Self-adaptive boiler combustion control method, computer equipment and readable storage medium

Abstract

The invention provides a self-adaptive boiler combustion control method, computer equipment and a readable storage medium. The method comprises the steps that boiler operation control parameters are obtained; based on the boiler operation control parameters, a plurality of multi-dimensional particles are determined, and each multi-dimensional particle is a set of parameter adjustment results of the boiler operation control parameters; based on a particle swarm optimization algorithm and a pre-trained Bayesian model, performing iterative updating on each multi-dimensional particle until the maximum number of iterations is reached or the multi-dimensional particles are converged to the optimal position; and determining the multi-dimensional particle with the lowest defect degree in the plurality of multi-dimensional particles after iteration updating as an optimization target of the boiler operation control parameters. According to the technical scheme, multi-target automatic dynamic optimization is conducted on boiler combustion, the working efficiency of the boiler is improved, and it is guaranteed that the NOx emission amount is reduced to be within a reasonable range on the basis that the working efficiency requirement is met.

Description

[Technical Field]

[0001] This application relates to the field of thermal power control technology, and in particular to an adaptive boiler combustion control method, computer equipment, and readable storage medium. [Background Technology]

[0002] In the field of thermal power control technology, the operation optimization of large power plant pulverized coal boilers faces many challenges. Traditional combustion optimization methods usually adopt a singular adjustment approach, optimizing only local links or equipment, which leads to contradictions between different optimization objectives. For example, it is difficult to simultaneously reduce nitrogen oxide emissions and improve boiler efficiency, resulting in poor overall optimization effects.

[0003] Meanwhile, the boiler combustion process is complex and variable, involving the dynamic coupling of multiple parameters. Existing technologies rely heavily on expert experience and on-site tests, lacking scientific and quantitative guidance, making it difficult to achieve precise adjustments.

[0004] Furthermore, existing combustion optimization models, such as fuzzy control and neural network algorithms, suffer from shortcomings in accuracy and generalization ability, particularly under conditions of frequent load adjustments and drastic coal quality changes. Early optimization techniques sought the optimal operating parameter range through combustion adjustment experiments, but limited experimental conditions prevented coverage of all operating conditions, and boiler equipment performance changes over time and during unit maintenance, leading to optimization results deviating from actual needs. Manual adjustments by operators are subjective and arbitrary, making it difficult to achieve multi-objective collaborative optimization and unable to cope with real-time changes in operating conditions.

[0005] Therefore, how to achieve multi-objective dynamic optimization of boiler combustion, taking into account both efficiency improvement and emission control, while reducing reliance on human experience, has become an urgent technical problem to be solved. [Summary of the Invention]

[0006] This application provides an adaptive boiler combustion control method, a computer device, and a readable storage medium. It aims to address the technical problem in related technologies where boiler combustion parameter optimization methods struggle to simultaneously achieve efficiency improvement and emission control, resulting in insufficient optimization capabilities.

[0007] In a first aspect, embodiments of this application provide an adaptive boiler combustion control method, including:

[0008] Obtain boiler operating control parameters;

[0009] Based on the boiler operation control parameters, multiple multidimensional particles are determined, wherein each multidimensional particle is a set of parameter adjustment results of the boiler operation control parameters;

[0010] Based on the particle swarm optimization algorithm and the pre-trained Bayesian model, each of the multidimensional particles is iteratively updated until the maximum number of iterations is reached or the multidimensional particles converge to the optimal position.

[0011] Among the multiple multidimensional particles after iterative updates, the multidimensional particle with the lowest defect degree is determined as the optimization target of the boiler operation control parameters.

[0012] Optionally, in one embodiment of this application, obtaining the boiler operation control parameters includes:

[0013] Real-time monitoring of several specified boiler operation control parameters;

[0014] When the change in any of the specified boiler operation control parameters is greater than or equal to a predetermined change threshold, all boiler operation control parameters are acquired.

[0015] In one embodiment of this application, optionally, the independent variable includes adjustable operating control parameters and non-adjustable operating control parameters, wherein,

[0016] The adjustable operating control parameters include: air-coal ratio, inlet oxygen content, total primary air volume, total secondary air volume, burnout air regulating damper opening, and burner secondary air regulating damper opening.

[0017] The non-adjustable operating control parameters include: unit power, total coal consumption, main steam flow rate, feedwater temperature, air preheater secondary air outlet temperature, air preheater primary air outlet temperature, coal mill air volume, coal mill coal consumption, and coal mill outlet temperature.

[0018] The dependent variables include: air preheater outlet flue gas temperature, boiler efficiency, CO content, NOx emissions, and fly ash combustible concentration.

[0019] In one embodiment of this application, optionally, determining multiple multidimensional particles based on the boiler operation control parameters includes:

[0020] Obtain the predetermined safety range for each of the aforementioned boiler operation control parameters;

[0021] Based on the predetermined safety range of each of the aforementioned boiler operation control parameters, multiple multidimensional particles are randomly generated. In the parameter adjustment result corresponding to each multidimensional particle, each of the aforementioned boiler operation control parameters is within its corresponding predetermined safety range. Each multidimensional particle includes the feature values ​​of the aforementioned boiler operation control parameters and the confidence level of each feature value. The confidence level is used to reflect the degree of contribution of the boiler operation control parameter to the boiler operation performance information when it is at its current position. The current position refers to the relative position of the feature value within the predetermined safety range corresponding to the boiler operation control parameter.

[0022] In one embodiment of this application, optionally, the iterative update of each of the multidimensional particles based on the particle swarm optimization algorithm and the pre-trained Bayesian model includes:

[0023] In each iteration update, for each multidimensional particle, the multidimensional particle is input into a pre-trained Bayesian model, and the Bayesian model outputs the boiler operation performance information corresponding to the multidimensional particle. The Bayesian model is used to reflect the correlation between the boiler operation control parameters and the boiler operation performance information.

[0024] Based on the boiler operation performance information and the multi-objective optimization control function in the particle swarm optimization algorithm, the defect degree of the multi-dimensional particle is determined. The defect degree is used to reflect the negative performance of the boiler operation after optimization to the parameter adjustment result represented by the multi-dimensional particle. The multi-objective optimization control function is used to reflect the correlation between the boiler operation performance information and the defect degree.

