Coal quality characteristic-based circulating fluidized bed boiler secondary air control method and system

Abstract

The application discloses a circulating fluidized bed boiler secondary air control method and system based on coal quality characteristics, relates to the technical field of automatic control, solves the technical problem that the existing manual control secondary air strategy makes it difficult to adjust the secondary air and match the secondary air with the furnace environment in time, and further causes the low efficiency of secondary air regulation, and comprises the following steps: acquiring industrial data of a fluidized bed boiler, constructing a digital twin model of the fluidized bed boiler based on the industrial data; acquiring coal quality data and furnace data of the coal to be added; inputting the coal quality data and the furnace data into the digital twin model to obtain corresponding simulated furnace data; generating a secondary air control scheme based on the simulated furnace data; regulating and controlling the fluidized bed boiler based on the secondary air control scheme; acquiring real-time furnace data, generating a secondary air adjustment scheme based on the real-time furnace data; and regulating and controlling the fluidized bed boiler based on the secondary air adjustment scheme; and the overall operation efficiency of the boiler is improved.

Description

Technical Field

[0001] This application belongs to the field of automatic control technology, specifically a method and system for controlling secondary air in a circulating fluidized bed boiler based on coal quality characteristics. Background Technology

[0002] Circulating fluidized bed (CFB) boilers are widely used in industrial power generation and heating due to their advantages such as strong fuel adaptability, high combustion efficiency, and low pollutant emissions. In the combustion control of CFB boilers, the proper proportion of secondary air has a decisive impact on in-furnace combustion efficiency, bed temperature stability, and pollutant emissions.

[0003] Existing secondary air regulation mainly relies on manual adjustments by operators based on experience; for example, adjusting different secondary air supply strategies according to the specific combustion conditions in the furnace. However, the combustion process of circulating fluidized bed boilers is characterized by high inertia, nonlinearity, and strong coupling. Differences in the quality of the coal entering the furnace, such as volatile matter, fixed carbon, and particle size distribution, increase the difficulty of control. Relevant operators often cannot respond to the furnace status in a timely manner. Secondly, when the quality of the added coal changes, the change in the combustion conditions in the furnace requires a long thermal inertia transfer process before it is reflected in the measurable parameters. This results in an inherent response lag in feedback control based on real-time parameters. By the time the controller detects a deviation, the actual combustion state has already significantly deviated from the target conditions. At this point, relying solely on manual control makes it difficult for secondary air regulation to match the furnace environment in a timely manner, leading to a decrease in the accuracy of secondary air regulation. Therefore, a secondary air control method for circulating fluidized bed boilers based on coal quality characteristics is needed. Summary of the Invention

[0004] This application provides a method and system for controlling secondary air in a circulating fluidized bed boiler based on coal quality characteristics. It solves the technical problem that existing manual control strategies for secondary air make it difficult to match secondary air regulation with the furnace environment in a timely manner, resulting in low efficiency of secondary air regulation.

[0005] To achieve the above objectives, this application adopts the following technical solution: The first aspect provides a method for controlling secondary air in circulating fluidized bed boilers based on coal quality characteristics, including: Acquire industrial data of the fluidized bed boiler, and construct a digital twin model of the fluidized bed boiler based on the industrial data; The process involves acquiring coal quality data and furnace data for the coal to be added. The coal quality data includes physical and chemical properties of the coal, with the physical properties including several physical parameters such as addition amount and median particle size distribution. The chemical properties include several chemical parameters such as volatile matter and fixed carbon. The coal quality data and furnace data are then input into a digital twin model to obtain corresponding simulated furnace data. A secondary air control scheme is generated based on the simulated furnace data. The fluidized bed boiler is then regulated based on the secondary air control scheme, and coal is added accordingly. Acquire real-time furnace data and generate a secondary air adjustment plan based on the real-time furnace data; then regulate the fluidized bed boiler based on the secondary air adjustment plan.

[0006] Based on the above technical solution, the circulating fluidized bed boiler secondary air control method and system based on coal quality characteristics provided in this application involves: acquiring industrial data of the fluidized bed boiler; constructing a digital twin model of the fluidized bed boiler based on the industrial data; acquiring coal quality data and furnace data of the coal to be added; inputting the coal quality data and furnace data into the digital twin model to obtain corresponding simulated furnace data; generating a secondary air control scheme based on the simulated furnace data; regulating the fluidized bed boiler according to the secondary air control scheme; acquiring real-time furnace data; generating a secondary air adjustment scheme based on the real-time furnace data; and regulating the fluidized bed boiler according to the secondary air adjustment scheme. This not only ensures timely control and adjustment of the secondary air but also guarantees the continuous stability of combustion during coal quality switching, effectively avoiding frequent oscillations of the actuators caused by instantaneous parameter fluctuations, extending the service life of key equipment such as dampers and fans, and improving the overall operating efficiency of the boiler.

[0007] In conjunction with the first aspect above, in one possible implementation, a secondary air control scheme is generated based on the simulated furnace data, including: Extract the parameter value set corresponding to each detection parameter from the simulated furnace data; generate a stable feature vector based on the parameter value set corresponding to each detection parameter. The stable feature vector is input into the secondary wind control scheme generation model to obtain the corresponding secondary wind control scheme; the secondary wind control scheme generation model is obtained by training an artificial intelligence model; the secondary wind control scheme includes the total secondary wind volume, the fan speed and the damper opening.

[0008] In conjunction with the first aspect above, in one possible implementation, generating a stable feature vector based on the parameter value group corresponding to each detection parameter includes: Extract the parameter values ​​collected at each time point from each parameter value group, and integrate the parameter values ​​corresponding to each detection parameter at the same time point into a feature vector according to a set order. The feature vectors are arranged in chronological order of their corresponding acquisition times. A predetermined number of feature vectors are obtained as the basic clustering vectors. The remaining feature vectors are divided into several perturbation vector groups. Each perturbation vector group includes the same number of feature vectors. The same number is set by experts based on experience, and this same number is less than the number of the previous basic clustering vectors. Cluster analysis is performed on each basic clustering vector to obtain several clusters, which are denoted as basic clusters. Feature vectors from perturbation vector groups are added sequentially to each basic clustering vector, and cluster analysis is performed to obtain several corresponding clusters. Clusters that do not have complete consistency among the basic clusters are denoted as perturbation clusters. Stable feature vectors are generated based on several basic clusters and perturbation clusters.

