A combustion air volume control system for a biomass boiler

Abstract

The application discloses a kind of combustion air volume control systems of biomass boiler, belong to biomass boiler technical field, including fuel parameter acquisition unit, for obtaining moisture content signal of biomass fuel, unit time supply amount signal and particle size characterization signal on the feeding path;Furnace and flue gas parameter acquisition unit, for obtaining furnace negative pressure signal, furnace temperature signal, exhaust gas oxygen content signal and exhaust gas carbon monoxide concentration signal;Wall heat load acquisition unit, for obtaining heat flux signal or wall temperature distribution signal of water-cooled wall, and generate slagging risk characterization quantity;By online obtaining and generating fuel reactivity index to moisture content, supply amount and particle size characterization parameter of biomass fuel, the distribution ratio of air supply and primary air secondary air can be pre-corrected with fuel quality fluctuation, so as to reduce the hysteresis caused by only relying on exhaust gas oxygen content or carbon monoxide single-point feedback and improve load tracking stability.

Description

Technical Field

[0001] This invention relates to the field of biomass boiler technology, and more specifically, to a combustion air volume control system for a biomass boiler. Background Technology

[0002] Biomass boilers are typically used as a type of energy-saving boiler and its auxiliary equipment. Their combustion process is highly sensitive to the air supply volume and the distribution ratio of primary and secondary air. In actual engineering operations, biomass fuel sources are complex and batches vary significantly. Fluctuations in moisture content and particle size distribution directly alter the volatile release rate, bed burnout rate, and local oxygen consumption intensity, causing the optimal excess air level and staged air distribution ratio under the same load to drift continuously. Meanwhile, after long-term operation, the fans and dampers in the air supply branch are prone to dust accumulation, leakage, mechanical lag, or decreased transmission efficiency, which causes the flow capacity of the execution link to drift. This makes it difficult for the fan speed command and damper opening command output by the controller to be stably mapped to the actual air volume, resulting in a deviation between the set air volume and the actual air volume. In existing technologies, common control methods often use the oxygen content of flue gas as the main feedback quantity, supplemented by carbon monoxide concentration or furnace negative pressure for correction, or use fixed empirical curves to adjust the ratio of primary and secondary air. However, such methods usually have problems such as feedback lag and failure to uniformly incorporate coupling factors. When fuel quality changes abruptly or the flow capacity of the actuator drifts, excessive air can easily lead to decreased efficiency, or local oxygen deficiency can lead to increased carbon monoxide and incomplete combustion. In some cases, when local jet or wall-mounted combustion is enhanced, it can even cause abnormal increase in the heat load of the water-cooled wall and increase the risk of slagging. Therefore, there is an urgent need for a biomass boiler combustion air volume control system that can adaptively optimize the air volume and the distribution ratio of each air outlet under the conditions of both fuel quality fluctuations and air supply execution link flow capacity drift. Summary of the Invention

[0003] To achieve the above objectives, the present invention provides the following technical solution: A combustion air volume control system for a biomass boiler includes: a fuel parameter acquisition unit, used to acquire biomass fuel moisture content signal, feed rate per unit time signal, and particle size characterization signal along the feeding path; The furnace and flue gas parameter acquisition unit is used to acquire furnace negative pressure signal, furnace temperature signal, flue gas oxygen content signal, and flue gas carbon monoxide concentration signal. The wall heat load acquisition unit is used to acquire the heat flux signal or wall temperature distribution signal of the water-cooled wall and generate a slagging risk characterization quantity. The fuel reactivity estimation unit is used to calculate the fuel reactivity index based on the moisture content signal, the feed rate per unit time signal, and the particle size characterization signal, and to update the fuel reactivity index online. The actuator health estimation unit is used to estimate the equivalent flow coefficient of the corresponding air supply branch and generate a health compensation factor online based on the fan current signal, fan speed signal, duct differential pressure signal and damper opening signal of each air supply branch. The combustion state prediction model unit is used to construct a prediction model that includes the bed oxygen consumption state, volatile matter release state and furnace effective temperature state, and inputs the fuel reactivity index and the equivalent flow coefficient as updatable parameters into the prediction model. The model predictive air distribution decision unit is used to solve the optimization problem based on the prediction model in the rolling time domain, so as to output the set value of the total primary air volume, the set value of the total secondary air volume, and the air volume distribution ratio of each air outlet. The constraints of the optimization problem include at least the range constraints of flue gas oxygen content, the upper limit constraints of carbon monoxide concentration, the range constraints of furnace negative pressure, and the constraint that the slagging risk characterization quantity does not exceed a preset threshold. The air distribution execution unit includes control interfaces for a primary air fan and a primary damper, as well as for a secondary air fan and a secondary damper. It is used to predict the output of the air distribution decision unit based on the model and generate fan speed control commands and damper opening control commands in combination with the health compensation factor, thereby realizing closed-loop control of the combustion air volume of the biomass boiler. Furthermore, the fuel parameter acquisition unit includes a near-infrared spectroscopy acquisition module, a microwave moisture content acquisition module, a weighing and metering module, and a particle image acquisition module; The near-infrared spectroscopy acquisition module is used to acquire the reflectance spectrum of biomass fuel within a preset wavelength range and extract spectral feature factors; the microwave moisture content acquisition module is used to acquire the fuel moisture content signal; the weighing and metering module is used to acquire the feed rate signal per unit time; and the particle image acquisition module is used to acquire fuel particle images and output particle size distribution characterization parameters. The fuel reactivity estimation unit is used to generate a fuel reactivity index based on the moisture content signal, the feed rate per unit time signal, the particle size distribution characterization parameters, and the spectral characteristic factors, and to recursively update the mapping weight of the fuel reactivity index based on the short-time response deviation between the flue gas oxygen content signal and the flue gas carbon monoxide concentration signal.

[0004] Furthermore, the actuator health estimation unit is configured to establish an air volume estimation model for each air supply branch; Based on the fan speed signal, fan current signal, duct differential pressure signal, duct temperature signal and damper opening signal of the air supply branch, the predicted air volume of the air supply branch is calculated. At the same time, the deviation between the measured air volume obtained by converting the duct differential pressure signal and the predicted air volume is used as the identification error, and the equivalent flow coefficient of the air supply branch is recursively updated to obtain the health compensation factor. The air distribution execution unit compensates and corrects the fan speed control command and damper opening control command according to the health compensation factor, so that the set value of the primary air total volume and the set value of the secondary air total volume still meet the output requirements of the model prediction air distribution decision unit under the conditions of damper lag, leakage or fan performance degradation.

[0005] Furthermore, the wall heat load acquisition unit is configured to divide the water-cooled wall into multiple monitoring zones along the furnace height direction and circumferential direction, and output the heat flux signal or wall temperature signal of each monitoring zone respectively. The system also includes a slagging risk quantification unit, which is used to generate a slagging risk characterization quantity based on the heat flux signal or wall temperature signal of each monitoring zone, wherein the slagging risk characterization quantity is obtained by weighting at least the zone peak heat load, the zone heat load spatial gradient and the duration exceeding the preset heat load threshold within the monitoring window. The model predictive air distribution decision unit is configured to use the slagging risk characterization quantity as a constraint quantity for rolling optimization, and when the slagging risk characterization quantity approaches the threshold, prioritize adjusting the secondary air volume distribution ratio and the local air outlet opening to reduce the heat flux or wall temperature peak of the high heat load zone attached to the wall, while maintaining the exhaust oxygen content range constraint and the carbon monoxide concentration upper limit constraint to meet the preset requirements.