[0025] Based on the historical optimal solution and the global optimal solution of the multidimensional particle, the position of each boiler operation control parameter in the multidimensional particle is updated.

[0026] In one embodiment of this application, optionally, before obtaining the boiler operation control parameters, the method further includes:

[0027] Obtain training samples of the boiler, wherein the training samples include: the boiler's operating control parameters and operating performance information in historical operation, and / or the boiler's operating control parameters and operating performance information in experimental operation;

[0028] The running control parameters in the training samples are set as independent variables, and the running performance information in the training samples is set as dependent variables. A Bayesian network structure is constructed with independent and dependent variables as nodes and the relationship between independent and dependent variables as edges.

[0029] Based on the training samples, the parameters of the Bayesian network structure are trained to obtain the Bayesian model.

[0030] In one embodiment of this application, optionally, training the parameters of the Bayesian network structure based on the training samples includes:

[0031] If the training samples are complete, the parameters of the Bayesian network structure are determined using the maximum likelihood estimation method.

[0032] If the training samples lack observation data at any time point, the expectation-maximization algorithm is used to determine the optimal approximate estimate of the parameters of the Bayesian network structure in the region.

[0033] In one embodiment of this application, optionally, determining the defect degree of the multidimensional particle based on the boiler operating performance information and the multi-objective optimization control function within the particle swarm optimization algorithm includes:

[0034] Obtain the boiler efficiency and NO from the boiler operation performance information. x Emissions;

[0035] Based on the boiler efficiency and NO x The emission rate, and the multi-objective optimization control function within the particle swarm optimization algorithm, are used to determine the defect degree of the multidimensional particle, wherein,

[0036] χ=k1NO x +k2(1-η),

[0037] χ represents the defect degree of the multidimensional particle, NOx represents the NOx emission in the boiler operation performance information in the current iteration, η represents the boiler efficiency in the boiler operation performance information in the current iteration, and k1 and k2 are the negative performances of the NOx emission and the subtraction of the boiler efficiency relative to 1 when the boiler operation is optimized to the current level, respectively.

[0038] In a second aspect, embodiments of this application provide a computer device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being configured to perform the method described in the first aspect above.

[0039] Thirdly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions for performing the method described in the first aspect above.

[0040] The above technical solution addresses the technical problem in related technologies where boiler combustion parameter optimization methods struggle to balance efficiency improvement and emission control, resulting in insufficient optimization capabilities. This solution, based on extensive historical and experimental boiler operating data, combines Bayesian networks and particle swarm optimization to predict and optimize the optimal solutions for boiler operating control parameters. This achieves a relatively balanced and optimal balance between boiler efficiency and NOx emissions, enabling multi-objective automated dynamic optimization of boiler combustion. This not only improves boiler efficiency but also ensures NOx emissions remain within a reasonable range while meeting efficiency requirements. Furthermore, it reduces reliance on manual experience, achieving real-time and effective control of the boiler's operating status and performance. [Attached Image Description]

[0041] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 A flowchart of an adaptive boiler combustion control method according to an embodiment of this application is shown;

[0043] Figure 2 A schematic diagram of the structure of a Bayesian network according to an embodiment of this application is shown;

[0044] Figure 3 A schematic diagram of a Bayesian network applied to boiler operation control according to an embodiment of this application is shown.

[0045] Figure 4 A schematic diagram of the overall interface of the intelligent combustion optimization control module according to an embodiment of this application is shown;

[0046] Figure 5 A schematic diagram of a combustion process model variable setting interface according to an embodiment of this application is shown;

[0047] Figure 6 A schematic diagram showing a comparison between predicted and actual values ​​of boiler efficiency according to an embodiment of this application is illustrated.

[0048] Figure 7 A schematic diagram showing the optimized air preheater outlet flue gas temperature reduction performance according to an embodiment of this application is illustrated.

[0049] Figure 8 A schematic diagram illustrating the principle of optimized nitrogen oxide emission concentration reduction according to an embodiment of this application is shown.

[0050] Figure 9 A block diagram of a computer device according to one embodiment of this application is shown;

[0051] Figure 10 A block diagram of a computer device according to another embodiment of this application is shown.

Detailed Implementation Methods

[0052] 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, not all, of the embodiments of the present invention. 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.

[0053] Figure 1 A flowchart of an adaptive boiler combustion control method according to an embodiment of this application is shown.

[0054] like Figure 1 As shown, an adaptive boiler combustion control method according to an embodiment of this application includes:

[0055] Step 102: Obtain boiler operation control parameters.

[0056] Acquiring boiler operation control parameters refers to collecting key parameter data, both adjustable and unadjustable, during boiler operation. These parameters include adjustable variables such as air-coal ratio, oxygen content, and air volume, as well as uncontrollable variables such as unit load and coal quality characteristics. Comprehensive acquisition of these parameters can accurately reflect the current operating status of the boiler. The purpose of this step is to provide a complete and accurate input data foundation for subsequent optimization, ensuring that the optimization system can operate based on real and reliable boiler operating status information and avoiding optimization deviations due to missing data.

[0057] In one possible design, several specified boiler operation control parameters can be monitored in real time. When the change of any of the specified boiler operation control parameters is greater than or equal to a predetermined change threshold, all boiler operation control parameters are obtained.

[0058] The independent variables include, but are not limited to, adjustable and non-adjustable operating control parameters. The adjustable operating control parameters include, but are not limited to: air-to-coal ratio, inlet oxygen content, total primary air volume, total secondary air volume, burnout air damper opening, and burner secondary air damper opening. The non-adjustable operating control parameters include, but are not limited to: unit power, total coal consumption, main steam flow rate, feedwater temperature, air preheater secondary air outlet temperature, air preheater primary air outlet temperature, pulverizer air volume, pulverizer coal consumption, and pulverizer outlet temperature. The dependent variables include, but are not limited to: air preheater outlet flue gas temperature, boiler efficiency, CO (carbon monoxide) content, NOx (nitrogen oxides) emissions, and fly ash combustible concentration. NOx is a general term for compounds formed by the combination of nitrogen and oxygen, including but not limited to nitric oxide (NO) and nitrogen dioxide (NO2), which are harmful gases produced during combustion.