[0009] In conjunction with the first aspect above, in one possible implementation, the generation of stable feature vectors based on several basic clusters and slightly disturbed dynamic clusters includes: Obtain the center vectors of several basic clusters and denote them as basic center vectors; obtain the center vectors corresponding to the disturbed clusters, calculate the Euclidean distance between each center vector and the basic center vectors, and classify the disturbed clusters corresponding to the center vectors whose Euclidean distance is less than a set distance threshold into the cluster evolution group corresponding to the basic clusters; Obtain the number of clusters in each cluster evolution group, identify the cluster evolution groups with a number of clusters greater than a set evolution number threshold, and extract the center vectors corresponding to the disturbed clusters in the cluster evolution groups. Calculate the dispersion of each center vector, including variance and standard deviation. The larger the variance, the greater the dispersion, and the more dispersed the corresponding center vectors. The cluster evolution group with the smallest dispersion is taken as the stable evolution group, and the mean vector of the center vectors corresponding to each disturbed cluster in the stable evolution group is recorded as the stable feature vector.

[0010] In conjunction with the first aspect above, in one possible implementation, the secondary wind control scheme generation model is obtained through training an artificial intelligence model, including: A number of historical furnace data and corresponding optimal secondary air control schemes are obtained. The historical furnace data includes several sets of parameter values ​​corresponding to detection parameters. The number and types of detection parameters in the historical furnace data, simulated furnace data, and actual furnace data are the same. The secondary air control scheme is the most suitable secondary air control scheme for the current furnace data obtained by experts adjusting the secondary air based on the historical furnace data. It includes several control parameters, such as the total secondary air volume, fan speed, and damper opening. Based on the parameter value sets corresponding to the detection parameters in the historical furnace data, a stable feature vector corresponding to the historical furnace data is generated. The stable feature vectors and their corresponding secondary air control schemes are integrated into several training data and test data. The artificial intelligence model is trained using training data and tested using testing data. The final result is a secondary wind control scheme generation model with stable feature vectors as input and secondary wind control schemes corresponding to the stable feature vectors as output.

[0011] In conjunction with the first aspect above, in one possible implementation, the generation of the secondary air adjustment scheme based on real-time furnace data includes: Extract the parameter value groups corresponding to each detection parameter in the real-time furnace data. The parameter value groups corresponding to each detection parameter in the real-time furnace data only include a set number of parameter values, and must include the parameter value at the current time. Extract the parameter values ​​at each time from each parameter array. Representative detection parameters are obtained by selecting the parameter values ​​corresponding to several times in the parameter value group corresponding to each detection parameter. Obtain a stable feature vector, which is used to generate a secondary wind control scheme; extract the parameter values ​​corresponding to the detection parameters in the stable feature vector and mark them as reference values ​​corresponding to the detection parameters; A secondary wind adjustment plan is generated based on the current parameter values ​​and reference values ​​of the representative detection parameters.

[0012] In conjunction with the first aspect above, in one possible implementation, the step of selecting representative detection parameters based on parameter values ​​corresponding to several time points in the parameter value group corresponding to each detection parameter includes: By substituting the parameter values ​​corresponding to several time points in the parameter value group corresponding to the detection parameter into the set stable change evaluation function, the degree of change corresponding to the detection parameter is obtained; one expression of the stable change evaluation function includes: ; in, The degree of variability represents the stability of changes in the detection parameter. The smaller the corresponding value, the more stable the change of the corresponding detection parameter, and the stronger its reference value; To detect the parameter corresponding to the parameter value group, the first... The parameter values ​​at time, and , This represents the number of parameter values ​​in the parameter value group. This represents the time interval between the acquisition of adjacent parameter values. The detection parameter corresponding to the smallest degree of change is selected as the representative detection parameter.

[0013] In conjunction with the first aspect above, in one possible implementation, generating the secondary wind adjustment scheme based on the parameter value representing the current moment of the detection parameter and the reference value includes: Get the current value of the parameter representing the detection parameter. and reference values Substituting these values ​​into the set adjustment ratio generation function yields the adjustment ratios corresponding to each control parameter; one expression of the adjustment ratio function includes: ; in, For the adjustment ratio of the control parameter corresponding to number n, The set unit deviation value, The unit adjustment ratio is set for the control parameter corresponding to number n. When the adjustment ratio When the value is positive, it indicates that the control parameter corresponding to number n needs to be adjusted in the positive direction, such as increasing the corresponding proportion of the total secondary air volume. When the value is negative, it means that the control parameter corresponding to number n needs to be adjusted negatively, such as reducing the proportion of the total secondary air volume. The adjustment ratios corresponding to each control parameter are integrated into a secondary air adjustment scheme.

[0014] In conjunction with the first aspect above, in one possible implementation, constructing a digital twin model of the fluidized bed boiler based on the industrial data includes: The entity data and component association mapping data of the fluidized bed boiler are extracted from industrial data. A three-dimensional model of the fluidized bed boiler is constructed based on the entity data. The association mapping relationship of each component in the three-dimensional model of the fluidized bed boiler is established based on the component association mapping data, thus obtaining a digital twin model of the fluidized bed boiler.