[0006] Furthermore, the model prediction air distribution decision unit is configured to construct a rolling optimization problem containing an objective function and constraints. The objective function includes at least a flue gas oxygen content deviation term to characterize combustion efficiency, a flue gas carbon monoxide concentration penalty term to characterize incomplete combustion, and a fan power penalty term to characterize air supply energy consumption. The flue gas carbon monoxide concentration penalty term adopts a logarithmic or piecewise penalty structure to increase the penalty weight when the carbon monoxide concentration is close to a preset upper limit. The constraints of the rolling optimization problem include at least the upper and lower limits of the total primary air volume and the total secondary air volume, the feasible region constraint of the air volume distribution ratio of each tuyer, the furnace negative pressure range constraint, the range of flue gas oxygen content constraint, the upper limit constraint of flue gas carbon monoxide concentration constraint, and the threshold constraint of the slagging risk characterization quantity. The model-predicted air distribution decision unit is also configured to adjust the weights of each penalty term in the objective function online based on the rate of change of boiler load and the rate of change of the fuel reactivity index, so as to simultaneously meet the combustion efficiency target and the stable combustion target under conditions of sudden load changes or fuel quality fluctuations.

[0007] Furthermore, the furnace and flue gas parameter acquisition unit is further used to acquire furnace sound pressure signals and furnace pressure pulsation signals; The combustion state prediction model unit is further configured to extract the low-spectrum energy ratio, the main frequency drift, and the peak-to-peak value of the pressure pulsation based on the furnace acoustic pressure signal and the furnace pressure pulsation signal, and generate combustion instability characterization quantities. The model-predicted wind distribution decision unit is further configured to use the combustion instability characterization quantity as the trigger quantity and constraint update quantity for rolling optimization. When the combustion instability characterization quantity exceeds a preset threshold, at least one of the following linked updates is performed: Increase the weight of the carbon monoxide concentration penalty term in the objective function or tighten the upper limit constraint of the carbon monoxide concentration in the flue gas, tighten the furnace negative pressure range constraint or increase the weight of the furnace negative pressure deviation penalty, limit the secondary air volume distribution ratio within the preset vibration suppression feasible domain and impose an upper limit constraint on the rate of change of the opening of the secondary air local tuyeres. Based on the rolling optimization results after the linkage update, the set value of the total primary air volume, the set value of the total secondary air volume, and the air volume distribution ratio of each air outlet are output to achieve early suppression of combustion instability and simultaneously meet the constraints of the range of exhaust oxygen content and the threshold constraints of the slagging risk characterization quantity.

[0008] Furthermore, the model prediction air distribution decision unit is further configured to introduce air supply branch coordination consistency constraints and consistency penalty terms in the rolling optimization, wherein the primary air branch and each secondary air outlet branch are defined as nodes to be allocated and a branch coupling relationship is established, so as to ensure that the air volume allocation ratio corresponding to each node remains consistent with the target air distribution ratio vector determined by the fuel reactivity index, under the premise of satisfying the set value of the total primary air volume and the set value of the total secondary air volume. The consistency penalty term is used to apply a weighted penalty to the deviation in the air volume allocation ratio between the coupled branch nodes, and the health compensation factor output by the actuator health estimation unit is used as the basis for updating the weighting coefficient, so that the branch with deteriorated health status is automatically reduced in allocation weight in the rolling optimization and the branch with better health status bears the compensation air volume. This allows for coordinated convergence of air distribution across multiple air outlets under fluctuating fuel quality conditions, while simultaneously satisfying constraints on the range of flue gas oxygen content, the upper limit of carbon monoxide concentration, the range of furnace negative pressure, and the threshold constraints of slagging risk characterization.

[0009] In summary, the present invention has the following beneficial effects: By acquiring biomass fuel moisture content, feed rate and particle size characterization parameters online and generating fuel reactivity index, the air supply volume and primary and secondary air distribution ratio can be corrected in advance according to fuel quality fluctuations, thereby reducing the lag caused by relying solely on single-point feedback of flue gas oxygen content or carbon monoxide and improving load tracking stability. By online identification of the fan speed, fan current, duct differential pressure and damper opening of the air supply branch, the equivalent flow coefficient is obtained and a health compensation factor is formed. This allows the flow capacity drift caused by actuator dust accumulation, leakage and hysteresis to be compensated in real time, thereby reducing the deviation between the set air volume and the actual air volume and improving the consistency of long-term operation. By incorporating the slagging risk characterization quantity constructed by the heat load or wall temperature of the water-cooled wall into the rolling optimization constraint, the air distribution adjustment can suppress the continuous accumulation of high heat load areas attached to the wall while meeting the range of oxygen content in the flue gas and the upper limit of carbon monoxide, thereby reducing the risk of slagging and local overheating and improving operational safety. By using the combustion instability characterization quantity formed by furnace sound pressure and pressure pulsation to trigger the adaptive update of weights and constraints within the same rolling optimizer, combustion pulsation and unstable conditions can be suppressed in advance, thereby balancing burnout efficiency and stable combustion objectives under conditions of fuel and load mutation. By introducing a consistency constraint for multi-ventilation distribution and updating the distribution weights with a health compensation factor, the air distribution between vents can converge collaboratively and be automatically redistributed when the performance of local branches deteriorates, thereby improving the robustness of multi-ventilation distribution and reducing manual adjustment costs. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of the present invention 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 the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic diagram of the overall structure of the biomass boiler combustion air volume control system of the present invention; Figure 2 This is a flowchart of the fuel parameter acquisition and online fuel reactivity index update process of the present invention; Figure 3 This is a schematic diagram illustrating the wall heat load zoning, slagging risk quantification, and constraint linkage of the present invention. Figure 4 This is a schematic diagram of the rolling optimized air distribution closed loop and the instability triggering and consistency constraints of the present invention. Detailed Implementation

[0012] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0013] Example: The following is in conjunction with the appendix Figure 1-4 The present invention will be described in further detail below.

