity, and oxygen concentration in flue gas. Among these, boiler load is the most important and has the greatest impact.

[0028] S102: Preprocess the operating data to construct state information characterizing the boiler combustion conditions.

[0029] Since data collected from industrial sites often contains noise, outliers, and differences in physical dimensions, directly using it for model training may lead to decreased prediction accuracy or difficulty in convergence. Therefore, the collected operating data is preprocessed to eliminate noise and outliers that may exist in the original data, unify the dimensions of different features, and extract key information that can effectively characterize the boiler combustion conditions, thereby constructing the state information required for subsequent prediction models.

[0030] Specifically, the preprocessing process includes data cleaning, normalization, and feature extraction. Data cleaning removes invalid data and outliers significantly deviating from the normal range caused by sensor malfunctions or communication anomalies during data acquisition, ensuring the reliability of the input data. The cleaned data is then normalized to map features of different dimensions (such as boiler load, coal feed rate, and wind speed) to the same numerical range, eliminating the adverse effects of dimensional differences on model training. Based on this, feature extraction is performed to select features closely related to boiler combustion performance from the original operating data, such as boiler load, coal feed rate, primary wind speed, secondary wind speed, and oxygen concentration in the flue gas. These features collectively constitute state information reflecting the boiler combustion conditions. This state information will serve as input to the subsequent conditional generative neural network to establish the mapping relationship between boiler operating parameters and optimization objectives.

[0031] S103: Based on state information, a first prediction model for predicting nitrogen oxide emission concentration and a second prediction model for predicting boiler thermal efficiency are established using a conditional generative neural network.

[0032] Among them, the conditional generative neural network is a conditional generative adversarial network, which includes a generator and a discriminator. The generator outputs the predicted value of nitrogen oxide emission concentration or boiler thermal efficiency based on the state information. The discriminator is used to distinguish the predicted value from the corresponding real value.

[0033] The first or second prediction model is obtained through training in the following manner: Based on state information, a conditional input vector is constructed, and the nitrogen oxide emission concentration or boiler thermal efficiency corresponding to the conditional input vector is used as the real output to form a conditional sample pair. Input the conditional input vector into the generator to generate the prediction output; The conditional input vector is combined with the true output, and the conditional input vector is combined with the predicted output. Both are then input into the discriminator for adversarial training. By alternately optimizing the objective functions of the generator and the discriminator, the generator learns the mapping relationship between the state information and the optimization objective. After the training reaches the preset convergence condition, the generator parameters are fixed and used as the corresponding first or second prediction model.

[0034] In the application, based on the preprocessed state information, a conditional generative neural network is used to establish a first prediction model for predicting nitrogen oxide emission concentration and a second prediction model for predicting boiler thermal efficiency. Specifically, this conditional generative neural network employs a conditional generative adversarial network structure. Through an adversarial learning mechanism between the generator and discriminator, it can effectively capture the complex nonlinear mapping relationship between boiler operating conditions and combustion performance indicators, thereby obtaining a high-precision prediction model.

[0035] Conditional Generative Adversarial Networks (GANs) consist of two core components: a generator and a discriminator. The generator takes state information as input and outputs a predicted value for the corresponding nitrogen oxide emission concentration or boiler thermal efficiency. The discriminator distinguishes between the generator's predicted value and the actual collected values. Through alternating adversarial training, the generator continuously approximates the distribution of real data, ultimately learning the mapping relationship from operating parameters to combustion performance indicators.

[0036] During model training, the conditional input vector is first fed into the generator, which then generates the corresponding predicted output based on this input. Subsequently, the conditional input vector is combined with both the true and predicted outputs to form true sample pairs and generated sample pairs. Both types of sample pairs are then input into the discriminator for discrimination. The discriminator outputs a discrimination result based on the input sample pair, indicating whether the sample pair comes from real data or is predicted data generated by the generator.

[0037] The training process employs an adversarial training approach with alternating optimization. In each iteration, the generator parameters are first fixed, while the discriminator parameters are updated to maximize the discriminator's ability to distinguish between real and generated sample pairs; the objective function is the adversarial loss. Subsequently, the discriminator parameters are fixed again, and the generator parameters are updated to generate predictive outputs that can deceive the discriminator as much as possible, while simultaneously ensuring that the predicted outputs numerically approximate the real outputs. The generator's objective function comprises two parts: first, the adversarial loss, which prevents the discriminator from classifying generated sample pairs as real samples; and second, the prediction consistency constraint, which minimizes the error between the generated and real outputs, guiding the generator to focus not only on distribution fitting but also on prediction accuracy. Through the joint optimization of the adversarial loss and the prediction consistency constraint, the generator can learn the mapping relationship between state information and nitrogen oxide emission concentration or boiler thermal efficiency.