[0015] Secondly, this application provides a secondary air control system for a circulating fluidized bed boiler based on coal quality characteristics, comprising: a data acquisition module, a data analysis module, and a control module; wherein, The data acquisition module includes a first acquisition unit and a second acquisition unit; The first acquisition unit is used to acquire coal quality data of the coal to be added; The second acquisition unit is used to acquire industrial data of fluidized bed boilers and corresponding in-furnace data of fluidized bed boilers; The data analysis module includes a first analysis unit and a second analysis unit; The first analysis unit is used to construct a digital twin model of the fluidized bed boiler based on the industrial data, and input the coal quality data and furnace data into the digital twin model to perform simulation operation and obtain the corresponding simulated furnace data; The second analysis unit is used to generate a secondary air control scheme based on the simulated furnace data, and to acquire real-time furnace data and generate a secondary air adjustment scheme based on the real-time furnace data. The control module is used to regulate the fluidized bed boiler based on the secondary air control scheme and to regulate the fluidized bed boiler based on the secondary air adjustment scheme.

[0016] This application provides a secondary air control method for circulating fluidized bed boilers based on coal quality characteristics. It can construct a digital twin model of the fluidized bed boiler and perform advanced simulation using coal quality data of the coal to be added. Before coal is introduced into the boiler, the rapid extrapolation capability of the digital twin allows for the pre-calculation of a secondary air control scheme adapted to the new coal quality conditions, fundamentally overcoming the response lag problem caused by the large inertia of the combustion system. Furthermore, incremental cluster analysis is used to distinguish between the basic state and the disturbance state, and the dispersion of cluster evolution is used as a stability criterion to select stable feature vectors. The principle is that if a certain type of operating condition can maintain a high cluster density and a low center drift after introducing coal quality disturbances, it indicates that the operating condition is a steady-state attractor of the system under given coal quality conditions. Taking its mean vector can filter out transient noise during the simulation process to the greatest extent, thereby improving the accuracy and robustness of the generated control scheme. At the real-time control level, the fluctuation of the rate of change of each detection parameter is quantified by a stable change evaluation function. The representative detection parameter with the highest signal-to-noise ratio is dynamically selected as the main control variable, and the corresponding value in the stable feature vector is used as the dynamic reference value. Combined with the deviation-based adjustment ratio generation function, the correction amount is distributed proportionally to each secondary air actuator, realizing adaptive fine-tuning with steady-state characteristics as the anchor point. This dual closed-loop mechanism of "feedforward simulation to set the benchmark and real-time feedback to correct deviation" not only ensures timely control and adjustment of the secondary air, but also ensures the continuous stability of combustion during coal quality switching. It effectively avoids frequent oscillations of actuators caused by instantaneous parameter fluctuations, extends the service life of key equipment such as dampers and fans, and improves the overall operating efficiency of the boiler.

[0017] It should be understood that the descriptions of technical features, technical solutions, beneficial effects, or similar language in this application do not imply that all features and advantages can be achieved in any single embodiment. Rather, it is understood that the description of a feature or beneficial effect means that a specific technical feature, technical solution, or beneficial effect is included in at least one embodiment. Therefore, the descriptions of technical features, technical solutions, or beneficial effects in this specification do not necessarily refer to the same embodiment. Furthermore, the technical features, technical solutions, and beneficial effects described in this embodiment can be combined in any suitable manner. Those skilled in the art will understand that embodiments can be implemented without one or more specific technical features, technical solutions, or beneficial effects of a particular embodiment. In other embodiments, additional technical features and beneficial effects may be identified in specific embodiments that do not embody all embodiments. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art 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.

[0019] Figure 1 This is a schematic diagram of the steps in the secondary air control method for a circulating fluidized bed boiler in this application; Figure 2 This is a schematic diagram of the module connections of the secondary air control system for the circulating fluidized bed boiler in this application. Detailed Implementation

[0020] The technical solutions of this application will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0021] Please see Figure 1 The first aspect of this application provides a method for controlling secondary air in a circulating fluidized bed boiler based on coal quality characteristics, including: The process involves acquiring industrial data of a fluidized bed boiler and constructing a digital twin model of the boiler based on this data. Specifically, industrial data forms the basis for building a high-fidelity digital twin, and its sources are diverse. For example, industrial data may include historical operating data from a distributed control system (DCS), design parameters from equipment ledgers, and real-time data collected by field sensors. During the construction process, not only is a three-dimensional geometric model of the boiler established using physical data, but more importantly, physical and logical coupling relationships between components are established based on component association mapping data, such as the aerodynamic relationship between the fan, damper, and furnace, and the thermodynamic relationship between coal quality, combustion, and heat load. It should be understood that although this embodiment uses a three-dimensional visualization model as an example, in other embodiments, the digital twin model can also be represented as a multi-dimensional mathematical mechanism model or a reduced-order model, as long as it has the simulation and deduction capability to input operating conditions and output simulated response data. This step provides a reliable virtual verification environment for subsequent advanced simulations.

[0022] The process involves acquiring coal quality data and furnace data for the coal to be added. The coal quality data includes physical and chemical properties of the coal, with the physical properties including several physical parameters such as addition amount and median particle size distribution. The chemical properties include several chemical parameters such as volatile matter and fixed carbon. The coal quality data and furnace data are then input into a digital twin model to obtain corresponding simulated furnace data. A secondary air control scheme is generated based on the simulated furnace data. The fluidized bed boiler is then regulated based on the secondary air control scheme, and coal is added accordingly. Specifically, coal quality data covers key indicators affecting combustion characteristics, such as the amount added and median particle size distribution in physical properties, and volatile matter and fixed carbon content in chemical properties. Furnace data reflects the current combustion baseline state, such as bed temperature, bed pressure, and oxygen content. After inputting these data into the digital twin model, the model can quickly deduce the combustion evolution trend after the new coal type is added to the furnace in virtual space, thereby obtaining simulated furnace data. The secondary air control scheme generated based on this simulated data is essentially a pre-control strategy for future operating conditions. In actual control, the system first adjusts the secondary air to the target value set in the pre-control scheme, such as adjusting the total amount of secondary air, the fan speed, or the damper opening, and then starts the coal feeder to add new coal. This "air-coal-waiting" control sequence effectively overcomes the large inertia lag problem of the circulating fluidized bed boiler combustion system and avoids drastic fluctuations in bed temperature or incomplete combustion caused by changes in coal quality. It should be noted that in actual engineering applications, the two can also be carried out simultaneously or with a slight overlap in timing, as long as the adjustment of the secondary air is based on the prediction of the new coal quality rather than a reactive response.