[0059] Optionally, the boiler described in this application includes a W-flame boiler that burns all anthracite. Of course, the technical solution described in this application is not only applicable to W-flame boilers that burn all anthracite, but can also be widely applied to many other types of coal-fired boilers. In the field of pulverized coal boilers, this technology is applicable to common boiler types such as opposed-wall combustion boilers and tangential combustion boilers. These boilers also face the contradiction between combustion efficiency and emission control when burning high-ash, low-volatile anthracite or lean coal. For circulating fluidized bed boilers, this technology can effectively optimize their unique fluidized combustion process, and is particularly suitable for adjusting operating parameters when burning low-calorific-value fuels such as coal gangue and coal slime. In terms of stoker boilers, this solution can be applied to equipment such as chain grate boilers, improving combustion conditions by optimizing coal feed rate, grate speed, and air distribution parameters through intelligent algorithms. It is worth noting that this solution is adaptable to boilers of different capacity levels; whether it is a small or medium-sized industrial boiler or a large power plant boiler, intelligent optimization of the combustion process can be achieved through this technology.

[0060] Anthracite, a special type of coal with high fixed carbon and low volatile matter, exhibits significantly different combustion characteristics compared to conventional bituminous coal. Under traditional combustion methods, anthracite often faces problems such as high fire temperature and difficulty in complete combustion, leading to low boiler efficiency and uncontrollable pollutant emissions. The technical solution presented in this application effectively solves the technical challenge of poor combustion stability of anthracite by dynamically adjusting combustion conditions in real time using key parameters such as the air-coal ratio and oxygen content, thereby significantly improving combustion efficiency while ensuring boiler output.

[0061] To address the issue of high NOx generation during anthracite combustion, this solution utilizes a pre-trained Bayesian model to accurately predict emission characteristics under different air distribution methods. By establishing a multi-objective optimization function, the system automatically optimizes the operating parameters to meet both efficiency requirements and reduce NOx emissions. This results in a lower NOx emission concentration in the anthracite boiler compared to before optimization, while also effectively improving boiler efficiency.

[0062] This scheme also specifically considers the safety boundary constraints of anthracite combustion, strictly limiting the variation of each parameter within the allowable range during the optimization process. By setting reasonable upper and lower limits for the air-to-coal ratio and oxygen control ranges, it avoids problems such as slagging and high-temperature corrosion caused by excessive pursuit of efficiency, while ensuring the long-term stable operation of the boiler. This intelligent optimization method, which balances economy and safety, provides reliable technical support for the efficient and clean utilization of anthracite.

[0063] Step 104: Based on the boiler operation control parameters, determine multiple multidimensional particles, wherein each multidimensional particle is a set of parameter adjustment results of the boiler operation control parameters.

[0064] Therefore, each possible parameter adjustment scheme is encoded as a multi-dimensional vector, with each dimension corresponding to the adjustment amount of a control parameter. The multi-dimensional particle represents a complete set of parameter optimization schemes. In this way, an optimization search space can be constructed, providing diverse candidate solutions for intelligent algorithms. By randomly generating multiple initial solutions, the search range is expanded, increasing the probability of finding the global optimum and avoiding optimization from getting trapped in local optima.

[0065] In one possible design, step 104 includes: obtaining the predetermined safety range of each of the multiple boiler operation control parameters; and randomly generating multiple multidimensional particles based on the predetermined safety range of each of the multiple boiler operation control parameters, wherein each of the boiler operation control parameters in the parameter adjustment result corresponding to each multidimensional particle is within its corresponding predetermined safety range.

[0066] Each of the multidimensional particles includes feature values ​​of multiple boiler operation control parameters and a confidence level for each feature value. The confidence level is used to reflect the degree of contribution of the boiler operation control parameter to the boiler operation performance information when it is in its current position. The current position refers to the relative position of the feature value within the predetermined safety range corresponding to the boiler operation control parameter.

[0067] Specifically, a single multidimensional particle is not simply defined by multiple boiler operation control parameters as multiple dimensions, but rather, while defining multiple boiler operation control parameters as multiple dimensions, a confidence level is set for each of these multiple dimensions. It can be said that a single multidimensional particle is a twenty-dimensional structured data, or a ten-dimensional structured data, and each of its dimensions is divided into a sub-dimension containing the associated boiler operation control parameters and a sub-dimension containing the confidence level corresponding to the boiler operation control parameters.

[0068] In particle swarm optimization (PSO), a single multidimensional particle contains not only the feature values ​​of boiler operation control parameters but also the confidence index corresponding to each feature value. The confidence index reflects the contribution of the parameter to the boiler's operational performance at its current value; it can be considered the weight of the boiler operation control parameter, and its value is closely related to the relative position of the boiler operation control parameter within a predetermined safe range. This dual-dimensional design makes each multidimensional particle actually constitute a more complex structured data space. During each iteration, the particle fitness is evaluated not only based on the changes in the parameter feature values ​​themselves but also by referring to the confidence index to adjust the magnitude and direction of parameter adjustments. Parameter dimensions with high confidence receive larger weight adjustments, guiding particles to move towards search regions with higher confidence. This mechanism effectively avoids the blind search problem that may occur in traditional optimization algorithms, making the parameter optimization process more accurate and efficient.

[0069] By incorporating confidence levels along with boiler operating control parameters into the iterative process, the introduction of confidence levels significantly enhances the intelligence of the optimization process, facilitating the differentiation of the importance of different boiler operating control parameters. This design also strengthens the ability to identify the coupling relationships between parameters, better balancing the interdependent indicators such as boiler efficiency and emission control in multi-objective optimization. Most importantly, the confidence level mechanism ensures that the optimization process always operates within a safe and reliable parameter range, avoiding unreasonable extreme parameter combinations and guaranteeing the stability of boiler operation. This multi-dimensional particle structure that integrates parameter values ​​and confidence levels makes the final optimization result both theoretically optimal and feasible in engineering practice.

[0070] Step 106: Based on the particle swarm optimization algorithm and the pre-trained Bayesian model, iteratively update each of the multidimensional particles until the maximum number of iterations is reached or the multidimensional particles converge to the optimal position.

[0071] This approach utilizes swarm intelligence algorithms to simulate bird flock foraging behavior, while simultaneously combining this with probabilistic graphical models to predict boiler performance under different parameter combinations. By continuously adjusting particle positions to find the optimal solution, efficient multi-objective collaborative optimization can be achieved. This retains the global search capability of swarm intelligence algorithms while leveraging Bayesian networks to accurately evaluate the merits of each solution, significantly improving optimization efficiency and accuracy.