[0014] Please see Figure 1-4 This invention provides a technical solution: a combustion air volume control system for a biomass boiler, such as... Figure 1-4 As shown, it includes: a fuel parameter acquisition unit, used to acquire the moisture content signal, feed rate per unit time signal, and particle size characterization signal of biomass fuel on the feeding path; The furnace and flue gas parameter acquisition unit is used to acquire furnace negative pressure signal, furnace temperature signal, flue gas oxygen content signal, and flue gas carbon monoxide concentration signal. The wall heat load acquisition unit is used to acquire the heat flux signal or wall temperature distribution signal of the water-cooled wall and generate a slagging risk characterization quantity. The fuel reactivity estimation unit is used to calculate the fuel reactivity index based on the moisture content signal, the feed rate per unit time signal, and the particle size characterization signal, and to update the fuel reactivity index online. The actuator health estimation unit is used to estimate the equivalent flow coefficient of the corresponding air supply branch and generate a health compensation factor online based on the fan current signal, fan speed signal, duct differential pressure signal and damper opening signal of each air supply branch. The combustion state prediction model unit is used to construct a prediction model that includes the bed oxygen consumption state, volatile matter release state, and furnace effective temperature state, and inputs the fuel reactivity index and equivalent flow coefficient as updatable parameters into the prediction model. The model-predictive air distribution decision unit is used to solve optimization problems based on the prediction model in the rolling time domain, so as to output the set value of the total primary air volume, the set value of the total secondary air volume, and the air volume distribution ratio of each air outlet. The constraints of the optimization problem include at least the range constraints of flue gas oxygen content, the upper limit constraints of carbon monoxide concentration, the range constraints of furnace negative pressure, and the constraint that the slagging risk characterization quantity does not exceed the preset threshold. The air distribution execution unit includes control interfaces for the primary air fan and primary damper, as well as for the secondary air fan and secondary damper. It is used to predict the output of the air distribution decision unit based on the model and generate fan speed control commands and damper opening control commands in combination with health compensation factors, thereby realizing closed-loop control of the combustion air volume of the biomass boiler. In this embodiment: the biomass boiler is a chain grate boiler or a circulating fluidized bed boiler. The boiler's primary air fan corresponds to the primary air duct and supplies air to the area below the grate or the bed through the primary air damper. The secondary air fan corresponds to the secondary air duct and supplies air to the upper part of the furnace in stages through several secondary air outlets. The induced draft fan maintains the negative pressure in the furnace through the induced draft regulating valve. The combustion air volume control system uses an industrial control computer or a programmable logic controller with floating-point computing capabilities as the computing and control carrier. It is connected to each sensor and actuator through a fieldbus or Ethernet. The sampling period is preferably 0.5s to 2s, and the control refresh period is preferably 1s. In this embodiment, the fuel parameter acquisition unit is arranged at the outlet of the feed belt or screw feeder, including a microwave moisture content sensor for outputting the fuel moisture content signal W, a weighing and metering device for outputting the feed rate signal ṁf per unit time, a particle image acquisition device for outputting particle size distribution characterization parameters d50 and d90, and optionally a near-infrared spectral acquisition device for outputting the spectral feature factor vector v; wherein the particle image acquisition device preferably adopts an industrial camera with a dust-resistant air curtain and a ring-shaped supplementary lighting structure, and obtains d50 and d90 by threshold segmentation and connected component statistics of continuous frame images; the near-infrared spectral acquisition device preferably covers the 900nm to 1700nm band, and extracts the principal component score related to the cellulose lignin content as v from the spectral lines. Wherein, d50 represents the equivalent particle size when the cumulative particle size distribution reaches 50%; d90 represents the equivalent particle size when the cumulative particle size distribution reaches 90%; and v represents the principal component score obtained from the near-infrared spectrum by principal component analysis and is related to the cellulose lignin content. The furnace and flue gas parameter acquisition units are arranged at key locations in the furnace and flue, including: a furnace negative pressure sensor outputting Pf, where Pf represents the furnace negative pressure sampling value at the pressure tapping point at the top of the furnace, in Pa, measured relative to atmospheric pressure, with a negative Pf value indicating negative pressure; a furnace temperature sensor outputting Tf, where Tf represents the furnace temperature sampling value at a representative point, in °C; a flue gas oxygen content probe outputting O2, where O2 represents the flue gas oxygen volume fraction sampling value, in %; and a flue gas carbon monoxide sensor outputting CO, where CO represents the flue gas carbon monoxide concentration sampling value, in ppm or mg per cubic meter. O2 and CO are preferably located in the flue gas section after secondary air mixing and where the flue gas is uniform to reduce local deviations. The furnace temperature is preferably sampled using a multi-point thermocouple array with median filtering for abnormal points. The furnace negative pressure is preferably sampled at the top of the furnace and equipped with a pressure tapping and condensation isolation structure. The wall heat load acquisition unit is used to acquire the heat flux signal or wall temperature distribution signal of the water-cooled wall and generate a slagging risk characterization quantity. In this embodiment, the water-cooled wall is divided into at least 8 monitoring zones along the furnace height and circumferential direction. Each zone is equipped with a heat flux meter or infrared thermometer array, and outputs the zone heat flux qk or zone wall temperature Twk. The peak heat load qkmax, spatial gradient gk, and over-threshold duration τk within the monitoring window Δt are combined to generate the slagging risk characterization quantity Sz, where qkmax is the maximum qk within Δt, gk is the normalized amount of the heat load difference between adjacent zones, τk is the cumulative time for qk to be higher than the threshold qth, and Sz can be the weighted sum of each zone, with the weight related to the zone position to highlight the high slagging sensitive area of ​​the furnace. Δt is preferably 20s to 120s, and qth is preferably set in combination with the fuel ash melting point and the boiler design wall temperature window. The fuel reactivity estimation unit is used to generate and update the fuel reactivity index Rf online based on fuel moisture content W, feed mass flow rate mf, particle size characteristics d50 and d90, and near-infrared spectral principal component score v. Here, W represents fuel moisture content or moisture mass fraction (%); mf represents feed mass flow rate per unit time (kg / s); d50 represents the equivalent particle size when the cumulative particle size distribution reaches 50%, and d90 represents the equivalent particle size when the cumulative particle size distribution reaches 90% (micrometers); v represents the principal component score obtained from the 900nm to 1700nm near-infrared spectrum through principal component analysis and is related to cellulose and lignin content; Rf is a dimensionless reactivity index. In this embodiment, a normalized feature vector x is first constructed, letting: ; in , , , , For calibration reference values; the fuel reactivity index is calculated using a weighted mapping as follows: ; Where Rf max With Rf min These are the upper and lower bounds of the reactivity index. to The coefficients are obtained through historical operating data; the online update uses a rolling window correction factor. To perform a smooth update, let: ; Where Tf is the measured furnace temperature, and CO is the measured carbon monoxide in the flue gas. tim T is the preset upper limit. ref For reference temperature, To correct the strength coefficient, It is a smoothing coefficient with a value ranging from 0 to 1; In this embodiment, a normalized feature vector x is first constructed, which includes a moisture content term x1, a feed rate term x2, a particle size term x3, and a spectral term x4. Here, x1 represents the normalized result of the fuel moisture content W relative to the reference moisture content Wref, x2 represents the feed mass flow rate, which is the normalized result relative to the reference flow rate, x3 represents the normalized result of the particle size feature obtained by weighting d50 and d90, and x4 represents the normalized result of the near-infrared principal component score relative to the reference score. The initial value of the fuel reactivity index is then calculated. The linear mapping between the parameter vector and x yields Rf0, which is equal to θ0 plus θ1 x1 plus θ2 x2 plus θ3 x3 plus θ4 x4. Here, θ is equal to the parameter vector composed of θ0 to θ4, where θ0 is a constant term, and θ1 to θ4 are the contribution coefficients of moisture content, feed rate, particle size, and spectral characteristics to fuel reactivity, respectively. The initial value of θ is obtained through offline calibration. During offline calibration, W, mf, d50, d90, v, and the corresponding steady-state combustion characterization quantities are collected under no less than twenty sets of fuel samples and typical load conditions. The least squares criterion is used to identify θ to minimize the prediction error of the combustion characterization quantities. To enable the fuel reactivity index Rf to have online adaptive correction capability, a short-time response deviation is introduced as a correction basis during boiler operation, with the refresh cycle controlled between 20s and 120s. At the end of each refresh cycle, the flue gas oxygen content and flue gas carbon monoxide concentration are collected, and the predicted changes in oxygen content and carbon monoxide given by the combustion state prediction model within the same refresh cycle are read simultaneously. The measured changes are obtained by the difference between the end points of two adjacent refresh cycles, that is, the change of the end point value of the current cycle relative to the end point value of the previous cycle is taken as the measured change of that cycle. The short-time response deviation is obtained using a dimensionless comprehensive deviation caliber. During calculation, the oxygen content channel and the carbon monoxide channel are processed separately. First, the absolute deviation between the measured change and the predicted change of each channel is calculated, and then normalized using a reference amplitude. The reference amplitude can be the typical change amplitude under rated operating conditions or the upper quartile of the change amplitude obtained from the statistics of the most recent hour to avoid the influence of occasional spikes. Finally, the two normalized deviations are weighted and summed to obtain the short-time response deviation. The sum of the oxygen weight and carbon monoxide weight is 1. The weights can be set to 0.30 to 0.60 for oxygen and 0.40 to 0.70 for carbon monoxide, depending on the control target, to enhance sensitivity to incomplete combustion risks. A larger short-time response deviation indicates a greater fitting error of the model to the short-time dynamic response, further characterizing a higher probability of fuel reactivity drift or health deterioration in the air distribution process. Based on this short-time response deviation, Rf is corrected online. During correction, the current Rf is used as the base value, and a reduction correction is performed by multiplying the correction step size by the short-time response deviation. A limiting process is introduced