[0038] The aforementioned alternating optimization process is repeated, allowing the generator and discriminator to co-evolve in a game-like process. As training iterations progress, the distribution of the generator's predicted output gradually approximates the distribution of the real data, and the discriminator's ability to distinguish between true and false data continuously improves. Training stops when the model training reaches a preset convergence condition. The preset convergence condition can be that the loss function tends to stabilize, the preset maximum number of iterations is reached, or other preset training termination criteria are met.

[0039] Once the training reaches the preset convergence condition, the generator's parameters are fixed, and it no longer participates in subsequent adversarial training. At this point, the generator can be used as the corresponding prediction model. For the nitrogen oxide emission concentration prediction task, the generator with fixed parameters after training becomes the first prediction model; for the boiler thermal efficiency prediction task, the generator with fixed parameters after training becomes the second prediction model. In subsequent applications, simply input the real-time state information as a conditional input vector into the corresponding generator to quickly output the corresponding predicted values ​​for nitrogen oxide emission concentration or boiler thermal efficiency.

[0040] During model training, the first step is to construct conditional samples. Specifically, the process of constructing conditional samples includes the following steps: The operating condition parameters and control parameters of the circulating fluidized bed boiler during the operational phase are obtained, and the operating condition parameters and control parameters are combined to form a conditional input vector c; wherein, the conditional input vector c is used to characterize the combustion condition characteristics of the boiler under a specific operating state. Obtain the output data y corresponding to the conditional input vector c, wherein the output data y is the nitrogen oxide emission concentration or the boiler thermal efficiency; The conditional input vector c and the output data y are normalized to construct a conditional sample pair (c,y), which is then used as the data basis for training the subsequent conditional generative adversarial neural network model.

[0041] The construction of a conditional generative adversarial neural network structure includes the following steps: Building generator networks The generator takes a condition variable c and a random noise vector z as input and outputs the corresponding prediction result. The calculation relationship is expressed as follows: Constructing a discriminator network The discriminator uses the condition variable c and the actual output y or the generated output. As input, and outputting the corresponding discrimination result, it is used to characterize the difference between the predicted result and the real sample. The calculation relationship is expressed as follows: The training of a conditional generative adversarial neural network specifically includes the following steps: The process of generating prediction results and performing discrimination calculations involves using a generator to produce prediction outputs given the condition variable c. And the real sample pair (c,y) and the generated sample pair The corresponding inputs are fed into the discriminator to obtain the corresponding discriminant output results; The steps for constructing the discriminator's objective function include building the discriminator's adversarial loss function, enabling the discriminator to distinguish between the real output and the generated output. The objective function is expressed as follows: The steps for constructing the generator's objective function include building the generator's joint loss function, introducing a prediction consistency constraint on top of the adversarial loss to reduce the error between the generated output and the true output. The objective function is expressed as follows: Where λ is the consistency constraint weight coefficient; The steps for constructing the overall objective function are as follows: The overall objective function can be expressed as: The adversarial training iterative steps involve repeating the above operations to continuously optimize the generator and discriminator through alternating training until the model training meets the preset convergence conditions. The prediction model determination step involves fixing the generator parameters after the conditional generative adversarial neural network model has been trained, and using the generator as the prediction model for the corresponding combustion index.

[0042] S104: Construct the objective function for combustion optimization of the target boiler.

[0043] The objective function can be either the prediction result of the first prediction model as the optimization objective and the prediction result of the second prediction model as the constraint, or the prediction result of the second prediction model as the optimization objective and the prediction result of the first prediction model as the constraint.

[0044] When the prediction results of the second prediction model are used as the optimization objective and the prediction results of the first prediction model are used as the constraint, the objective function is to maximize the predicted value of boiler thermal efficiency, and the constraint is that the nitrogen oxide emission concentration does not exceed the preset emission limit.

[0045] When the predicted boiler combustion thermal efficiency is used as the objective function, the objective function is expressed as: in, The vector of boiler combustion parameters to be optimized. To obtain the predicted thermal efficiency under the corresponding operating conditions output by the thermal efficiency prediction model, boundary constraints are applied to each optimization parameter: And introduce nitrogen oxide emission concentration constraints: in, Given the preset emission limits, the subsequently improved embodied intelligent gray wolf optimization algorithm optimizes the objective function under the above parameter constraints and emission constraints to obtain the boiler combustion parameter tuning result that meets the emission requirements and has the best thermal efficiency.