[0072] Step 108: Among the multiple multidimensional particles after iterative updates, determine the multidimensional particle with the lowest defect degree as the optimization target of the boiler operation control parameters.

[0073] Determining the multidimensional particle with the lowest defect degree as the optimization target after iterative updates means comprehensively considering boiler efficiency and emission indicators. This allows us to select the optimal parameter combination as the final optimization result. In other words, the defect degree quantifies the comprehensive performance defects of each scheme, ensuring the feasibility and superiority of the optimization result in actual operation. It can automatically select the optimal operating scheme that balances economy and environmental protection, providing scientific and reasonable control parameter settings for the boiler.

[0074] In one possible design, during each iteration, for each multidimensional particle, the particle is input into a pre-trained Bayesian model, which outputs boiler operation performance information corresponding to the multidimensional particle. Then, based on the boiler operation performance information and the multi-objective optimization control function within the particle swarm optimization algorithm, the defect degree of the multidimensional particle is determined. Finally, based on the historical optimal solution and the global optimal solution of the multidimensional particle, the position of each boiler operation control parameter in the multidimensional particle is updated.

[0075] The Bayesian model is used to reflect the correlation between the boiler operation control parameters and the boiler operation performance information. The defect degree is used to reflect the negative performance after the boiler operation is optimized to the parameter adjustment results represented by the multi-dimensional particle. The multi-objective optimization control function is used to reflect the correlation between the boiler operation performance information and the defect degree.

[0076] Bayesian networks simulate human cognitive reasoning, using directed acyclic graphs (DAGs) and a set of conditional probability functions to model uncertain causal relationships. A Bayesian network definition includes a DAG and a set of conditional probability tables. Each node in the DAG represents a random variable, which can be directly observable or hidden, while directed edges represent conditional dependencies between random variables. Each element in the conditional probability table corresponds to a unique node in the DAG, storing the joint conditional probability of that node with respect to all its direct predecessors.

[0077] It is important to note that once the value of each node's direct predecessor is determined, that node's condition is independent of all its non-direct predecessors. This clarifies the computable joint probability distribution of a Bayesian network. Generally, the multivariate non-independent joint conditional probability distribution is obtained by the following formula:

[0078] P(x1,x2,...,x n )=P(x1)P(x2|x1)P(x3|x1,x2)...P(x n |x1,x2,...,x n-1 ),

[0079] In Bayesian networks, due to the aforementioned properties, the joint conditional probability distribution of any combination of random variables can be simplified to:

[0080]

[0081] Among them, Parents(x i ) represents the variable x i The probability values ​​for the joint of the direct predecessor nodes can be found in the corresponding conditional probability table obtained by training from historical data.

[0082] As can be seen, Bayesian networks greatly reduce the computational cost of joint conditional probabilities, thus ensuring the computational speed of real-time optimization control operations.

[0083] like Figure 2 In the example shown, the Bayesian network describes the causal relationships between five random variables W, X, Y, V, and Z. Each node represents a discrete random variable, and each single arrow connecting two variables indicates a conditional dependency. For example, W→Y means that W is the "cause" variable and Y is the "effect" variable.

[0084] In addition to the directed acyclic graph representing the structure of the Bayesian network, each node has a conditional probability table (CPT) representing the conditional probability of each state of the node given all possible combinations of the dependent variables. For example, if node W has two states, w1 and w2, then its resultant node Y has conditional probability values ​​including P(y1|w1), P(y2|w1), P(y1|w2), and P(y2|w2).

[0085] From the structure of the Bayesian network, we know that variable X is not the "effect" node of variable Y, and variable Y has only one direct predecessor node W. According to the Markov assumption, P(Y|W,X) = P(Y|W). Substituting into the previous equation, we get:

[0086] P(W,X,Y,V,Z)=P(W)P(X)P(Y|W)P(V|Y)P(Z|X,Y),

[0087] In summary, Bayesian networks are mainly used for probabilistic reasoning and decision-making. They infer unobservable random variables from observable random variables when information is incomplete. There can be more than one unobservable random variable. Generally, the unobservable variables are initially set to random values, and then probabilistic reasoning is performed.

[0088] Figure 3 This illustration shows a schematic diagram of a Bayesian network applied to boiler operation control according to an embodiment of this application. Figure 3 For the multi-input multi-output combustion process model of a boiler, the structure of its Bayesian network can be constructed based on the independent and dependent variables in the boiler operation control parameters and boiler operation performance information.

[0089] like Figure 3 As shown, each node in a Bayesian network represents a random variable related to the boiler combustion process, including combustion control variables and boiler operating state parameters. An arrow connecting two nodes indicates that the two random variables are causally related or unconditionally independent. If there are no arrows connecting the variables within nodes, their random variables are said to be conditionally independent. If two nodes are connected by a single arrow, it means that one node is the "cause" and the other is the "effect," and the two nodes will generate a conditional probability value.

[0090] For example, changes in combustion control parameters such as primary air velocity and volume, and burner sway angle can directly affect the CO content measured at the boiler outlet. Therefore, these combustion control parameters are all "cause" nodes of CO content, i.e., the aforementioned direct precursor nodes. It can be deduced that each combustion control parameter and CO content have a causal relationship or are not conditionally independent. Therefore, the two nodes representing the combustion control parameter and CO content are connected by a directed edge with a single arrow.

[0091] In the multi-input multi-output combustion process model of a boiler, changes in combustion control variables directly lead to changes in the boiler's operating state. Therefore, in the Bayesian network, the combustion control variables serve as "cause" nodes, the boiler's operating state parameters serve as "effect" nodes, and the directed edges between nodes represent which combustion control variables influence each boiler's operating state parameter. The input variables (independent variables) and output variables (dependent variables) of the boiler combustion process model have been described above and will not be repeated here.

[0092] Based on the boiler's historical operating data and relevant experimental data, Bayesian networks can automatically learn model parameters offline, which represent the probability distribution of the strength of the dependencies between various combustion variables, organically combining prior information with sample knowledge.