to ensure that the corrected Rf always falls between the preset minimum and maximum Rf values. The correction step size is a positive number and can be set to 0.01 to 0.20 to balance response speed and stability. The limited correction result is used as an observation of Rf and is used to update the mapping parameters between Rf and characteristic quantities online. The mapping adopts a recursive least squares online update mechanism and introduces a forgetting factor to enhance the tracking ability of operating condition changes. The forgetting factor is set to 0.90 to 0.98; The recursive update process is implemented according to the following executable steps: First, construct a regression vector, whose components include a constant term and four characteristic quantities. The four characteristic quantities are preferably a combination of fuel moisture content, instantaneous feed rate, furnace negative pressure and water-cooled wall heat load or flue gas temperature to cover disturbances on the fuel side and thermal side; then, use the parameter vector of the previous period to perform multiplication and addition operations on the current regression vector to obtain the estimated value of the current Rf observation; use the difference between the current Rf observation and the estimated value as the residual; then calculate the gain vector based on the covariance matrix of the previous period and the current regression vector. The gain vector is obtained by first using the covariance matrix... The intermediate vector is obtained by performing matrix-vector multiplication with the regression vector. Then, a scalar division is performed using the inner product of the forgetting factor and the regression vector (weighted by covariance) as the normalized denominator. Subsequently, the parameter vector is additively updated using the gain vector and residuals. Finally, the covariance matrix is ​​recursively updated, and a forgetting factor is introduced for amplification to absorb new data. The initial values ​​of the parameter vector and covariance matrix can be obtained offline through least squares identification within 20 to 60 minutes after startup stabilization or given based on engineering experience, and are continuously updated online during operation to achieve adaptive correction of Rf and long-term drift suppression. The actuator health estimation unit is used to estimate the equivalent flow coefficient Cv of the supply air branch online and generate a health compensation factor; in this embodiment, an air volume estimation relationship Qhat is established between the primary air branch and each secondary air branch, and Qhat is calculated by the following formula: Qhat = Cv × u × (√(ΔP ÷ T)); where u is the normalized control quantity of damper opening or fan speed, ranging from 0 to 1; ΔP is the measured differential pressure of the differential pressure sensor in this branch duct, in Pa; T is the measured or equivalent temperature of the temperature sensor in this branch duct, in K; Cv is the equivalent flow coefficient, used to characterize the change in flow capacity caused by damper leakage, jamming, and dust accumulation. To obtain the measured air volume Qmeas, it is preferable to set up a standard pressure tapping element in each branch and establish a differential pressure air volume calibration curve during the commissioning period. The calibration curve is obtained by: collecting the differential pressure ΔPi and temperature Ti of the branch at multiple steady-state operating points, and obtaining the reference air volume Qstd,i of the branch using a reference air volume measurement method, which can be a Pitot tube grid measurement method, a standard orifice plate flow meter, or a Venturi flow meter; then, using least squares fitting to obtain the calibration coefficient Kcal of the branch, so that the square error of the following formula at each operating point is minimized, Qstd,i is approximately equal to Kcal multiplied by the square root of ΔPi divided by Ti; after calibration, the measured air volume Qmeas is calculated online, and Qmeas is defined as Kcal multiplied by the square root of ΔP divided by T; thus, Qmeas can be obtained and calculated. The identification error is defined as εQ, which is defined as Qmeas minus Qhat. Recursive least squares with a forgetting factor is used to update Cv online, ensuring εQ converges in the sliding window sense. The input to recursive least squares is the regressor ψk, defined as uk multiplied by the square root of ΔPk divided by Tk. The output of recursive least squares is the updated Cvk plus 1. The error is defined as ekCv equal to Qmeas, k minus ψk multiplied by Cvk. The forgetting factor ranges from 0.90 to 0.98. When the deviation of Cvk from its initial calibration value Cvcal exceeds the preset proportional threshold, a health compensation factor H is output. H is the ratio of Cvcal to Cvk or its limiting form, and H is used for subsequent command compensation. Command compensation can be achieved by multiplying the control quantity ucmd by H and then limiting it to obtain ucomp, so that under the same target air volume requirement, the flow capacity reduction caused by damper leakage or dust accumulation can be automatically compensated. The combustion state prediction model unit constructs a prediction model including the bed oxygen consumption state, volatile matter release state, and furnace effective temperature state, and inputs Rf and Cv as updatable parameters. In this embodiment, the state vector is s, which includes oxygen consumption intensity So, volatile matter release intensity Sv, and effective temperature Se; the control vector is u, which includes the total primary air volume Qa, the total secondary air volume Qb, and the secondary air distribution ratio vector β; the branch equivalent flow coefficient is used to map the control quantity to the actual air volume, and the actual air volume is determined by the equivalent flow coefficient, the control quantity, the branch differential pressure, and the branch temperature; the combustion state is updated according to the discrete state recursive model. The inputs for the state recursion include actual air volume, feed mass flow rate, and fuel reactivity index; flue gas oxygen content, flue gas carbon monoxide, furnace negative pressure, and furnace temperature are calculated from the state variables according to the output mapping model; the parameters of both the state recursion model and the output mapping model are calibrated through step tests during boiler commissioning and are recursively updated based on prediction errors during operation; Rf is used to adjust the coupling coefficient between volatile matter release and oxygen consumption, and Cv is used to constrain the mapping from control variables to actual air volume; the initial values ​​of the model parameters can be calibrated through step tests during boiler commissioning, and adaptively updated through updates of Rf and Cv during the online phase; The model predicts the air distribution decision unit and solves the optimization problem in the rolling time domain, outputting the setpoints Qa, Qb, and β. In this embodiment, the prediction time domain Np is preferably 20 to 60 steps, corresponding to 20 to 120 seconds, and the control time domain Nc is preferably 5 to 20 steps. The objective function J includes at least a flue gas oxygen content deviation term, a flue gas carbon monoxide penalty term, and a supply air energy consumption penalty term, and may include an air volume change rate penalty term to reduce actuator wear. The air distribution execution unit includes control interfaces for the primary air fan and primary damper, and for the secondary air fan and secondary damper, used to generate control commands based on the rolling optimization output and combined with a health compensation factor. In this embodiment, the control commands include at least a primary damper opening setpoint, a primary air fan speed setpoint, a secondary damper opening setpoint vector, and a secondary air fan speed setpoint. The target primary air volume and target secondary air volume obtained from the rolling optimization are used as inputs to the inverse model, while the measured differential pressure and temperature values ​​of the corresponding branches are read. The inverse model converts the target air volume into a normalized control quantity based on the branch differential pressure, branch temperature, and the equivalent flow coefficient calibrated during the commissioning phase. The actuator health estimation unit identifies the current equivalent flow coefficient online and generates... A health compensation factor is generated, which is equal to the ratio of the initially calibrated equivalent flow coefficient to the currently identified equivalent flow coefficient. This health compensation factor is multiplied by the normalized control quantity output by the inverse model and then limited to obtain the compensated control quantity. When the rolling optimization simultaneously outputs the secondary air distribution ratio vector, the total secondary air control quantity is distributed to each secondary damper opening setpoint according to the distribution ratio. The primary damper opening setpoint and the secondary damper opening setpoint are obtained by linear mapping according to the normalized control quantity, and the fan speed setpoint is obtained by linear mapping according to the normalized control quantity between the minimum speed and the maximum speed. To avoid actuator oscillation, upper limits are set for the damper opening change rate and the fan speed change rate to ensure that the opening change does not exceed 2% per unit time. The speed variation shall not exceed 1% to 3% of the rated speed. After execution, the furnace and flue gas parameter acquisition unit and the wall heat load acquisition unit shall collect the exhaust oxygen content, carbon monoxide concentration, furnace negative pressure, furnace temperature and slagging characterization, and use them as feedback inputs to the combustion state prediction model and rolling optimizer to update the fuel reactivity index and equivalent flow coefficient, thereby forming a closed-loop control from fuel identification to prediction optimization to executable compensation. Through the above embodiments, when the moisture content of biomass fuel increases or the particle size distribution becomes coarser, leading to changes in volatile matter release and burnout characteristics, the fuel reactivity index Rf is updated online and entered into the prediction model. Rolling optimization automatically increases the secondary air ratio or adjusts the distribution ratio of each air outlet to reduce CO and maintain O2 within the target range. When damper leakage or fan ash accumulation causes the equivalent flow coefficient Cv to decrease, the health compensation factor H corrects the command to maintain the actual air volume. When the local heat load of the water-cooled wall increases, causing Sz to approach the threshold, the optimizer adjusts β to reduce the heat load in the wall-adhesive area while satisfying O2 and CO constraints, thereby taking into account both burnout efficiency and slagging risk control. The rolling optimizer uses exhaust oxygen content and carbon monoxide concentration as optimization targets. By dynamically adjusting the total secondary air volume and the air volume distribution ratio of each secondary air outlet, it achieves the control objective of simultaneously reducing carbon monoxide emissions and maintaining oxygen content within the set target range. When the system detects changes in biomass fuel, increased slagging load, or drift in combustion reactivity, leading to an increase in carbon monoxide concentration or a deviation from the target oxygen level, the optimizer prioritizes increasing the total secondary air volume to enhance oxygen supply capacity and promote complete combustion. When the total secondary air volume has reached the set upper limit, the optimizer further redistributes the air volume among multiple air outlets, prioritizing the increase in the air supply ratio of outlets closer to the main combustion zone or anoxic areas to enhance local mixing and combustion efficiency. The optimizer simulates different air volume and distribution combinations in each rolling prediction cycle, predicts their impact on exhaust indicators, and automatically selects the optimal air volume combination that simultaneously reduces carbon monoxide and keeps oxygen content close to the target range as the control command, forming a dynamic closed-loop regulation mechanism.