[0046] When the prediction results of the first prediction model are used as the optimization objective and the prediction results of the second prediction model are used as the constraint, the objective function is to minimize the predicted value of nitrogen oxide emission concentration, and the constraint is that the boiler thermal efficiency is not lower than the preset thermal efficiency value.

[0047] When the objective function is the predicted value of nitrogen oxide emissions from boiler combustion, the objective function is expressed as: in, The vector of boiler combustion parameters to be optimized. The predicted nitrogen oxide concentration values ​​under the corresponding operating conditions are output by the nitrogen oxide emission concentration prediction model, while boundary constraints are applied to each optimization parameter: And a thermal efficiency value constraint is introduced: in, Given a preset thermal efficiency value, the subsequently improved embodied intelligent gray wolf optimization algorithm iteratively searches the objective function under the above parameter constraints and emission constraints to obtain the boiler combustion parameter tuning result that can improve the boiler combustion thermal efficiency and achieve the optimal nitrogen oxide emission concentration.

[0048] In application, the objective function can be set in two ways. One approach prioritizes improving boiler thermal efficiency. In this case, the prediction results of the second prediction model (thermal efficiency prediction model) are used as the optimization objective to maximize thermal efficiency. Simultaneously, the prediction results of the first prediction model (nitrogen oxide emission concentration prediction model) are used as constraints to ensure that the nitrogen oxide emission concentration generated by the optimized operating parameters does not exceed the emission limits set by environmental regulations or the company itself. The other approach prioritizes reducing pollutant emissions. In this case, the prediction results of the first prediction model are used as the optimization objective to minimize nitrogen oxide emission concentration. Simultaneously, the prediction results of the second prediction model are used as constraints to ensure that the optimized operating parameters maintain boiler thermal efficiency at or above a preset benchmark value, thereby guaranteeing the economical operation of the unit.

[0049] For the mode with thermal efficiency as the optimization objective, the objective function is to maximize the predicted thermal efficiency value output by the second prediction model. Constraints include that the predicted nitrogen oxide emission concentration output by the first prediction model does not exceed a preset emission limit. Furthermore, the boiler operating parameters to be optimized must also meet their respective value range boundaries. For the mode with nitrogen oxide emission as the optimization objective, the objective function is to minimize the predicted nitrogen oxide emission concentration output by the first prediction model. Constraints include that the predicted thermal efficiency value output by the second prediction model is not lower than a preset thermal efficiency value, and each operating parameter must also meet its value range boundaries.

[0050] Both optimization modes seek the optimal value of the objective function while satisfying constraints. Preset emission limits and thermal efficiency values ​​can be set according to actual operating requirements, environmental standards, or enterprise performance indicators, while the boundary of the value range is determined based on the boiler equipment's design parameters and safe operating range. By constructing such an objective function, the multi-objective problem of combustion optimization can be transformed into a single-objective constrained optimization problem, providing a clear evaluation standard for subsequent parameter tuning using optimization algorithms. This flexible objective function construction method allows this approach to adapt to optimization needs in different scenarios, suitable for both pursuing energy efficiency improvements and meeting environmental requirements, thus possessing broad applicability.

[0051] S105: An improved embodied intelligent gray wolf optimization algorithm is used to iteratively optimize the boiler operating parameters of the target boiler in order to obtain the optimal operating parameters that satisfy the constraints and make the objective function optimal.

[0052] Among them, the improved embodied intelligence gray wolf optimization algorithm endows each search individual with environmental perception capabilities, so as to adaptively adjust the position update strategy according to changes in the search environment, and integrates teaching and learning mechanisms in the iterative process to maintain population diversity.

[0053] In the application, based on the aforementioned objective function, an improved embodied intelligence gray wolf optimization algorithm is used to iteratively optimize the operating parameters of the target boiler to obtain the optimal operating parameters that satisfy the constraints and maximize the objective function. This algorithm introduces the concept of embodied intelligence into the traditional gray wolf optimization algorithm, endowing each search individual with environmental awareness, enabling it to adaptively adjust its position update strategy according to changes in the search environment. Simultaneously, it integrates teaching and learning mechanisms during the iteration process to maintain population diversity, thereby effectively balancing global search and local exploitation capabilities, avoiding getting trapped in local optima, and improving the stability and accuracy of the optimization process.

[0054] Specifically, the improved embodied intelligence gray wolf optimization algorithm first endows each search individual with environmental awareness, including: For each search individual, the corresponding environmental perception quantity is determined by an exponentially weighted moving average based on the change in the objective function between the current iteration and the previous iteration; the environmental perception quantity is used to characterize the intensity of dynamic changes in the environment in which the corresponding search individual is located.