[0023] The system acquires real-time furnace data and generates a secondary air adjustment scheme based on this data. The fluidized bed boiler is then regulated based on this secondary air adjustment scheme. Although digital twin models can provide high-precision predictions, numerous unmeasurable disturbances exist in actual industrial settings, and the model itself has simplification errors. For example, the coal quality varies significantly within a batch, leading to slight differences between the coal input and the given coal quality data for each batch. This makes it difficult to achieve optimal results using only feedforward control. Therefore, this embodiment introduces a real-time feedback correction mechanism. Real-time furnace data refers to the combustion state parameters actually collected by sensors after new coal is introduced into the furnace. The system compares the real-time data with the expected state, dynamically generates a secondary air adjustment scheme, and fine-tunes the results of the feedforward control. For example, when the real-time bed temperature is lower than the simulated predicted value, the adjustment scheme automatically increases the secondary air volume to enhance combustion; conversely, it reduces the air volume. This adjustment is continuous or periodic, ensuring that the control strategy can adaptively track changes in actual operating conditions.

[0024] In one possible implementation, a secondary air control scheme is generated based on simulated furnace data, including: extracting parameter value groups corresponding to each detected parameter in the simulated furnace data; generating a stable feature vector based on the parameter value groups corresponding to each detected parameter; inputting the stable feature vector into a secondary air control scheme generation model to obtain the corresponding secondary air control scheme; the secondary air control scheme generation model is obtained through training an artificial intelligence model; the secondary air control scheme includes the total secondary air volume, fan speed, and damper opening. Specifically, the simulated furnace data output by the digital twin model is usually a high-dimensional time-series signal containing rich transient information. If the raw, noisy simulation data is directly used as input to the artificial intelligence model, it is easy to cause frequent oscillations in the generated control commands, which cannot reflect the essential steady-state requirements of the combustion process. Therefore, this invention introduces a stable feature vector as an intermediate representation layer. This stable feature vector is not a simple statistical average, but a mathematical expression extracted from the dynamically evolving simulation data through a specific incremental clustering algorithm, which can characterize the final convergence state of the system under given coal quality conditions. In this way, transient noise in the simulation process can be effectively filtered out, providing robust and physically meaningful input features for subsequent AI decision-making, thereby ensuring that the generated control parameters such as total secondary air volume, fan speed, and damper opening are both accurate and stable.

[0025] In one possible implementation, a stable feature vector is generated based on the parameter value group corresponding to each detection parameter, including: extracting the parameter values ​​collected at each time in each parameter value group, and integrating the parameter values ​​corresponding to each detection parameter at the same time into a feature vector in a set order. The feature vectors are arranged in chronological order of their corresponding acquisition times. A predetermined number of feature vectors are obtained as the basic clustering vectors. The remaining feature vectors are divided into several perturbation vector groups. Each perturbation vector group includes the same number of feature vectors. The same number is set by experts based on experience, and this same number is less than the number of the previous basic clustering vectors. Cluster analysis is performed on each basic clustering vector to obtain several clusters, which are denoted as basic clusters. Feature vectors from perturbation vector groups are added sequentially to each basic clustering vector, and cluster analysis is performed to obtain several corresponding clusters. Clusters that do not have complete consistency among the basic clusters are denoted as perturbation clusters. Stable feature vectors are generated based on several basic clusters and perturbation clusters.

[0026] In this process, it is first necessary to understand the physical meaning of "temporal integration". The combustion state of a circulating fluidized bed boiler is determined by multiple parameters, and a single parameter cannot fully describe the operating conditions. Therefore, by splicing multiple detection parameters such as bed temperature, bed pressure, oxygen content, and CO concentration at the same moment into a multi-dimensional feature vector in a fixed order, a state point in the boiler combustion state space is actually constructed. As time goes by, these state points form a trajectory in space.

[0027] In practice, a predetermined number of feature vectors, such as the first 30% or the first 500 sampling points, are selected as the basic clustering vectors. This data typically corresponds to the initial stage of simulation or a relatively stable baseline condition, used to establish the system's "background state" or "initial attractor." The remaining feature vectors are divided equally to form a set of perturbation vectors, which simulates the continuous excitation experienced by the system after changes in coal quality characteristics. This division is not arbitrary but based on the physical characteristics of the large inertial system of a CFB boiler: the system's response to new coal quality is a gradual process, and subsequent perturbation vector sets actually record the dynamic path of the system's migration from the old steady state to the new steady state. By re-clustering after each perturbation vector set is added and identifying the "perturbation clusters" that have changed relative to the basic clusters, the algorithm can capture the subtle drifts in the system state caused by the influence of coal quality. This incremental processing method, compared to traditional full-scale one-time clustering, better preserves causal evolution information in the time dimension and avoids the ambiguity of steady-state characteristics caused by data aliasing.