[0015] like Figure 1-4 As shown, the fuel parameter acquisition unit includes a near-infrared spectroscopy acquisition module, a microwave moisture content acquisition module, a weighing and metering module, and a particle image acquisition module. The near-infrared spectroscopy acquisition module is used to acquire the reflectance spectrum of biomass fuel within a preset wavelength range and extract spectral feature factors; the microwave moisture content acquisition module is used to acquire the fuel moisture content signal; the weighing and metering module is used to acquire the feed rate signal per unit time; and the particle image acquisition module is used to acquire fuel particle images and output particle size distribution characterization parameters. The fuel reactivity estimation unit is used to generate a fuel reactivity index based on the moisture content signal, the feed rate per unit time signal, the particle size distribution characterization parameters, and the spectral characteristic factors. The mapping weight of the fuel reactivity index is recursively updated based on the short-time response deviation between the flue gas oxygen content signal and the flue gas carbon monoxide concentration signal.

[0016] In this embodiment: the fuel parameter acquisition unit is installed in a closed detection chamber at the end of the feeding belt or above the screw feed outlet. The detection chamber is equipped with an air curtain or low-pressure purging to reduce dust adhesion. The microwave moisture content acquisition module performs online moisture content measurement on the fuel passing through the detection chamber and outputs a moisture content signal W. Preferably, the correspondence between W and microwave phase difference or attenuation is established by calibrating the fuel in the same batch offline drying method and storing the calibration curve in the controller. The weighing and metering module uses a belt scale or hopper weighing plus time difference method to output a feed rate signal ṁf per unit time. The sampling period is preferably 0.5s to 2s and the instantaneous peaks are filtered by median. The particle image acquisition module uses an industrial camera and a supplementary lighting structure to acquire particle surface images. It obtains the equivalent particle size through threshold segmentation and connected component statistics, and calculates the particle size distribution characterization parameters d50 and d90. It also sets a minimum effective particle area threshold for the image calculation results to suppress powder noise. The near-infrared spectroscopy acquisition module can be installed in the same detection cavity. It preferably acquires the reflectance spectrum in the 900nm to 1700nm band and extracts the spectral feature factor vector v. The spectral feature factor vector v is preferably the spectral principal component score or a combination of reflectance of several characteristic wavelengths, thereby characterizing the changes in volatile matter release characteristics caused by the difference in the ratio of cellulose to lignin. In this embodiment, the fuel reactivity estimation unit constructs a fuel feature vector x using W, ṁf, d50, d90, and v, and generates a fuel reactivity index Rf. The feature vector x includes at least a moisture content feature x1, a feed rate feature x2, a particle size feature x3, and a spectral feature x4, wherein the particle size feature x3 can be obtained by a weighted combination of d50 and d90. The fuel reactivity index Rf can be calculated as Rf equal to the inner product of θ and x, where θ is the mapping weight vector. To achieve online recursive updates, the fuel reactivity estimation unit calculates the short-time response deviation e between the exhaust oxygen content signal O2 and the exhaust carbon monoxide concentration signal CO in each control refresh cycle. The short-time response deviation e is preferably calculated by the change in O2 and CO within a short window and the predicted change. The difference between quantities constitutes the short window, which is preferably 10s to 30s and can use the moving average to suppress random fluctuations. The fuel reactivity estimation unit uses e as the correction basis and adopts a recursive least squares or recursive update method with a forgetting factor to update θ. The forgetting factor is preferably 0.90 to 0.98 and upper and lower limits are set for θ and Rf to avoid divergence caused by abnormal fuel or sensor noise. Thus, when the moisture content of biomass fuel increases or the particle size becomes coarser, the short-time response deviation of O2 and CO will trigger the adaptive correction of θ, so that Rf can be updated online with changes in fuel quality and provide real-time available fuel reactivity input for model prediction and wind distribution decision, thereby achieving pre-compensation for fuel fluctuations without increasing parallel control loops.

[0017] like Figure 1-4 As shown, the actuator health estimation unit is configured to establish an airflow estimation model for each air supply branch; Based on the fan speed signal, fan current signal, duct differential pressure signal, duct temperature signal and damper opening signal of the air supply branch, the predicted air volume of the air supply branch is calculated. At the same time, the deviation between the measured air volume obtained by converting the duct differential pressure signal and the predicted air volume is used as the identification error, and the equivalent flow coefficient of the air supply branch is recursively updated to obtain the health compensation factor. The air distribution execution unit compensates and corrects the fan speed control command and damper opening control command based on the health compensation factor, so that the set values ​​of the primary air total volume and the secondary air total volume still meet the output requirements of the model prediction air distribution decision unit under the conditions of damper sluggishness, leakage or fan performance degradation. In this embodiment, the actuator health estimation unit establishes isomorphic online identification and compensation links for the primary air branch and each secondary air branch. Each air branch collects fan speed signals, damper position signals, duct differential pressure signals, and duct temperature signals. The controller calculates the predicted air volume Qmeas for the branch based on the sensor signals. Qmeas can be mapped to air volume by the linear coefficient calibrated during the commissioning period, or it can be calculated based on the physical model of pressure difference and temperature. Both methods have the same effect in this embodiment. The controller also calculates the predicted air volume Qhat based on the current equivalent flow coefficient Cv, the current damper or fan control quantity, and the current differential pressure signal. The actuator health estimation unit uses Qmeas minus Qhat as the identification error and recursively updates the Cv parameter to achieve adaptive estimation of duct flow capacity. In this embodiment, when Cv deviates from its initial calibration value Cv0, the actuator health estimation unit generates a health compensation factor H, which is preferably the ratio of Cv0 to the current Cv and is subjected to amplitude limiting. When the air distribution execution unit converts the target air volume Qset output by the model prediction air distribution decision unit into the fan speed and damper opening, it incorporates H into the target air volume or into the inverse mapping function to achieve command compensation, so that the actual air volume can still return to near Qset under actuator degradation conditions. To ensure stability, the rate of change of damper opening and fan speed after compensation are set to not exceed the preset upper limit. If the absolute value of the identification error e is still greater than the preset threshold within multiple consecutive refresh cycles, the health status mark of the branch is triggered, so that the model prediction air distribution decision unit automatically tends to be conservative in the allocation ratio of the branch in subsequent rolling optimization.