[0055] In the application, the individual gray wolf (the search individual) is treated as an embodied individual and given the ability to perceive the environment. An exponentially weighted moving average mechanism is introduced to model the dynamic changes of the search environment. The specific calculation formula is as follows: in, In the t-th iteration, the first... Environmental perception of each individual searcher; It is a smoothing factor; This represents the change in the objective function of the t-th iteration relative to the (t-1)-th iteration, used to characterize the intensity of environmental change.

[0056] Based on the obtained environmental perception data, the position update strategy is adaptively adjusted according to changes in the search environment, specifically including: For each individual being searched, the current search stage is determined based on the rate of change of the corresponding environmental perception quantity, and a location update strategy corresponding to the current search stage is executed to update the location; the location update strategy includes at least one of the following: a circling strategy, a pursuit strategy, and a hovering strategy.

[0057] Specifically, when environmental perception indicates that the population is in a relatively stable search state, a surrounding strategy is triggered, causing search individuals to move closer to the center of dominant individuals in the current population to enhance local exploitation capabilities. When environmental perception indicates that the population is approaching a potential optimal region, a chasing strategy is triggered, introducing a random perturbation term while adjusting the step size to enhance the algorithm's ability to escape local optima and expand the search range. When environmental perception indicates that the population search is stagnating, a hovering strategy is triggered, constructing a local perturbation radius in the current search region to cause search individuals to search in multiple directions around a representative solution, thereby restoring population vitality and avoiding premature convergence.

[0058] When environmental perception indicates that the population is in a relatively stable state, a circling strategy is triggered, setting a step size factor α to control the extent to which individual gray wolves move towards the center position. First, the center position of the three alpha wolves is calculated: Then update the current position according to the update rules. Get a new position : in, These represent the current position vectors of the three alpha wolves; Indicates surrounding the central position; This represents the current position vector of the individual to be updated; This represents the updated position vector.

[0059] When the environmental perception indicates that the algorithm is approaching a potential optimal region, a pursuit strategy is triggered. A random perturbation term is introduced while adjusting the step size factor α to enhance the algorithm's ability to escape local optima. The specific update formula is as follows: Where N(0,I) represents a Gaussian random vector with a mean of 0 and a covariance of the identity matrix; range represents the range vector of values ​​for the search variable.

[0060] When environmental perception indicates that the population search has stagnated, a hovering strategy is triggered. This strategy introduces a random direction vector to allow individuals to explore uniformly within their local neighborhood, thereby restoring population diversity. The specific formula is as follows: Where D represents the dimension of the optimization variable; This represents a unit direction vector randomly generated in D-dimensional space; The norm representing the range scale is used to generate a local search radius that fits the variable scale.

[0061] During the iteration process, the algorithm also incorporates teaching and learning mechanisms to maintain population diversity, specifically including: When the population diversity falls below a preset threshold, a teaching and learning mechanism is triggered. This mechanism includes a teaching phase and a learning phase. In the teaching phase, the search individual in the current iteration population that satisfies the constraints and has the optimal objective function value is designated as the teacher individual, which guides other search individuals to update their positions. In the learning phase, several learning operations are performed: each learning operation randomly selects two search individuals for pairing, and the weaker search individual learns from the stronger one to update its position. The number of learning operations can be flexibly set according to the actual situation.

[0062] During the teaching phase, the best individual in the current population is selected as the teacher individual, and the learning individuals are guided to update according to the teaching update method. The specific formula is as follows: in, Represents the individual teacher's position vector; Indicates the first The position vectors of each search entity; Indicates the first The updated position vector of each searched individual; Represents the location vector of the population mean; This is the teaching update coefficient; This is the scaling factor; is a random vector; ⊙ represents Hadamard element-wise multiplication; TF is the teaching factor.

[0063] During the learning phase, students will learn from teachers and interact with each other; two students will be randomly selected. and And p is not equal to q; if p is better than q, then the learning update rule is as follows: In the formula, Indicates the first The position vector of each search entity Indicates the first The position vector of each search entity Represents an individual fitness value, Represents an individual fitness value, Represents a uniformly random number between 0 and 1. Indicates the first The updated position vector of each searched individual.

[0064] It is understandable that when two randomly selected search individuals both meet the constraints and have the same objective function value, meaning that their quality is comparable and there is no clear inferior or superior one, no position update is performed, and the original position remains unchanged.

[0065] Finally, the update results from the teaching and learning phases are used to update the population individuals in a unified manner, and the construction of a new generation of population is completed to enter the next round of iterative search.