[0028] Building upon this, in order to accurately pinpoint the true steady state from numerous evolutionary states, this embodiment further defines a stability discrimination mechanism; specifically, it generates stable feature vectors based on several basic clusters and slightly disturbed dynamic clusters, including: Obtain the center vectors of several basic clusters and denote them as basic center vectors; obtain the center vectors corresponding to the disturbed clusters, calculate the Euclidean distance between each center vector and the basic center vectors, and classify the disturbed clusters corresponding to the center vectors whose Euclidean distance is less than a set distance threshold into the cluster evolution group corresponding to the basic clusters; The number of clusters in each cluster evolution group is obtained. Cluster evolution groups with a number of clusters greater than a set evolution number threshold are obtained. The center vectors corresponding to the disturbed clusters in the cluster evolution groups are extracted, and the dispersion of each center vector is calculated. The dispersion includes variance, standard deviation, etc. For example, the larger the variance, the greater the dispersion, and the more dispersed the corresponding center vectors are. The cluster evolution group corresponding to the minimum dispersion is taken as the stable evolution group, and the mean vector of the center vectors corresponding to each disturbed cluster in the stable evolution group is recorded as the stable feature vector. It can be understood that since most of the feature vectors selected in the clustering data used in the basic clusters will only affect some clusters when the subsequent disturbed vectors are added for re-clustering, and this effect is generally small. If a basic cluster is constantly evolving and the center vector of the cluster is always in a stable state, it means that the parameters in the simulation furnace data are tending to a stable state, and the center vector of this cluster is approximately in this stable state. The Euclidean distance threshold acts as a "state association filter"; only when the center of a disturbance cluster is sufficiently close to the base center is it considered an "evolution" of the base state rather than a jump to a completely new and unrelated operating mode; this ensures that we are tracking the dynamic adjustment process of the same combustion mode. Secondly, the evolution quantity threshold is used to eliminate random disturbances; if a state appears only in a very small number of disturbance groups, it may be merely a numerical oscillation in the simulation or a very brief transitional state, lacking representativeness as a control baseline; only those evolution groups that persist through multiple consecutive disturbance stages are considered the true response trend of the system. Finally, "minimum dispersion" is the core criterion for identifying steady-state attractors. In a dynamic system, steady state means that the system has the ability to self-recover or converge after being disturbed. Dispersion, such as variance or standard deviation, quantifies the degree of dispersion of cluster centers during the evolution process. The smaller the dispersion, the more closely the system state fluctuates around a specific value. Conversely, evolution groups with large dispersion indicate that the system is still in a state of violent adjustment or non-convergence. Therefore, selecting the evolution group corresponding to the minimum dispersion and using the mean of its center vector as the final stable feature vector essentially extracts the target operating point of boiler combustion under the coal quality condition from complex nonlinear simulation data. It should be understood that although this embodiment uses Euclidean distance and variance as examples, in other embodiments, Mahalanobis distance, cosine similarity, or other metrics such as entropy and coefficient of variation can also be used to evaluate the degree of dispersion, as long as they can achieve the technical objective of "screening out the state evolution trajectory with the strongest convergence".

[0029] In one possible implementation, the secondary wind control scheme generation model is obtained through training an artificial intelligence model, including: A number of historical furnace data and corresponding optimal secondary air control schemes are obtained. The historical furnace data includes several sets of parameter values ​​corresponding to detection parameters. The number and types of detection parameters in the historical furnace data, simulated furnace data, and actual furnace data are the same. The secondary air control scheme is the most suitable secondary air control scheme for the current furnace data obtained by experts adjusting the secondary air based on the historical furnace data. It includes several control parameters, such as the total secondary air volume, fan speed, and damper opening. Based on the parameter value sets corresponding to the detection parameters in the historical furnace data, a stable feature vector corresponding to the historical furnace data is generated. The stable feature vectors and their corresponding secondary air control schemes are integrated into several training data and test data. The artificial intelligence model is trained using training data and tested using testing data. The final result is a secondary wind control scheme generation model with stable feature vectors as input and secondary wind control schemes corresponding to the stable feature vectors as output. The artificial intelligence model includes deep neural network models or recurrent neural network models, etc.

[0030] Specifically, the model training in this embodiment is not simply mathematical fitting, but a process of digitally encapsulating expert experience and control laws from the industrial field. The optimal secondary air control scheme serves as a label for supervised learning, and its source has clear technical attributes. For example, this scheme could be a secondary air adjustment record selected from the historical database of a distributed control system (DCS) corresponding to the period when boiler combustion efficiency was highest and pollutant emissions met standards under specific coal quality conditions; or it could be an ideal control strategy derived by senior operation experts based on historical furnace data. This ensures that the model learns not simple statistical correlations, but control logic that conforms to the combustion mechanism. This means that the model's input space eliminates transient noise interference, directly representing the steady-state nature of combustion, thereby significantly reducing the model's learning difficulty and dependence on the number of samples.

[0031] During the data preparation phase, the integrated dataset is divided into training data and validation data according to a preset ratio, such as 8:2 or 7:3. The training data is used to drive the artificial intelligence model to update its weight parameters, gradually approximating the nonlinear mapping relationship from the stable feature vector to the secondary wind control scheme. The validation data is independent of the training process and is specifically used to evaluate the model's generalization ability under unseen operating conditions, preventing overfitting. It should be understood that although this embodiment uses a fixed ratio as an example, in practical applications, more rigorous evaluation strategies such as cross-validation can also be used to ensure the robustness of the model's performance.

[0032] Regarding the specific selection of the artificial intelligence model, this invention is not limited to a single architecture. For example, deep neural networks (DNNs) can be used to capture the complex nonlinear relationship between high-dimensional feature vectors and multivariable control outputs; or, considering the time-dependent nature of the combustion process, recurrent neural networks (RNNs) and their variants (such as LSTM and GRU) can be used so that the model can better understand the dynamic evolution trend implicit within the stable feature vectors. Regardless of the specific model structure used, its core function is to establish a bridge from "simulated steady-state characteristics" to "physical control commands." The model's output is explicitly defined as physical quantities that can directly affect the actuators, such as the total secondary air volume, fan speed, and damper opening. Through the above training and verification process, the final secondary air control scheme generation model is essentially a soft measurement controller that integrates expert knowledge and data-driven advantages. It can automatically infer a matching secondary air pre-control strategy based on the new coal quality steady-state characteristics predicted by the digital twin model within milliseconds, thereby solving the technical problems of traditional control methods being unable to cope with frequent coal quality fluctuations, and the lag and insufficient accuracy of manual adjustments.