[0018] like Figure 1-4 As shown, the wall heat load acquisition unit is configured to divide the water-cooled wall into multiple monitoring zones along the furnace height and circumferential directions and output the heat flux signal or wall temperature signal of each monitoring zone respectively. The system also includes a slagging risk quantification unit, which is used to generate a slagging risk characterization quantity based on the heat flux signal or wall temperature signal of each monitoring zone. The slagging risk characterization quantity is obtained by weighting the zone peak heat load, zone heat load spatial gradient and the duration of exceeding the preset heat load threshold within the monitoring window. The model predicts the air distribution decision unit and is configured to use the slagging risk characterization quantity as the constraint quantity for rolling optimization. When the slagging risk characterization quantity approaches the threshold, the secondary air volume distribution ratio and the local air outlet opening are adjusted first to reduce the heat flux or wall temperature peak of the high heat load zone attached to the wall, while keeping the exhaust oxygen content range constraint and the carbon monoxide concentration upper limit constraint in line with the preset requirements. In this embodiment, the wall heat load acquisition unit divides the water-cooled wall into zones along the height and circumferential directions of the furnace. The preferred zoning method is to divide the height into four zones: upper, middle, lower, and throat-adjacent area, and to divide the circumferential direction into four sectors: the fire-facing side, the fire-repellent side, and the left and right sidewalls, thus forming at least sixteen monitoring zones. Each monitoring zone is equipped with a heat flux measurement point or a wall temperature measurement point and outputs the corresponding zone's heat flux signal qk or wall temperature signal Twk. The heat flux measurement point can be a water-cooled wall-attached heat flux meter, and its zero point and sensitivity are calibrated under boiler cold and steady-state conditions. The wall temperature measurement point can be a high-temperature resistant infrared thermometer that collects the radiation temperature of the zone and converts it into a wall temperature signal Twk after emissivity correction. The sampling period for each zone is preferably 0.5s to 2s, and qk or Twk is filtered using a moving average to suppress smoke and dust obstruction and random disturbances. In this embodiment, the slagging risk quantification unit constructs a slagging risk characterization quantity Sz using the peak heat load component, spatial gradient component, and over-threshold duration component within the monitoring window Δt. For each zone k, the peak heat load qkmax and over-threshold duration τk are calculated. The threshold qth or Twth is preferably determined during the commissioning phase based on fuel ash melting point test results and the boiler's allowable wall temperature window. The spatial gradient component is preferably obtained by normalizing the average heat load difference between the zone and adjacent zones. Sz is preferably generated by weighted summation of zones, with the weight wk related to the zone location to increase the weight of the throat-adjacent zone and the fire-facing zone. To avoid false triggering of instantaneous spikes, Sz can be further... The first step involves exponential smoothing to obtain Szf, which is then used as a constraint input to the model's predictive air distribution decision unit. In the rolling optimization, the model's predictive air distribution decision unit uses the constraint that Szf does not exceed the threshold Szth. When Szf approaches Szth, it prioritizes applying an upper limit constraint to the allocation ratio of local air outlets corresponding to high-risk zones in the secondary air distribution ratio vector and restricts the rate of change of the opening of these local air outlets. At the same time, it compensates for the allocation of air outlets corresponding to non-high-risk zones to maintain the range constraints of flue gas oxygen content and the upper limit constraints of carbon monoxide concentration. This achieves closed-loop control of slagging risk constraints based on water-cooled wall heat load zone feedback, enabling zone overheating to be suppressed in advance by the air distribution strategy before slagging forms.

[0019] like Figure 1-4 As shown, the model predicts the air distribution decision unit and is configured to construct a rolling optimization problem containing an objective function and constraints. The objective function includes at least a flue gas oxygen content deviation term to characterize combustion efficiency, a flue gas carbon monoxide concentration penalty term to characterize incomplete combustion, and a fan power penalty term to characterize air supply energy consumption. The flue gas carbon monoxide concentration penalty term adopts a logarithmic or piecewise penalty structure to increase the penalty weight when the carbon monoxide concentration is close to the preset upper limit. The constraints of the rolling optimization problem include at least the upper and lower limits of the total primary air volume and the total secondary air volume, the feasible region constraint of the air volume distribution ratio of each tuyer, the range of furnace negative pressure, the range of flue gas oxygen content, the upper limit constraint of flue gas carbon monoxide concentration, and the threshold constraint of slagging risk characterization quantity. The model-predicted air distribution decision unit is also configured to adjust the weights of each penalty term in the objective function online based on the rate of change of boiler load and the rate of change of fuel reactivity index, so as to simultaneously meet the combustion efficiency target and the stable combustion target under the conditions of sudden load changes or fuel quality fluctuations. In this embodiment, the model prediction air distribution decision unit uses the predicted quantity output by the combustion state prediction model unit as the optimization input, solves the rolling optimization problem in each control refresh cycle and outputs the primary air total air volume setpoint Qa, the secondary air total air volume setpoint Qb and the air volume distribution ratio vector β of each air outlet. The prediction time domain Np is preferably 20 to 60 steps and corresponds to the control refresh cycle of 20 to 120 seconds. The control time domain Nc is preferably 5 to 20 steps to reduce the solution burden. The objective function J consists of at least a flue gas oxygen content deviation term J1, a flue gas carbon monoxide penalty term J2, and an air supply energy consumption penalty term J3. An air volume change rate penalty term J4 can be added to suppress frequent actuator actions. J1 is the weighted sum of squares of the deviations between the predicted flue gas oxygen content O2pred and the target oxygen content O2star. O2star is preferably a settable value in the range of 4% to 8%. J3 is preferably based on the estimated power of the primary fan and the secondary fan as the penalty quantity, and the power estimation relationship is established in the controller through the fan speed and current. The carbon monoxide penalty term J2 in the flue gas uses a logarithmic or segmented penalty structure to improve the critical region suppression capability. The logarithmic penalty can be the weighted sum of logarithm 1 plus COpred divided by COref, where COref is the reference concentration and is preferably 50ppm to 200ppm. The segmented penalty can be set to use linear penalty when COpred does not exceed the first threshold and quadratic penalty with increased weight when COpred exceeds the first threshold or approaches the upper limit, thereby significantly improving the optimizer's priority for CO suppression when the risk of incomplete combustion increases. In this embodiment, the constraints of the rolling optimization problem include at least the upper and lower limits of the total primary air volume Qa and the total secondary air volume Qb, the feasible region constraint of the air volume allocation ratio β of each tuyer, the range constraint of the furnace negative pressure Pf, the range constraint of the flue gas oxygen content O2, the upper limit constraint of the flue gas carbon monoxide concentration CO, and the threshold constraint of the slagging risk characterization quantity Sz; wherein the O2 range is preferably 4% to 10%, the upper limit of CO is preferably 200ppm to 800ppm, the Pf range is preferably -80Pa to -10Pa, and the Sz threshold Szth is determined during the commissioning stage based on the zone heat load threshold and the duration threshold; the feasible region constraint of β preferably includes the minimum and maximum allocation ratios of each tuyer and includes the upper limit of the difference between the allocation ratios of adjacent tuyeres to avoid excessively strong local jets; to ensure the feasibility of execution, the rate of change constraint is applied to Qa, Qb, and β to limit the adjustment range within each refresh cycle; In this embodiment, the model predicts the air distribution decision unit to adjust the weights in the objective function online based on the boiler load change rate and the change rate of the fuel reactivity index Rf. The boiler load change rate can be obtained from the change rate of steam flow or main steam pressure, and the change rate of the fuel reactivity index is recursively updated. When the load change rate or the Rf change rate exceeds the threshold, the weight of the flue gas carbon monoxide penalty term J2 is increased, and the furnace negative pressure deviation penalty is increased or the Pf range constraint is tightened. At the same time, a more conservative margin factor is applied to the slagging risk constraint Sz to reduce the probability of wall-adhesive overheating. When the load change rate and the Rf change rate are in a stable range, the weight of J2 is reduced and the weight of the air supply energy consumption penalty term J3 is increased to pursue better energy-saving effects. Through the above online weight adjustment method, the rolling optimization prioritizes complete and stable combustion during fuel or load mutations, and prioritizes reducing excess air and air supply energy consumption during steady-state operation.