[0066] The integration of environmental perception, adaptive strategy switching, and the teaching and learning mechanism enables the improved embodied intelligence gray wolf optimization algorithm to achieve a dynamic balance between global search capability and local fine-grained search capability. In the early stages of iteration, the algorithm maintains strong exploration capabilities through environmental perception and strategy switching; in the later stages, the teaching and learning mechanism guides the population to gradually approach the optimal solution while avoiding getting trapped in local optima. The algorithm iteratively executes the above steps until a preset termination condition is met, such as reaching the maximum number of iterations or the objective function value converging. Finally, it outputs the optimal operating parameters that satisfy the constraints and optimize the objective function, serving as the parameter tuning scheme for boiler combustion optimization. The operating parameters obtained through this improved algorithm can maximize thermal efficiency or minimize nitrogen oxide emissions while meeting emission limits or thermal efficiency requirements, thus providing reliable decision support for the safe, efficient, and low-pollution operation of boilers.

[0067] S106: Apply the optimal operating parameters to the target boiler operation, and update the first prediction model, the second prediction model, and / or the improved embodied intelligent gray wolf optimization algorithm based on the feedback data after operation.

[0068] The optimized operating parameters are then applied to the actual operation of the target boiler. These optimal operating parameters are the combination of parameters that satisfies the preset constraints and optimizes the objective function. Applying them to the boiler enables the combustion process to achieve the desired balance between thermal efficiency and nitrogen oxide emissions.

[0069] After the boiler operates according to optimal parameters, it is necessary to continuously collect feedback data during its operation. This feedback data includes, at least, boiler load data, as well as the actual nitrogen oxide emission concentration and boiler thermal efficiency at the corresponding time, similar to the aforementioned operating condition data. This feedback data accurately reflects the actual combustion performance of the boiler under optimal parameters and serves as the basis for evaluating the optimization effect and making subsequent updates.

[0070] Based on the collected feedback data, the first prediction model, the second prediction model, and / or the improved embodied intelligent gray wolf optimization algorithm are updated. For the prediction model, the newly acquired feedback data can be used as new samples for incremental training or fine-tuning, enabling the model to adapt to changes in boiler operating conditions and maintain prediction accuracy. For the improved embodied intelligent gray wolf optimization algorithm, relevant parameters in the algorithm (such as environmental perception smoothing factor, teaching factor, etc.) can be adaptively adjusted according to the actual optimization effect reflected by the feedback data, making the algorithm's search behavior more in line with the characteristics of the current boiler.

[0071] Through this operation-feedback-based update mechanism, the predictive model and optimization algorithm can dynamically adjust as the boiler's operating status changes, forming a closed-loop continuous optimization process. When the boiler's combustion characteristics deviate due to equipment aging, changes in fuel characteristics, or alterations in environmental conditions, the feedback data can promptly capture these changes and update the model and algorithm to adapt to the new operating conditions, thereby ensuring the stability and reliability of the optimization effect during long-term operation. This closed-loop feedback mechanism enables this invention not only to achieve single-stage optimization but also to achieve continuous optimization and stable control of the boiler combustion process, effectively solving the problems of disconnect between prediction and control and insufficient adaptability to changes in operating conditions in existing technologies.

[0072] As an exemplary implementation, when updating the first prediction model, the second prediction model, and / or the improved embodied intelligent gray wolf optimization algorithm based on the feedback data after operation, it is triggered after the feedback data accumulates to a preset amount, or when the error between the model prediction value and the actual value corresponding to the feedback data exceeds a preset error threshold.

[0073] After applying the optimal operating parameters to the target boiler and collecting feedback data, a reasonable update trigger mechanism is needed to continuously optimize the prediction model and optimization algorithm. Specifically, when the number of newly collected feedback data samples reaches a preset number, it indicates that sufficient new information is available for adjusting the model or algorithm. At this point, an update operation is triggered to incrementally train the first and / or second prediction models, or to adaptively adjust the control parameters of the improved embodied intelligent gray wolf optimization algorithm. On the other hand, if during operation the deviation between the model's predicted value and the actual collected value exceeds a preset error threshold, it indicates that the current model can no longer accurately reflect the actual combustion characteristics of the boiler. Even if the feedback data has not yet accumulated to the preset number, an update is immediately triggered to avoid deviations in subsequent optimization decisions due to model inaccuracy. The preset number and preset error threshold can be set according to the actual operating characteristics and accuracy requirements of the boiler. For example, the number threshold can be set to 50 or 100 samples. The error threshold can be set to a relative error of 5% or a certain limit of the absolute error. This dual-trigger mechanism ensures the timeliness of updates while avoiding the computational overhead caused by frequent updates. This allows the prediction model and optimization algorithm to continuously track changes in boiler operating conditions at a moderate frequency, thereby maintaining their prediction accuracy and optimization performance, and achieving long-term stable optimization control of the boiler combustion process.