[0093] For example, combustion control variables, acting as cause nodes in a Bayesian network, directly drive changes in operating state parameters, which then act as effect nodes. Total air volume, a key control variable, is divided into six adjustment ranges within the range of 3.1e5 to 3.8e5, each corresponding to a different combustion intensity strategy. When the air volume increases from 3.4e5 to 3.5e5, the flue gas temperature drops from 125℃ to 124℃. Flue gas oxygen content, another core control variable, has twelve adjustment gradients within the range of 4 to 5.2. An increase in oxygen content from 4.1 to 5.1 causes NOx emissions to decrease from -300 to -345. Simultaneously, the secondary air pressure at the boiler wall inlet forms six control zones within the range of 0.25 to 0.55. When the secondary air pressure is adjusted from 0.35 to 0.39, the boiler efficiency increases from 92.4% to 92.6%. The synergistic effect of these control variables achieves optimal operating conditions under specific combinations. For example, when the total air volume is stable at 24.9% within the range of 3.5e5 to 3.6e5, the oxygen content is maintained at a confidence value of 70.33 between 4.34 and 5.18, and the secondary air pressure is controlled at ±0.013 of 0.382, the system output parameters reach a multi-objective optimal state, manifested as a boiler efficiency of 92.44, NOx emissions not exceeding -300, and flue gas temperature below 128℃. This fully verifies the directed edge weight relationship between control variable nodes and state variable nodes in the Bayesian network. A typical causal path is shown where the total air volume of 3.6e5, as an input control variable, affects the state variable of flue gas temperature of 127℃, ultimately impacting the output parameter of boiler efficiency of 92.44. Meanwhile, the oxygen content of 5.02, as another control variable, affects the excess air coefficient of 1.288 by adjusting the state variable of NOx emissions of -300. These data relationships strictly follow the modeling principle that changes in combustion control variables directly lead to changes in operating conditions.

[0094] The determination of parameters for a boiler combustion process model based on a Bayesian network involves learning and training the network across the entire sample space of historical boiler operating data, given a predetermined Bayesian network structure, to obtain a conditional probability table. In this table, each probability value can be represented as P(X...). i |parents(X i Then the joint probability assignment of the Bayesian network can be expressed as:

[0095]

[0096] Among them, X j For each relative to X i The "dependent" variable.

[0097] When the structure of a Bayesian network is known, the determination of parameters falls into two categories: one where the training data sample space is complete, and the other where the sample space is missing. Because the Bayesian network algorithm has a strong ability to handle uncertainty, it can model even when the data sample space is missing. Therefore, a boiler combustion object model based on a Bayesian network can more objectively reflect the true state of the boiler combustion process.

[0098] If the training samples are complete, the parameters of the Bayesian network structure are trained by using the maximum likelihood estimation method to determine the parameters of the Bayesian network structure; if the training samples are missing observation data at any time point, the parameters of the Bayesian network structure are trained by using the expectation-maximization algorithm to determine the optimal regional estimate of the parameters of the Bayesian network structure.

[0099] When the historical data observations are complete, the parameters of the Bayesian network are calculated using Maximum Likelihood Estimation (MLE). The training sample data is known to be D = {x1,...,x...}. m}, where x l =(x l1 ,...,x ln ) T Assume the parameter set is Θ = (θ1,...,θ2) n ), where θ i Represents variable X i The conditional probability vector. Then the log-likelihood function reflecting the parameters of the training sample data is:

[0100]

[0101] Wherein, pa(X) i ) represents X i The "dependent" variable, that is, its direct predecessor variable, D i N represents a sample of the boiler's historical operating data, where N represents the total number of historical operating data samples.

[0102] When the historical operating data is incomplete, such as missing data at certain time points, and if some combustion process variables cannot be observed, the EM algorithm can be used to determine the best approximate estimate of the parameters in the region.

[0103] The steps of the EM algorithm are as follows: First, give an initial value for the parameter to be estimated. Then, use this initial value and other observations to calculate the conditional expectation value of other unobserved nodes. Next, treat the estimated value as the observation value and substitute this complete sample of observations into the maximum likelihood estimation formula of the model.

[0104] In an embodiment of this application, the system obtains a Bayesian network-based combustion process model of a boiler through offline learning of the boiler's operating history data. When one or more combustion parameters change, this model can deduce the maximum probability values ​​of other adjustable variables, as well as the probability distributions of boiler efficiency and NOx emissions under that operating condition.

[0105] Once the model structure and parameters are determined, the Bayesian network-based combustion process model will serve as the objective function in the optimal control, used to evaluate the merits of each set of candidate combustion control variables. Through Bayesian inference, the corresponding boiler operating state can be derived from each set of candidate combustion control variables, thereby selecting the optimal combustion control variable, which can improve boiler combustion efficiency while reducing NOx emissions. This scheme uses variable elimination in its Bayesian inference, that is, removing irrelevant variables by adding joint probability distributions, thus obtaining the conditional probability of any variable. This includes two cases: one is predicting the boiler operating state affected by the observed combustion control variable, i.e., prediction support; the other is inferring the changes in the combustion control variable that caused the observed boiler operating state, i.e., diagnostic support.

[0106] Therefore, the beneficial effects of the boiler combustion performance prediction model based on Bayesian networks are as follows.

[0107] Boiler combustion performance prediction models based on Bayesian networks can reflect the causal characteristics of how boiler efficiency and NOx emissions vary with different air distribution and combustion operation modes. This not only provides a better understanding of the boiler combustion process but also enables accurate predictions even under conditions of significant external disturbances.

[0108] Furthermore, Bayesian networks can easily handle incomplete data. For example, artificial neural networks often exhibit significant biases in their predictive structures when dealing with missing values ​​for certain variables. Bayesian networks, however, possess strong capabilities in handling uncertainty; the algorithm can model data even when the sample space is incomplete and provides a more intuitive probabilistic correlation model. Therefore, a boiler combustion model based on Bayesian networks can more objectively reflect the true state of the boiler combustion process.

[0109] The structure and parameters of the Bayesian network are obtained entirely through learning and training from historical operating data of boiler combustion. Therefore, its modeling algorithm has strong adaptability and is suitable for boilers of different manufacturers, models, capacities and fuels. After changing the research object, there is no need to redesign and compile the main algorithm for model building.