[0020] like Figure 1-4 As shown, the furnace and flue gas parameter acquisition unit is further used to acquire furnace sound pressure signals and furnace pressure pulsation signals; The combustion state prediction model unit is further configured to extract the low-spectrum energy ratio, the main frequency drift and the peak-to-peak value of the pressure pulsation based on the furnace sound pressure signal and the furnace pressure pulsation signal, and generate combustion instability characterization quantities. The model-predicted air distribution decision unit is further configured to use combustion instability characterization as the trigger and constraint update quantity for rolling optimization. When the combustion instability characterization exceeds a preset threshold, at least one of the following linked updates is performed: Increase the weight of the carbon monoxide concentration penalty term in the objective function or tighten the upper limit constraint of the carbon monoxide concentration in the flue gas, tighten the furnace negative pressure range constraint or increase the weight of the furnace negative pressure deviation penalty, limit the secondary air volume distribution ratio within the preset vibration suppression feasible domain and impose an upper limit constraint on the rate of change of the opening of the secondary air local tuyeres. Based on the rolling optimization results after linkage update, the set value of the total primary air volume, the set value of the total secondary air volume, and the air volume distribution ratio of each air outlet are output to achieve early suppression of combustion instability and simultaneously meet the constraints of the range of exhaust oxygen content and the threshold constraints of slagging risk characterization quantity. In this embodiment, the furnace and flue gas parameter acquisition unit is further configured to acquire furnace sound pressure signal and furnace pressure pulsation signal. The sound pressure signal is preferably acquired by a high-temperature resistant acoustic probe and connected to the furnace through a sampling tube with cooling and purging to avoid dust blockage. The pressure pulsation signal is preferably acquired by a differential pressure or piezoelectric dynamic pressure sensor and arranged in the upper part of the furnace or the uniform mixing section of the flue. The sampling frequency of sound pressure and pressure pulsation is preferably 200Hz to 2kHz and synchronized with the controller clock. After the sampled data enters the controller, it is first bandpass filtered to remove power frequency and high frequency noise and then feature extraction is performed using a sliding window. The sliding window length is preferably 2s to 10s and the window update step size is preferably 0.5s to 2s, so as to obtain real-time stability information that can be used for rolling optimization triggering and constraint update without increasing the parallel control loop. In this embodiment, the combustion state prediction model unit is further configured to generate a combustion instability characterization quantity based on the furnace sound pressure signal and the furnace pressure pulsation signal. Specifically, it performs rapid spectral analysis on the sound pressure sequence and pressure sequence within each sliding window to obtain the spectral energy distribution. The low-frequency energy ratio E is defined as the ratio of low-frequency band energy to full-frequency band energy, wherein the low-frequency band is preferably 0.5Hz to 20Hz or pre-tuned according to the boiler structure resonance characteristics. The dominant frequency drift D is defined as the absolute value of the difference between the dominant frequency of the current window and the dominant frequency of the previous window. The pressure pulsation peak value Ppp is defined as the difference between the maximum and minimum values ​​of the pressure sequence within the window. Then, the combustion instability characterization quantity U is generated in a weighted manner. U is preferably equal to a multiplied by E plus b multiplied by D plus c multiplied by Ppp, wherein abc are the weights obtained by step feeding or step air volume test during the boiler commissioning stage and stored in the controller parameter table. A first-order smoothing is set for U to obtain Uf to suppress occasional spikes. In this embodiment, the model-predicted air distribution decision unit is further configured to use Uf as the trigger and constraint update quantity for rolling optimization and write it into the same rolling optimization solution process. When Uf does not exceed the preset threshold Uth, rolling optimization is performed using the objective function weights and constraint ranges, and Qa, Qb, and β are output. When Uf exceeds Uth, at least one or more of the following linked updates are executed and the rolling optimization problem is continued to be solved without establishing a new parallel control loop: First, increase the weight of the flue gas carbon monoxide penalty term or tighten the upper limit constraint of carbon monoxide to a lower critical value to suppress the pulsation induced by incomplete combustion; Second, increase the weight of the furnace negative pressure deviation penalty or tighten the furnace negative pressure range constraint and simultaneously increase the weight of the furnace negative pressure deviation penalty term. The high rate of change penalty of the induced draft control quantity is used to suppress pressure oscillation. Third, the secondary air distribution ratio β is limited to the preset vibration suppression feasible domain and a stricter upper limit is applied to the rate of change of the local secondary air outlet opening to reduce the excitation of the flame by jet disturbance. Fourth, a margin factor is added to the constraint of the slagging risk characterization quantity Sz to avoid the risk of superposition of high heat load on the wall in unstable state. After the model predicts the air distribution decision unit completes the above linkage update, it outputs new QaQb and β and is executed by the air distribution execution unit. Then, the furnace and flue gas parameter acquisition unit continues to feed back O2COPf and Uf to enter the next round of rolling optimization, thereby realizing the early identification of combustion instability and the adaptive update of constraints within the same optimizer.

[0021] like Figure 1-4 As shown, the model prediction air distribution decision unit is further configured to introduce air supply branch coordination consistency constraints and consistency penalty terms in rolling optimization. The primary air branch and each air outlet branch of the secondary air are defined as nodes to be allocated and a branch coupling relationship is established. This is used to ensure that the air volume allocation ratio of each node remains consistent with the target air distribution ratio vector determined by the fuel reactivity index, under the premise of satisfying the set value of the total primary air volume and the set value of the total secondary air volume. The consistency penalty term is used to apply a weighted penalty to the deviation in the air volume allocation ratio between the coupled branch nodes, and the health compensation factor output by the actuator health estimation unit is used as the basis for updating the weighting coefficient, so that the branch with deteriorated health status is automatically reduced in allocation weight in the rolling optimization and the branch with better health status bears the compensation air volume. This enables coordinated convergence of air distribution across multiple air outlets under fluctuating fuel quality conditions, while simultaneously satisfying constraints on the range of flue gas oxygen content, the upper limit of carbon monoxide concentration, the range of furnace negative pressure, and the threshold constraints of slagging risk characterization. In this embodiment, when solving the rolling optimization problem, the model prediction air distribution decision unit uniformly numbers the primary air branches and the secondary air outlet branches as a set of nodes, with each node including at least a primary air node and multiple secondary air nodes. Based on the geometric position of the air outlet and the wind box structure, a node coupling relationship matrix A is established. When two secondary air nodes belong to the same wind box or the center distance between their corresponding air outlets is less than a preset distance threshold, the corresponding element in A is set to be coupled effectively; otherwise, it is set to be coupled ineffectively. The model prediction air distribution decision unit calls a set of target air distribution ratio template vectors β targets based on the fuel reactivity index and the boiler load interval. The target air distribution ratio template vectors are stored as lookup table data according to load segments and fuel reactivity segments. When the slagging risk characterization quantity is close to the threshold, the target air distribution ratio of the secondary air nodes corresponding to the wall-adhering sensitive zone is reduced and corrected to obtain the target air distribution ratio vector β targets used for consistent convergence. In this embodiment, the model-predicted air distribution decision unit introduces a consistency penalty term and consistency constraints into the objective function, ensuring that the air volume allocation ratio of each node converges towards the β target while maintaining smooth consistency among coupled nodes, provided that the total primary air volume and total secondary air volume are satisfied. The consistency penalty term comprises at least two parts: a weighted sum of the squared differences in allocation ratios between coupled node pairs, and a weighted sum of the squared deviations of the allocation ratios of each node from the corresponding components of the β target. The weighting coefficients are updated using the health compensation factor output by the actuator health estimation unit. Specifically, a health score G is calculated for each air supply branch, and the health score G is taken as the reciprocal of the health compensation factor and adjusted between zero and one. The system limits the health score G so that when damper leakage or fan performance degradation leads to an increase in the health compensation factor, the health score G decreases accordingly. The model predicts that the air distribution decision unit uses the health score G to update the consistency penalty weight and simultaneously updates the upper limit of the allocation ratio of the corresponding node in the branch. This causes the nodes with lower health scores to have their allocation weights automatically reduced and their upper limit of allocation ratios tightened in the rolling optimization. At the same time, the constraint that the sum of the secondary air allocation ratios is one ensures that the remaining nodes with higher health scores bear the compensation allocation. Thus, when the equivalent flow coefficient of a certain secondary air outlet branch decreases, the rolling optimization will automatically lower the allocation ratio of that air outlet and raise the allocation ratio of other air outlets in adjacent or the same air box during the same solution process.