[0074] To better understand the above method, the technical solution of the present invention will be illustrated below through a specific application example.

[0075] First, 26 characteristics affect the thermal efficiency and nitrogen oxide emission concentration of circulating fluidized bed boilers, mainly including boiler load, coal feed rate, primary air velocity, secondary air velocity, and oxygen concentration in flue gas. A total of 2880 data samples were collected from a 330MW circulating fluidized bed boiler at different times, and divided into training and test sets at an 8:2 ratio.

[0076] Secondly, the conditional generative adversarial neural network model proposed in this invention is used to model the thermal efficiency and nitrogen oxide emission concentration of the circulating fluidized bed boiler, respectively. That is, a predictive model for the nitrogen oxide (NOx) emission concentration of the circulating fluidized bed boiler is trained, and a predictive model for the thermal efficiency of the circulating fluidized bed boiler is trained. The model parameters are set as follows: the sample size t is set to 2880, the number of conditional features x is set to 26, the input noise dimension z of the generator is set to 100, and the number of neurons in the hidden layer of the generator is [not specified]. The number of neurons in the hidden layer of the discriminator is set to 128. The number of training iterations n for the generator and discriminator is set to 128, and the number of iterations n for the generator and discriminator is set to 20. In addition, the output layer activation function of the generator is the tanh function, and the output layer activation function of the discriminator is the sigmoid function.

[0077] Finally, the fitting curves of the conditional generative adversarial neural network model proposed in this invention for real-time prediction of the thermal efficiency and nitrogen oxide emission concentration of circulating fluidized bed boilers are plotted as follows: Figure 3 and Figure 4 As shown; the fitted curves of the improved embodied intelligent gray wolf optimization algorithm proposed in this invention for optimizing the thermal efficiency and nitrogen oxide emission concentration of a circulating fluidized bed boiler are plotted as follows. Figure 5 and Figure 6 As shown.

[0078] Depend on Figure 3 and Figure 4 It can be seen that, based on the conditional generative adversarial neural network model proposed in this invention, the output value of the real-time prediction model of the circulating fluidized bed boiler combustion system (i.e., Figure 3 and Figure 4 The predicted values ​​(in the model) closely approximate the actual values, thus verifying that the method for predicting the combustion thermal efficiency and nitrogen oxide emission concentration of circulating fluidized bed boilers proposed in this invention is an effective and reliable technical solution.

[0079] Depend on Figure 5 and Figure 6 It can be seen that, based on the improved embodied intelligent gray wolf optimization algorithm proposed in this invention, after optimizing the key operating parameters of the circulating fluidized bed boiler, the thermal efficiency is improved and the emission concentration is reduced while maintaining the fluctuation characteristics of the operating conditions. This verifies that the improved embodied intelligent gray wolf optimization algorithm proposed in this invention can effectively improve the combustion thermal efficiency of the boiler and reduce the nitrogen oxide emission concentration of the circulating fluidized bed boiler without affecting the stability of boiler operation, and has good optimization effect and engineering application value.

[0080] Corresponding to the aforementioned application function implementation method embodiments, the present invention also provides a boiler combustion optimization-integrated intelligent prediction and parameter tuning system and corresponding embodiments.

[0081] Please see Figure 7 , Figure 7 A schematic diagram of the modular structure of a customized intelligent prediction and parameter tuning system for boiler combustion optimization.

[0082] The boiler combustion optimization system incorporates intelligent prediction and parameter tuning, including: Sensing module 71 is used for: Acquire operating data of the target boiler during its operation phase; operating data includes boiler load data, nitrogen oxide emission concentration, and boiler thermal efficiency; Preprocess the operating data to construct state information characterizing the boiler combustion conditions; Learning module 72 is used to build a first prediction model for predicting nitrogen oxide emission concentration and a second prediction model for predicting boiler thermal efficiency based on state information and using a conditional generative neural network. Decision optimization module 73 is used for: Construct an objective function for combustion optimization of the target boiler; the objective function takes the prediction result of the first prediction model as the optimization objective and the prediction result of the second prediction model as the constraint, or takes the prediction result of the second prediction model as the optimization objective and the prediction result of the first prediction model as the constraint. An improved embodied intelligent gray wolf optimization algorithm is used to iteratively optimize the boiler operating parameters of the target boiler to obtain the optimal operating parameters that satisfy the constraints and make the objective function optimal. The improved embodied intelligent gray wolf optimization algorithm endows each search individual with environmental perception ability to adaptively adjust the position update strategy according to changes in the search environment, and integrates teaching and learning mechanisms during the iteration process to maintain population diversity. The feedback update module 74 is used to apply the optimal operating parameters to the operation of the target boiler and update the first prediction model, the second prediction model and / or the improved embodied intelligent gray wolf optimization algorithm based on the feedback data after operation.