[0110] Of course, Bayesian networks can quickly infer the maximum probability values ​​of adjustable variables such as coal feeder, burner, air volume, and swing angle based on known boiler combustion parameters such as unit load and coal quality, as well as the probability distribution of boiler efficiency and NOx emissions under that operating condition. They can also flexibly increase or decrease input and output parameters according to user requirements, with no limit on the number of input and output parameters, and changing the number of parameters does not affect the calculation speed of Bayesian inference in real-time optimization control operations.

[0111] Furthermore, combining Bayesian networks with the genetic algorithm described later can effectively avoid overfitting and local extrema during training. Unlike artificial neural networks, Bayesian networks increase both generalization ability and model accuracy simultaneously. Bayesian networks can update their parameters in real time based on new training samples, and with an increasing number of training samples, they can more accurately reflect the causal relationships between variables.

[0112] Furthermore, the principle of the particle swarm optimization algorithm is as follows.

[0113] In practical applications of particle optimization algorithms, the potential solution to each optimization problem can be imagined as a particle in a D-dimensional search space. Each particle has a fitness value determined by the objective function. These particles fly in the search space at a certain speed, the magnitude and direction of which are dynamically adjusted based on the particle's own flight experience and the flight experience of the entire population. Subsequently, all particles will follow the current optimal particle in the search space.

[0114] Suppose we are looking for a minimum value in a problem where we need to find the optimal solution x. In a D-dimensional search space, there is a swarm of N particles, where the i-th particle is represented as a D-dimensional vector. i = 1, 2, ..., N, that is, the position of the i-th particle in the D-Wade search space is In other words, the position of each particle is a potential solution to the optimization problem. Substituting the values ​​into the objective function allows us to calculate the fitness value, which is then used to measure performance. The merits and demerits. Let the fitness value of each particle be Fitness. i (i∈[1,N]). The velocity of the i-th particle is also a D-dimensional vector, denoted as... Let the optimal position found so far by the i-th particle be denoted as . The optimal position found so far by the entire particle swarm is g. best =(g1,g2,...,g D The operational mode of each particle depends not only on its own flight experience (i.e., p) best It is also affected by the flight experience of the entire population (i.e., g). best Therefore, the particle swarm optimization algorithm can guarantee that the final result is the global optimum, rather than being trapped in a local optimum. The process of the particle swarm optimization algorithm is described below.

[0115] (1) Set the initial values ​​for N particles, each particle x i Let i represent a potential solution to the optimization problem, where i ∈ [1, N]. For the minimum optimization problem, the fitness value of each particle is... i The optimal position of each particle And the optimal position g of the entire population best Set all to infinity.

[0116] (2) When the number of iterations t reaches the set maximum number of iterations t max Previously, or if a certain termination condition was not met, the following steps were repeated in each iteration:

[0117] Calculate the fitness value for each particle. i =f(x) i );

[0118] Update the best position found so far for each particle.

[0119] Update the optimal position found so far for the entire particle swarm.

[0120] The particle is moved according to the following formula.

[0121] x i,t+1 =x i,t +u i,t+1 ,,

[0122] Where, u i,t+1 Defined as:

[0123]

[0124] u i,t u represents the velocity of the i-th particle during time interval t. i,t+1ω represents the velocity of the i-th particle in the next time interval t+1, where ω is a constant less than 1, used to reflect the influence of the particle's velocity in time interval t on its velocity in the next time interval t+1. i,t This represents the current position of the i-th particle. Learning factors c1 and c2 are the weights of these variables in determining the flight speed. r1 and r2 are random constants between [0,1], adding a random element to the algorithm.

[0125] (3) When the iteration ends, the optimal solution x that satisfies the multidimensional objective function f(x) can be obtained.

[0126] Based on the above principles, firstly, a search space is defined to determine the parameter range of the optimization problem, setting boundary conditions for subsequent particle motion. Next, the initial positions and velocities of the particles are randomly set; that is, the particle swarm is randomly initialized within the search space, with each particle representing a potential solution and assigned an initial velocity. Further, the fitness value of each particle is evaluated, and the global and local optimal positions are stored. This involves calculating the objective function value (fitness) corresponding to each particle's current position, while simultaneously recording the individual's historical best position and the swarm's global best position. The optimal solution is the current global best position, which can be tentatively set as the best position among all particles.

[0127] At this point, the termination condition is verified to determine if the preset termination criteria are met, such as reaching the maximum number of iterations or achieving sufficiently high solution quality. If the termination condition is not met, the iteration count is incremented by 1, and the optimization process continues.

[0128] Finally, based on the fitness value, the current optimal position globally and locally is updated. This involves comparing the old and new fitness values ​​to update the historical optimal records for individuals and the population. When updating the flight velocity and position of each particle, the particle's direction of motion and step size can be adjusted based on individual and population optimal information, guiding it to explore more favorable regions.

[0129] When the termination condition is met, the algorithm ends and outputs the optimal solution.

[0130] In one possible design, independent variables (i.e., operating quantities) may include, but are not limited to: unit load, coal quality, primary air volume, secondary air volume at each level, pulverizer outlet air volume, flue gas oxygen content, feeder bias, and burner organization, etc., while dependent variables (i.e., control quantities) may include, but are not limited to: NOx emissions, flue gas temperature, fly ash carbon content, excess air coefficient, and boiler efficiency, etc. Among these, boiler efficiency and NOx emissions can be selected. x Emissions are used as two dependent variables to evaluate the results of boiler parameter optimization.

[0131] Specifically, boiler efficiency and NO can be obtained from the boiler operation performance information. xEmissions, and based on the boiler efficiency and NO x The emission amount, and the multi-objective optimization control function within the particle swarm optimization algorithm, are used to determine the defect degree of the multidimensional particle.

[0132] χ=k1NO x +k2(1-η),

[0133] χ represents the defect degree of the multidimensional particle, NOx represents the NOx emission in the boiler operation performance information in the current iteration, η represents the boiler efficiency in the boiler operation performance information in the current iteration, and k1 and k2 are the negative performances of the NOx emission and the subtraction of the boiler efficiency relative to 1 when the boiler operation is optimized to the current level, respectively.