[0033] In one possible implementation, a secondary air adjustment scheme is generated based on real-time furnace data, including: extracting parameter value groups corresponding to each detection parameter in the real-time furnace data, wherein the parameter value groups corresponding to each detection parameter in the real-time furnace data only include a set number of parameter values, and must include the parameter values ​​at the current moment; and extracting the parameter values ​​at each moment from each parameter array. Representative detection parameters are obtained by selecting the parameter values ​​corresponding to several times in the parameter value group corresponding to each detection parameter. Obtain a stable feature vector, which is used to generate a secondary wind control scheme; extract the parameter values ​​corresponding to the detection parameters in the stable feature vector and mark them as reference values ​​corresponding to the detection parameters; A secondary wind adjustment plan is generated based on the current parameter values ​​and reference values ​​of the representative detection parameters.

[0034] Specifically, this embodiment constructs a real-time feedback correction path that runs parallel to and complements feedforward control. In the aforementioned embodiment, the system has generated a secondary air pre-control scheme adapted to the new coal quality through digital twin simulation. However, due to the presence of sensor noise, unmodeled disturbances, and random fluctuations in the microscopic properties of coal in actual industrial settings, it is difficult to maintain a long-term accurate steady state solely through feedforward. Therefore, a dynamic adjustment mechanism based on real-time data is introduced. It should be noted that the "reference value" used for comparison here is not the fixed setpoint in traditional control, but a stable eigenvector. This means that the goal of feedback control is to dynamically track the theoretical steady-state point predicted by the digital twin and adapted to the current coal quality characteristics. This design makes real-time adjustment no longer a blind regression to an empirical constant, but a purposeful approximation of the optimal operating condition derived from simulation, thereby achieving seamless integration of feedforward prediction and feedback correction at the data level.

[0035] Furthermore, in order to accurately identify the dominant variable that best characterizes the current combustion state from numerous real-time monitoring parameters, this embodiment employs a parameter screening mechanism based on stability assessment. As one implementation method, in one possible approach, the selection of representative detection parameters based on parameter values ​​corresponding to several time points within the parameter value groups corresponding to each detection parameter includes: By substituting the parameter values ​​corresponding to several time points in the parameter value group corresponding to the detection parameter into the set stable change evaluation function, the degree of change corresponding to the detection parameter is obtained; one expression of the stable change evaluation function includes: ; in, The degree of variability represents the stability of changes in the detection parameter. The smaller the corresponding value, the more stable the change of the corresponding detection parameter, and the stronger its reference value; To detect the parameter corresponding to the parameter value group, the first... The parameter values ​​at time, and , This represents the number of parameter values ​​in the parameter value group. This represents the time interval between the acquisition of adjacent parameter values. The detection parameter corresponding to the smallest degree of change is selected as the representative detection parameter.

[0036] In practical implementation, the system iterates through all available detection parameters, such as bed temperature, bed pressure, oxygen content, CO concentration, and furnace negative pressure, calculates their values ​​within the most recent time window, and automatically selects the parameter with the smallest value as the representative detection parameter. For example, during the transition phase after a coal quality change, if the oxygen content signal fluctuates violently due to local turbulence, while the bed temperature also changes but its rate of change maintains a relatively uniform upward or downward trend, the system will determine that the bed temperature is smaller and select it as the representative parameter. The advantage of this dynamic screening mechanism is that it can adaptively find the "anchor point" variable with the highest signal-to-noise ratio and best characterize the essential state of combustion under different operating conditions, effectively avoiding erroneous adjustments caused by individual sensor failures or abnormal fluctuations in non-critical parameters, and significantly improving the robustness of feedback control. It should be understood that although this embodiment uses the standard deviation of the rate of change as the evaluation index, in other embodiments, the coefficient of variation of the rate of change, entropy value, or other statistical quantities that can characterize the stationarity of time-series signals can also be used to achieve the same technical purpose.

[0037] In one possible implementation, generating the secondary wind adjustment scheme based on the parameter value representing the current moment of the detection parameter and the reference value includes: obtaining the current parameter value representing the detection parameter. and reference values Substituting these values ​​into the set adjustment ratio generation function yields the adjustment ratios corresponding to each control parameter; one expression of the adjustment ratio function includes: ; in, For the adjustment ratio of the control parameter corresponding to number n, The set unit deviation value, The unit adjustment ratio is set for the control parameter corresponding to number n. When the adjustment ratio When the value is positive, it indicates that the control parameter corresponding to number n needs to be adjusted in the positive direction, such as increasing the corresponding proportion of the total secondary air volume. When the value is negative, it means that the control parameter corresponding to number n needs to be adjusted negatively, such as reducing the proportion of the total secondary air volume. The adjustment ratios corresponding to each control parameter are integrated into a secondary air adjustment scheme.

[0038] In one possible implementation, constructing a digital twin model of the fluidized bed boiler based on industrial data includes: extracting entity data and component association mapping data of the fluidized bed boiler from industrial data; constructing a three-dimensional model of the fluidized bed boiler based on the entity data; and establishing the association mapping relationship between each component in the three-dimensional model of the fluidized bed boiler based on the component association mapping data, thereby obtaining the digital twin model of the fluidized bed boiler. It is understood that the construction of the digital twin model is a relatively existing technology and will not be elaborated on further here.