[0083] In one embodiment, when the first prediction model or the second prediction model is being trained, the learning module 72 performs the following operations: Based on state information, a conditional input vector is constructed, and the nitrogen oxide emission concentration or boiler thermal efficiency corresponding to the conditional input vector is used as the real output to form a conditional sample pair. Input the conditional input vector into the generator to generate the prediction output; The conditional input vector is combined with the true output, and the conditional input vector is combined with the predicted output. Both are then input into the discriminator for adversarial training. By alternately optimizing the objective functions of the generator and the discriminator, the generator learns the mapping relationship between the state information and the optimization objective. After the training reaches the preset convergence condition, the generator parameters are fixed and used as the corresponding first or second prediction model.

[0084] In one embodiment, when endowing each search individual with environmental awareness, the decision optimization module 73 is used to: For each search individual, the corresponding environmental perception quantity is determined by an exponentially weighted moving average based on the change in the objective function between the current iteration and the previous iteration; the environmental perception quantity is used to characterize the intensity of dynamic changes in the environment in which the corresponding search individual is located.

[0085] In one embodiment, when adaptively adjusting the position update strategy according to changes in the search environment, the decision optimization module 73 is used to: For each individual being searched, the current search stage is determined based on the rate of change of the corresponding environmental perception quantity, and a location update strategy corresponding to the current search stage is executed to update the location; the location update strategy includes at least one of the following: a circling strategy, a pursuit strategy, and a hovering strategy.

[0086] In one embodiment, when integrating teaching and learning mechanisms during the iteration process to maintain population diversity, the decision optimization module 73 is used to: When the population diversity falls below a preset threshold, the teaching and learning mechanism is triggered; the teaching and learning mechanism includes a teaching phase and a learning phase. During the teaching phase, the search individual in the current iterative population that satisfies the constraints and has the optimal objective function value is identified as the teacher individual, and the teacher individual guides other search individuals to update their positions. During the learning phase, several learning operations are performed: each learning operation randomly selects two search individuals for pairing, and the inferior search individual in the pair learns from the superior search individual to update its position.

[0087] Regarding the system in the above embodiments, the specific manner in which each unit module performs operations has been described in detail in the embodiments related to the method, and will not be elaborated further here.

[0088] The various embodiments of the present invention have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A method for intelligent prediction and parameter tuning of boiler combustion optimization, characterized in that, include: Acquire operating data of the target boiler during its operation phase; the operating data includes boiler load data, nitrogen oxide emission concentration, and boiler thermal efficiency. The operating data is preprocessed to construct state information characterizing the boiler combustion conditions; Based on the state information, a first prediction model for predicting nitrogen oxide emission concentration and a second prediction model for predicting boiler thermal efficiency are established using a conditional generative neural network. Construct an objective function for combustion optimization of the target boiler; the objective function takes the prediction result of the first prediction model as the optimization objective and the prediction result of the second prediction model as the constraint, or takes the prediction result of the second prediction model as the optimization objective and the prediction result of the first prediction model as the constraint. An improved embodied intelligent gray wolf optimization algorithm is used to iteratively optimize the boiler operating parameters of the target boiler to obtain the optimal operating parameters that satisfy the constraints and make the objective function optimal. The improved embodied intelligent gray wolf optimization algorithm endows each search individual with environmental perception capabilities to adaptively adjust the position update strategy according to changes in the search environment, and integrates teaching and learning mechanisms during the iteration process to maintain population diversity. The optimal operating parameters are applied to the operation of the target boiler, and the first prediction model, the second prediction model, and / or the improved embodied intelligent gray wolf optimization algorithm are updated based on the feedback data after operation.

2. The method for intelligent prediction and parameter tuning of boiler combustion optimization according to claim 1, characterized in that, The conditional generative neural network is a conditional generative adversarial network, which includes a generator and a discriminator; The generator outputs predicted values ​​of nitrogen oxide emission concentration or boiler thermal efficiency based on the state information; the discriminator is used to distinguish the predicted values ​​from the corresponding actual values.