[0134] In summary, this technical solution proposes an adaptive intelligent boiler combustion optimization control system, possessing all the beneficial effects of any of the aforementioned embodiments. Based on research into intelligent modeling and optimization algorithms, this system establishes a combustion system model through boiler operation data analysis and obtains the optimal combination of operating modes under various conditions, thus acquiring the boiler's basic operation optimization strategy. According to the boiler's operating conditions and combustion status, it optimizes and adjusts adjustable variables such as the burner secondary air damper, inlet oxygen content, and burnout layer damper in real time, ultimately achieving closed-loop combustion optimization control. This aims to improve boiler efficiency, reduce flue gas NOx emissions, improve the boiler combustion system's operating condition, enhance the boiler's automation level, and reduce the workload of operators. The overall interface of its intelligent combustion optimization control module is shown below. Figure 4 As shown, the basic functions of this module include:

[0135] First, a boiler combustion process model was established using algorithms such as Bayesian methods.

[0136] After offline modeling and importing historical data for analysis, the causal relationships between the input and output parameters are established. Clicking "Variable Maintenance" at the top of the main interface brings up the settings interface for all input and output variables of the boiler combustion process model, i.e., the combustion process model variable settings interface. Figure 5 The interface shown is for setting variables in the combustion process model. In this interface, you can select model variables, as well as characteristic values ​​such as upper and lower limits of the variables.

[0137] Second, store and read in the existing model.

[0138] After incorporating historical boiler operation data and establishing a model, the model's stability reached 98.5%, essentially concluding the modeling process. The model was then validated and compared using 2000 sets of historical data, combined with... Figure 6The predicted and actual boiler efficiency values ​​shown in the figure are compared, and the accuracy of the predicted and actual values ​​reaches 97.4%. The overall accuracy of the model meets the standard and can be used for performance prediction and optimization of input parameters.

[0139] In the boiler process modeling parameter configuration interface, you can set the reading step size of historical data, the model prediction calculation frequency, the save path of historical data, and the storage path of the model parameter CPT table file. After completing the settings, you can click "Model Load" in the boiler process modeling parameter configuration interface to call the Bayesian network modeling algorithm to build the boiler combustion process model. After modeling is completed, you can modify the storage and reading path of the model parameter CPT table.

[0140] Third, it stores, retrieves, and dynamically displays optimization suggestions for adjustable variables, as well as important parameters describing the boiler combustion status.

[0141] Fourth, it stores, retrieves, and dynamically displays optimization targets such as boiler efficiency and NOx emissions.

[0142] This system uses a particle swarm optimization algorithm to automatically adjust the inlet oxygen content, secondary air and burnout air ratio, and other relevant controllable parameters of the coal-fired boiler combustion optimization system according to different operating conditions during load and coal quality changes. This makes the target parameters such as efficiency and emissions move towards a more optimized performance, thereby achieving the goals of emission reduction and efficiency improvement. Figure 7 A schematic diagram showing the optimized air preheater outlet flue gas temperature reduction performance according to an embodiment of this application is illustrated, as follows: Figure 7 As shown, after optimization, the flue gas temperature at the air preheater outlet decreased significantly. Through testing, it was verified that under the same operating conditions, the efficiency of the optimized boiler was reduced by an average of 0.6% compared to that before optimization. Figure 8 A schematic diagram illustrating the principle of optimized nitrogen oxide emission concentration reduction according to an embodiment of this application is shown, as follows. Figure 8 As shown, NOx emissions decreased by 5%-10% at the same time.

[0143] In summary, this technical solution establishes a multi-input multi-output intelligent control model for the boiler combustion system based on a large amount of historical boiler operation data and artificial intelligence algorithms such as continuous-time Bayesian network prediction. This model first undergoes offline learning using historical operating data and relevant experimental data, and then continues to be updated, corrected, and optimized based on real-time boiler operation data, thus establishing a continuously evolving dynamic optimization model. Simultaneously, a particle swarm optimization algorithm with dynamically adjusted inertial weights is applied to the real-time combustion optimization system of a power plant pulverized coal boiler. This algorithm dynamically changes particle weights based on particle swarm evolution rate and aggregation factors, exhibiting dynamic adaptive optimization characteristics, and its search speed and accuracy are significantly higher than existing optimization algorithms.

[0144] The control strategy of this technical solution overcomes the various defects of traditional optimization methods, making the model's convergence accuracy and computation speed superior to the current mainstream algorithms. Moreover, it can effectively balance the two contradictory optimization objectives of boiler efficiency and NOx emissions, thereby improving boiler efficiency and reducing flue gas NOx emissions.

[0145] Furthermore, the aforementioned intelligent boiler combustion optimization control system was applied to a 600MW unit boiler. Through on-site testing by boiler professionals before and after optimization, the boiler efficiency increased by 0.3%-0.6% and NOx emissions decreased by an average of 9% after optimization. Based on this, it is estimated that annual standard coal consumption can be reduced by 4,000 tons, fuel costs can be reduced by approximately 1 million yuan, and annual NOx emissions can be reduced by 100 tons. This achieves both improved boiler efficiency and reduced pollutant emissions, demonstrating significant environmental and economic benefits.

[0146] In another embodiment, this application provides a computer device, which may be a server, and its internal structure diagram may be as follows. Figure 9 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with external clients via a network connection. When the computer program is executed by the processor, it can implement the methods described in any of the above embodiments.

[0147] In one embodiment, this application also provides a computer device, which can be a client, and its internal structure diagram can be as follows: Figure 10 As shown, the computer device includes a processor, memory, network interface, display screen, and input device connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it can implement the methods described in any of the above embodiments.

[0148] Any of the computer devices described in the embodiments of this application exist in various forms, including but not limited to:

[0149] (1) Mobile communication devices: These devices are characterized by their mobile communication capabilities and primarily aim to provide voice and data communication. These terminals include smartphones, multimedia phones, feature phones, and low-end phones.

[0150] (2) Ultra-mobile personal computer devices: These devices fall under the category of personal computers, possessing computing and processing capabilities, and generally also have mobile internet access features. These terminals include PDAs, MIDs, and UMPCs, etc.

[0151] (3) Portable entertainment devices: These devices can display and play multimedia content. This category includes: audio and video players, handheld game consoles, e-books, as well as smart toys, wearable devices, and portable car navigation devices.

[0152] (4) Server: A device that provides computing services. The components of a server include a processor, hard disk, memory, system bus, etc. Servers are similar to general computer architectures, but because they need to provide highly reliable services, they have higher requirements in terms of processing power, stability, reliability, security, scalability, and manageability.