[0039] Specifically, the digital twin model in this embodiment is not merely a static 3D model for visual display, but a virtual-real mapping system with dynamic simulation and deduction capabilities. Its construction is based on the in-depth mining and structured reorganization of "industrial data." This industrial data has multiple sources, including historical operating data accumulated over a long period in distributed control systems (DCS), design ledgers and nameplate parameters of equipment at the time of manufacture, real-time data collected by field sensors, and CAD design drawings of the boiler body and auxiliary equipment. These data collectively constitute the "gene pool" of the digital twin. Based on this, physical data is mainly used to construct the 3D geometric model of the fluidized bed boiler. This includes not only the spatial location and dimensions of key components such as the furnace, separator, return feeder, secondary air fan, and dampers, but also physical parameters such as material properties, wall thickness, and volume, providing accurate geometric boundary conditions for subsequent heat transfer and flow calculations. More crucially, the processing of component association mapping data is essential. This data defines the logical and physical coupling relationships between the various components within the 3D model, and is the core element that gives the model "life." For example, at the aerodynamic correlation level, a fluid dynamic mapping needs to be established between the secondary fan speed, damper opening, and furnace inlet air pressure and velocity; at the thermal correlation level, a combustion heat release mapping needs to be established between coal feed rate, coal quality characteristics such as volatile matter and calorific value, and bed temperature and steam load; at the control correlation level, the transfer function and time lag characteristics between actuator actions and sensor responses also need to be reflected. By embedding these correlation mapping relationships into the three-dimensional model, the digital twin can, like a real boiler, dynamically deduce the future furnace state evolution based on built-in physical laws and data-driven models when it receives coal quality data to be added and current furnace data as input excitation, thereby outputting high-fidelity simulated furnace data. It should be understood that although this embodiment uses a three-dimensional geometric model combined with physical mechanism mapping as an example for explanation, in other embodiments, the digital twin model can also adopt a reduced-order model, a pure data-driven neural network model, or a hybrid model that combines the two, as long as it can achieve the function of predicting operating conditions based on coal quality characteristics. This refined construction approach ensures that the generated secondary air pre-control scheme is based on an accurate prediction of the boiler's actual physical response, rather than theoretical calculations divorced from reality, thereby fundamentally improving the engineering applicability and robustness of the control strategy.

[0040] Please see Figure 2 Secondly, this application provides a secondary air control system for a circulating fluidized bed boiler based on coal quality characteristics, comprising: a data acquisition module, a data analysis module, and a control module; wherein, The data acquisition module includes a first acquisition unit and a second acquisition unit; The first acquisition unit is used to acquire coal quality data of the coal to be added; The second acquisition unit is used to acquire industrial data of fluidized bed boilers, as well as corresponding in-furnace data of fluidized bed boilers; The data analysis module includes a first analysis unit and a second analysis unit; The first analysis unit is used to construct a digital twin model of the fluidized bed boiler based on the industrial data, and input the coal quality data and furnace data into the digital twin model to perform simulation operation and obtain the corresponding simulated furnace data. The second analysis unit is used to generate a secondary air control scheme based on the simulated furnace data, and to acquire real-time furnace data and generate a secondary air adjustment scheme based on the real-time furnace data. The control module is used to regulate the fluidized bed boiler based on the secondary air control scheme and the secondary air adjustment scheme.

[0041] Some of the data in the above formula are calculated by removing dimensions and taking their numerical values. The formula is the closest to the real situation obtained by software simulation of a large amount of collected data. The preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.

[0042] How this application works: By constructing a digital twin model of a fluidized bed boiler and combining it with coal quality data of the coal to be added for advanced simulation, a secondary air control scheme adapted to the new coal quality conditions can be pre-calculated using the rapid extrapolation capability of the digital twin before the coal is introduced into the furnace. This fundamentally overcomes the response lag problem caused by the large inertia of the combustion system. On this basis, incremental cluster analysis is used to distinguish between the basic state and the disturbance state, and the dispersion of the cluster evolution is used as a stability criterion to select stable feature vectors. The principle is that if a certain type of operating condition can still maintain a high cluster density and a low center drift after the introduction of coal quality disturbance, it indicates that the operating condition is a steady-state attractor of the system under the given coal quality conditions. Taking its mean vector can filter out transient noise in the simulation process to the greatest extent, thereby improving the accuracy and robustness of the control scheme generation. At the real-time control level, the fluctuation of the rate of change of each detection parameter is quantified by a stable change evaluation function. The representative detection parameter with the highest signal-to-noise ratio is dynamically selected as the main control variable, and the corresponding value in the stable feature vector is used as the dynamic reference value. Combined with the deviation-based adjustment ratio generation function, the correction amount is distributed proportionally to each secondary air actuator, realizing adaptive fine-tuning with steady-state characteristics as the anchor point. This dual closed-loop mechanism of "feedforward simulation to set the benchmark and real-time feedback to correct deviation" not only ensures timely control and adjustment of the secondary air, but also ensures the continuous stability of combustion during coal quality switching. It effectively avoids frequent oscillations of actuators caused by instantaneous parameter fluctuations, extends the service life of key equipment such as dampers and fans, and improves the overall operating efficiency of the boiler.

[0043] The above embodiments are only used to illustrate the technical methods of this application and are not intended to limit it. Although this application has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of this application without departing from the spirit and scope of the technical methods of this application.

Claims

1. A method for controlling secondary air in a circulating fluidized bed boiler based on coal quality characteristics, characterized in that, include: Acquire industrial data of the fluidized bed boiler, and construct a digital twin model of the fluidized bed boiler based on the industrial data; Obtain coal quality data and furnace data for the coal to be added; The coal quality data and furnace in-furnace data are input into the digital twin model to obtain the corresponding simulated furnace in-furnace data; A secondary air control scheme is generated based on the simulated furnace data. Fluidized bed boiler is regulated based on a secondary air control scheme; Acquire real-time furnace internal data and generate a secondary air adjustment plan based on the real-time furnace internal data; The fluidized bed boiler is regulated based on the secondary air adjustment scheme.

2. The method for secondary air control of a circulating fluidized bed boiler based on coal quality characteristics according to claim 1, characterized in that, A secondary air control scheme is generated based on the simulated furnace data, including: Extract the parameter value set corresponding to each detection parameter from the simulated furnace data; generate a stable feature vector based on the parameter value set corresponding to each detection parameter. The stable feature vector is input into the secondary wind control scheme generation model to obtain the corresponding secondary wind control scheme; the secondary wind control scheme generation model is obtained by training an artificial intelligence model; the secondary wind control scheme includes the total secondary wind volume, the fan speed and the damper opening.