3. The method for intelligent prediction and parameter tuning of boiler combustion optimization according to claim 2, characterized in that, The first prediction model or the second prediction model is trained in the following way: Based on the state information, a conditional input vector and the nitrogen oxide emission concentration or boiler thermal efficiency corresponding to the conditional input vector are constructed as the actual output, forming a conditional sample pair. The conditional input vector is input into the generator to generate a prediction output; The conditional input vector is combined with the true output, and the conditional input vector is combined with the predicted output. Both are then input into the discriminator for adversarial training. By alternately optimizing the objective functions of the generator and the discriminator, the generator learns the mapping relationship between the state information and the optimization objective. After the training reaches the preset convergence condition, the parameters of the generator are fixed and used as the corresponding first prediction model or second prediction model.

4. The method for intelligent prediction and parameter tuning of boiler combustion optimization according to claim 1, characterized in that, Empower each search entity with environmental awareness, including: For each search individual, the corresponding environmental perception quantity is determined by an exponentially weighted moving average based on the change in the objective function between the current iteration and the previous iteration; the environmental perception quantity is used to characterize the intensity of dynamic changes in the environment in which the corresponding search individual is located.

5. The method for intelligent prediction and parameter tuning of boiler combustion optimization according to claim 4, characterized in that, The location update strategy is adaptively adjusted based on changes in the search environment, including: For each individual being searched, the current search stage is determined based on the rate of change of the corresponding environmental perception quantity, and a location update strategy corresponding to the current search stage is executed to update the location; the location update strategy includes at least one of a circling strategy, a pursuit strategy, and a hovering strategy.

6. The method for intelligent prediction and parameter tuning of boiler combustion optimization according to claim 1, characterized in that, Integrating teaching and learning mechanisms during the iterative process to maintain population diversity includes: When the population diversity falls below a preset threshold, a teaching and learning mechanism is triggered; the teaching and learning mechanism includes a teaching phase and a learning phase. During the teaching phase, the search individual in the current iterative population that satisfies the constraints and has the optimal objective function value is identified as the teacher individual, and the teacher individual guides other search individuals to update their positions. During the learning phase, several learning operations are performed: each learning operation randomly selects two search individuals for pairing, and the inferior search individual in the pair learns from the superior search individual to update its position.

7. The method for intelligent prediction and parameter tuning of boiler combustion optimization according to claim 1, characterized in that, When the prediction result of the second prediction model is used as the optimization objective and the prediction result of the first prediction model is used as the constraint, the objective function is to maximize the predicted value of boiler thermal efficiency, and the constraint is that the nitrogen oxide emission concentration does not exceed the preset emission limit.

8. The method for intelligent prediction and parameter tuning of boiler combustion optimization according to claim 1, characterized in that, When the prediction result of the first prediction model is used as the optimization objective and the prediction result of the second prediction model is used as the constraint, the objective function is to minimize the predicted value of nitrogen oxide emission concentration, and the constraint is that the boiler thermal efficiency is not lower than the preset thermal efficiency value.

9. The method for intelligent prediction and parameter tuning of boiler combustion optimization according to claim 1, characterized in that, When updating the first prediction model, the second prediction model, and / or the improved embodied intelligent gray wolf optimization algorithm based on the feedback data after operation, it is triggered after the feedback data accumulates to a preset amount, or when the error between the model prediction value and the actual value corresponding to the feedback data exceeds a preset error threshold.

10. A boiler combustion optimization-integrated intelligent prediction and parameter tuning system, characterized in that, include: The sensing module is used for: Acquire operating data of the target boiler during its operation phase; the operating data includes boiler load data, nitrogen oxide emission concentration, and boiler thermal efficiency. The operating data is preprocessed to construct state information characterizing the boiler combustion conditions; The learning module is used to establish a first prediction model for predicting nitrogen oxide emission concentration and a second prediction model for predicting boiler thermal efficiency based on the state information using a conditional generative neural network. The decision optimization module is used for: Construct an objective function for combustion optimization of the target boiler; the objective function takes the prediction result of the first prediction model as the optimization objective and the prediction result of the second prediction model as the constraint, or takes the prediction result of the second prediction model as the optimization objective and the prediction result of the first prediction model as the constraint. An improved embodied intelligent gray wolf optimization algorithm is used to iteratively optimize the boiler operating parameters of the target boiler to obtain the optimal operating parameters that satisfy the constraints and make the objective function optimal. The improved embodied intelligent gray wolf optimization algorithm endows each search individual with environmental perception capabilities to adaptively adjust the position update strategy according to changes in the search environment, and integrates teaching and learning mechanisms during the iteration process to maintain population diversity. The feedback update module is used to apply the optimal operating parameters to the operation of the target boiler, and to update the first prediction model, the second prediction model, and / or the improved embodied intelligent gray wolf optimization algorithm based on the feedback data after operation.

Images