[0153] (5) Other electronic devices with data interaction functions.

[0154] Additionally, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which are used to perform the following steps:

[0155] Obtain boiler operating control parameters;

[0156] Based on the boiler operation control parameters, multiple multidimensional particles are determined, wherein each multidimensional particle is a set of parameter adjustment results of the boiler operation control parameters;

[0157] Based on the particle swarm optimization algorithm and the pre-trained Bayesian model, each of the multidimensional particles is iteratively updated until the maximum number of iterations is reached or the multidimensional particles converge to the optimal position.

[0158] Among the multiple multidimensional particles after iterative updates, the multidimensional particle with the lowest defect degree is determined as the optimization target of the boiler operation control parameters.

[0159] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0160] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0161] Depending on the context, the word "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0162] The terminology used in the embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. The singular forms “a,” “the,” and “the” used in the embodiments of this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise.

[0163] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0164] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in a combination of hardware and software functional units.

[0165] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0166] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. An adaptive boiler combustion control method, characterized in that, include: Obtain boiler operating control parameters; Based on the boiler operation control parameters, multiple multidimensional particles are determined, wherein each multidimensional particle is a set of parameter adjustment results of the boiler operation control parameters; Based on the particle swarm optimization algorithm and the pre-trained Bayesian model, each of the multidimensional particles is iteratively updated until the maximum number of iterations is reached or the multidimensional particles converge to the optimal position. Among the multiple multidimensional particles after iterative updates, the multidimensional particle with the lowest defect degree is determined as the optimization target of the boiler operation control parameters.

2. The method according to claim 1, characterized in that, The acquisition of boiler operation control parameters includes: Real-time monitoring of several specified boiler operation control parameters; When the change in any of the specified boiler operation control parameters is greater than or equal to a predetermined change threshold, all boiler operation control parameters are acquired.

3. The method according to claim 2, characterized in that, The independent variables include adjustable operating control parameters and non-adjustable operating control parameters, wherein, The adjustable operating control parameters include: air-coal ratio, inlet oxygen content, total primary air volume, total secondary air volume, burnout air regulating damper opening, and burner secondary air regulating damper opening. The non-adjustable operating control parameters include: unit power, total coal consumption, main steam flow rate, feedwater temperature, air preheater secondary air outlet temperature, air preheater primary air outlet temperature, coal mill air volume, coal mill coal consumption, and coal mill outlet temperature. The dependent variables include: air preheater outlet flue gas temperature, boiler efficiency, CO content, NOx emissions, and fly ash combustible concentration.

4. The method according to claim 1, characterized in that, The determination of multiple multidimensional particles based on the boiler operation control parameters includes: Obtain the predetermined safety range for each of the aforementioned boiler operation control parameters; Based on the predetermined safety ranges of the various boiler operation control parameters, multiple multidimensional particles are randomly generated, among which... In the parameter adjustment results corresponding to each of the multidimensional particles, each of the boiler operation control parameters is within its corresponding predetermined safety range; Each of the multidimensional particles includes feature values ​​of multiple boiler operation control parameters and a confidence level for each feature value. The confidence level is used to reflect the degree of contribution of the boiler operation control parameter to the boiler operation performance information when it is in its current position. The current position refers to the relative position of the feature value within the predetermined safety range corresponding to the boiler operation control parameter.

5. The method according to any one of claims 1 to 4, characterized in that, The particle swarm optimization algorithm and pre-trained Bayesian model iteratively update each of the multidimensional particles, including: In each iteration update, for each multidimensional particle, the multidimensional particle is input into a pre-trained Bayesian model, and the Bayesian model outputs the boiler operation performance information corresponding to the multidimensional particle. The Bayesian model is used to reflect the correlation between the boiler operation control parameters and the boiler operation performance information. Based on the boiler operation performance information and the multi-objective optimization control function in the particle swarm optimization algorithm, the defect degree of the multi-dimensional particle is determined. The defect degree is used to reflect the negative performance of the boiler operation after optimization to the parameter adjustment result represented by the multi-dimensional particle. The multi-objective optimization control function is used to reflect the correlation between the boiler operation performance information and the defect degree. Based on the historical optimal solution and the global optimal solution of the multidimensional particle, the position of each boiler operation control parameter in the multidimensional particle is updated.

6. The method according to claim 5, characterized in that, Before obtaining the boiler operation control parameters, the following is also included: Obtain training samples of the boiler, wherein the training samples include: the boiler's operating control parameters and operating performance information in historical operation, and / or the boiler's operating control parameters and operating performance information in experimental operation; The running control parameters in the training samples are set as independent variables, and the running performance information in the training samples is set as dependent variables. A Bayesian network structure is constructed with independent and dependent variables as nodes and the relationship between independent and dependent variables as edges. Based on the training samples, the parameters of the Bayesian network structure are trained to obtain the Bayesian model.

7. The method according to claim 6, characterized in that, The step of training the parameters of the Bayesian network structure based on the training samples includes: If the training samples are complete, the parameters of the Bayesian network structure are determined using the maximum likelihood estimation method. If the training samples lack observation data at any time point, the expectation-maximization algorithm is used to determine the optimal approximate estimate of the parameters of the Bayesian network structure in the region.

8. The method according to claim 5, characterized in that, The determination of the defect degree of the multidimensional particle based on the boiler operation performance information and the multi-objective optimization control function within the particle swarm optimization algorithm includes: Obtain the boiler efficiency and NO from the boiler operation performance information. x Emissions; Based on the boiler efficiency and NO x The emission rate, and the multi-objective optimization control function within the particle swarm optimization algorithm, are used to determine the defect degree of the multidimensional particle, wherein, x=k1NO x +k2(1-n), χ represents the defect degree of the multidimensional particle, NOx represents the NOx emission in the boiler operation performance information in the current iteration, η represents the boiler efficiency in the boiler operation performance information in the current iteration, and k1 and k2 are the negative performances of the NOx emission and the subtraction of the boiler efficiency relative to 1 when the boiler operation is optimized to the current level, respectively.

9. A computer device, characterized in that, include: At least one processor; And, a memory communicatively connected to the at least one processor; The memory stores instructions executable by the at least one processor, the instructions being configured to cause the processor to perform the method according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The device stores computer-executable instructions configured to perform the method as described in any one of claims 1 to 8.

Images