3. The method for secondary air control of a circulating fluidized bed boiler based on coal quality characteristics according to claim 2, characterized in that, The process of generating a stable feature vector based on the parameter value group corresponding to each detection parameter includes: Extract the parameter values ​​collected at each time point from each parameter value group, and integrate the parameter values ​​corresponding to each detection parameter at the same time point into a feature vector according to a set order. The feature vectors are arranged in chronological order of their corresponding acquisition times. A predetermined number of feature vectors are obtained as the basic clustering vectors. The remaining feature vectors are divided into several perturbation vector groups, each perturbation vector group containing the same number of feature vectors. Cluster analysis is performed on each basic clustering vector to obtain several clusters, which are denoted as basic clusters. Feature vectors from perturbation vector groups are added sequentially to each basic clustering vector, and cluster analysis is performed to obtain several corresponding clusters. Clusters that do not have complete consistency among the basic clusters are denoted as perturbation clusters. Stable feature vectors are generated based on several basic clusters and perturbation clusters.

4. The method for secondary air control of a circulating fluidized bed boiler based on coal quality characteristics according to claim 3, characterized in that, The generation of stable feature vectors based on several basic clusters and slightly disturbed dynamic clusters includes: Obtain the center vectors of several basic clusters and denote them as basic center vectors; obtain the center vectors corresponding to the disturbed clusters, calculate the Euclidean distance between each center vector and the basic center vectors, and classify the disturbed clusters corresponding to the center vectors whose Euclidean distance is less than a set distance threshold into the cluster evolution group corresponding to the basic clusters; Obtain the number of clusters in each cluster evolution group, obtain the cluster evolution group with a number of clusters greater than a set evolution number threshold, and extract the center vectors corresponding to the disturbed clusters in the cluster evolution group, calculate the dispersion of each center vector; take the cluster evolution group corresponding to the minimum dispersion as the stable evolution group, and denote the mean vector of the center vectors corresponding to each disturbed cluster in the stable evolution group as the stable feature vector.

5. The method for secondary air control of a circulating fluidized bed boiler based on coal quality characteristics according to claim 2, characterized in that, The secondary wind control scheme generation model is obtained through training an artificial intelligence model, including: Acquire several historical furnace in-furnace data and the optimal secondary air control scheme corresponding to the historical furnace in-furnace data; generate stable feature vectors corresponding to the historical furnace in-furnace data based on the parameter value groups corresponding to several detection parameters in the historical furnace in-furnace data; integrate several stable feature vectors and their corresponding secondary air control schemes into several training data and test data. The artificial intelligence model is trained using training data and tested using testing data. The final result is a secondary wind control scheme generation model with stable feature vectors as input and secondary wind control schemes corresponding to the stable feature vectors as output.

6. The method for secondary air control of a circulating fluidized bed boiler based on coal quality characteristics according to claim 1, characterized in that, The secondary air adjustment scheme generated based on real-time furnace data includes: Extract the parameter value groups corresponding to each detection parameter in the real-time furnace data; extract the parameter values ​​at each time point in each parameter array; Representative detection parameters are obtained by selecting the parameter values ​​corresponding to several times in the parameter value group corresponding to each detection parameter. Obtain stable feature vectors; extract the parameter values ​​corresponding to the detection parameters in the stable feature vectors and mark them as reference values ​​for the corresponding detection parameters; A secondary wind adjustment plan is generated based on the current parameter values ​​and reference values ​​of the representative detection parameters.

7. The method for secondary air control of a circulating fluidized bed boiler based on coal quality characteristics according to claim 6, characterized in that, The method of selecting representative detection parameters based on parameter values ​​at several times from the parameter value groups corresponding to each detection parameter includes: Substitute the parameter values ​​at several times in the parameter value group corresponding to the detection parameter into the set stable change evaluation function to obtain the degree of change of the detection parameter. The detection parameter corresponding to the smallest degree of change is selected as the representative detection parameter.

8. The method for secondary air control of a circulating fluidized bed boiler based on coal quality characteristics according to claim 6, characterized in that, The process of generating a secondary wind adjustment scheme based on the current parameter values ​​and reference values ​​representing the detection parameters includes: Get the current value of the parameter representing the detection parameter. and reference values Substitute these values ​​into the set adjustment ratio generation function to obtain the adjustment ratio corresponding to each control parameter. The adjustment ratios corresponding to each control parameter are integrated into a secondary air adjustment scheme.

9. The method for secondary air control of a circulating fluidized bed boiler based on coal quality characteristics according to claim 1, characterized in that, Constructing a digital twin model of the fluidized bed boiler based on the aforementioned industrial data includes: The entity data and component association mapping data of the fluidized bed boiler are extracted from industrial data. A three-dimensional model of the fluidized bed boiler is constructed based on the entity data. The association mapping relationship of each component in the three-dimensional model of the fluidized bed boiler is established based on the component association mapping data, thus obtaining a digital twin model of the fluidized bed boiler.

10. A secondary air control system for a circulating fluidized bed boiler based on coal quality characteristics, characterized in that, include: The module consists of a data acquisition module, a data analysis module, and a control module; among which, The data acquisition module includes a first acquisition unit and a second acquisition unit; The first acquisition unit is used to acquire coal quality data of the coal to be added; The second acquisition unit is used to acquire industrial data of fluidized bed boilers and corresponding in-furnace data of fluidized bed boilers; The data analysis module includes a first analysis unit and a second analysis unit; The first analysis unit is used to construct a digital twin model of the fluidized bed boiler based on the industrial data, and input the coal quality data and furnace data into the digital twin model to perform simulation operation and obtain the corresponding simulated furnace data; The second analysis unit is used to generate a secondary air control scheme based on the simulated furnace data, and to acquire real-time furnace data and generate a secondary air adjustment scheme based on the real-time furnace data. The control module is used to regulate the fluidized bed boiler based on the secondary air control scheme and the secondary air adjustment scheme.

Images