[0022] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. A combustion air volume control system for a biomass boiler, characterized in that, include: The fuel parameter acquisition unit is used to acquire the moisture content signal, feed rate per unit time signal, and particle size characterization signal of biomass fuel along the feeding path. The furnace and flue gas parameter acquisition unit is used to acquire furnace negative pressure signal, furnace temperature signal, flue gas oxygen content signal, and flue gas carbon monoxide concentration signal; The wall heat load acquisition unit is used to acquire the heat flux signal or wall temperature distribution signal of the water-cooled wall and generate a slagging risk characterization quantity. The fuel reactivity estimation unit is used to calculate the fuel reactivity index based on the moisture content signal, the feed rate per unit time signal, and the particle size characterization signal, and to update the fuel reactivity index online. The actuator health estimation unit is used to estimate the equivalent flow coefficient of the corresponding air supply branch and generate a health compensation factor online based on the fan current signal, fan speed signal, duct differential pressure signal and damper opening signal of each air supply branch. The combustion state prediction model unit is used to construct a prediction model that includes the bed oxygen consumption state, volatile matter release state and furnace effective temperature state, and inputs the fuel reactivity index and the equivalent flow coefficient as updatable parameters into the prediction model. The model predictive air distribution decision unit is used to solve the optimization problem based on the prediction model in the rolling time domain, so as to output the set value of the total primary air volume, the set value of the total secondary air volume, and the air volume distribution ratio of each air outlet. The constraints of the optimization problem include at least the range constraints of flue gas oxygen content, the upper limit constraints of carbon monoxide concentration, the range constraints of furnace negative pressure, and the constraint that the slagging risk characterization quantity does not exceed a preset threshold. The air distribution execution unit includes control interfaces for the primary air fan and primary damper, as well as the secondary air fan and secondary damper. It is used to predict the output of the air distribution decision unit based on the model and generate fan speed control commands and damper opening control commands in combination with the health compensation factor, thereby realizing closed-loop control of the combustion air volume of the biomass boiler.

2. The combustion air volume control system for a biomass boiler according to claim 1, characterized in that, The fuel parameter acquisition unit includes a near-infrared spectroscopy acquisition module, a microwave moisture content acquisition module, a weighing and metering module, and a particle image acquisition module. The near-infrared spectroscopy acquisition module is used to acquire the reflectance spectrum of biomass fuel within a preset wavelength range and extract spectral feature factors; the microwave moisture content acquisition module is used to acquire the fuel moisture content signal; the weighing and metering module is used to acquire the feed rate signal per unit time; and the particle image acquisition module is used to acquire fuel particle images and output particle size distribution characterization parameters. The fuel reactivity estimation unit is used to generate a fuel reactivity index based on the moisture content signal, the feed rate per unit time signal, the particle size distribution characterization parameters, and the spectral characteristic factors, and to recursively update the mapping weight of the fuel reactivity index based on the short-time response deviation between the flue gas oxygen content signal and the flue gas carbon monoxide concentration signal.

3. The combustion air volume control system for a biomass boiler according to claim 1, characterized in that, The actuator health estimation unit is configured to establish an air volume estimation model for each air supply branch; Based on the fan speed signal, fan current signal, duct differential pressure signal, duct temperature signal and damper opening signal of the air supply branch, the predicted air volume of the air supply branch is calculated. At the same time, the deviation between the measured air volume obtained by converting the duct differential pressure signal and the predicted air volume is used as the identification error, and the equivalent flow coefficient of the air supply branch is recursively updated to obtain the health compensation factor. The air distribution execution unit compensates and corrects the fan speed control command and damper opening control command according to the health compensation factor, so that the set value of the primary air total volume and the set value of the secondary air total volume still meet the output requirements of the model prediction air distribution decision unit under the conditions of damper lag, leakage or fan performance degradation.

4. A combustion air volume control system for a biomass boiler according to any one of claims 2-3, characterized in that, The wall heat load acquisition unit is configured to divide the water-cooled wall into multiple monitoring zones along the furnace height direction and circumferential direction, and output the heat flux signal or wall temperature signal of each monitoring zone respectively. The system also includes a slagging risk quantification unit, which is used to generate a slagging risk characterization quantity based on the heat flux signal or wall temperature signal of each monitoring zone, wherein the slagging risk characterization quantity is obtained by weighting at least the zone peak heat load, the zone heat load spatial gradient and the duration exceeding the preset heat load threshold within the monitoring window. The model predictive air distribution decision unit is configured to use the slagging risk characterization quantity as a constraint quantity for rolling optimization, and when the slagging risk characterization quantity approaches the threshold, prioritize adjusting the secondary air volume distribution ratio and the local air outlet opening to reduce the heat flux or wall temperature peak of the high heat load zone attached to the wall, while maintaining the exhaust oxygen content range constraint and the carbon monoxide concentration upper limit constraint to meet the preset requirements.

5. The combustion air volume control system for a biomass boiler according to claim 3, characterized in that, The model predicts the air distribution decision unit and is configured to construct a rolling optimization problem containing an objective function and constraints. The objective function includes at least a flue gas oxygen content deviation term to characterize combustion efficiency, a flue gas carbon monoxide concentration penalty term to characterize incomplete combustion, and a fan power penalty term to characterize air supply energy consumption. The flue gas carbon monoxide concentration penalty term adopts a logarithmic or piecewise penalty structure to increase the penalty weight when the carbon monoxide concentration is close to a preset upper limit. The constraints of the rolling optimization problem include at least the upper and lower limits of the total primary air volume and the total secondary air volume, the feasible region constraint of the air volume distribution ratio of each tuyer, the furnace negative pressure range constraint, the range of flue gas oxygen content constraint, the upper limit constraint of flue gas carbon monoxide concentration constraint, and the threshold constraint of the slagging risk characterization quantity. The model-predicted air distribution decision unit is also configured to adjust the weights of each penalty term in the objective function online based on the rate of change of boiler load and the rate of change of the fuel reactivity index, so as to simultaneously meet the combustion efficiency target and the stable combustion target under conditions of sudden load changes or fuel quality fluctuations.

6. The combustion air volume control system for a biomass boiler according to claim 5, characterized in that, The furnace and flue gas parameter acquisition unit is further used to acquire furnace sound pressure signal and furnace pressure pulsation signal; The combustion state prediction model unit is further configured to extract the low-spectrum energy ratio, the main frequency drift, and the peak-to-peak value of the pressure pulsation based on the furnace acoustic pressure signal and the furnace pressure pulsation signal, and generate combustion instability characterization quantities. The model-predicted wind distribution decision unit is further configured to use the combustion instability characterization quantity as the trigger quantity and constraint update quantity for rolling optimization. When the combustion instability characterization quantity exceeds a preset threshold, at least one of the following linked updates is performed: Increase the weight of the carbon monoxide concentration penalty term in the objective function or tighten the upper limit constraint of the carbon monoxide concentration in the flue gas, tighten the furnace negative pressure range constraint or increase the weight of the furnace negative pressure deviation penalty, limit the secondary air volume distribution ratio within the preset vibration suppression feasible domain and impose an upper limit constraint on the rate of change of the opening of the secondary air local tuyeres. Based on the rolling optimization results after the linkage update, the set value of the total primary air volume, the set value of the total secondary air volume, and the air volume distribution ratio of each air outlet are output to achieve early suppression of combustion instability and simultaneously meet the constraints of the range of exhaust oxygen content and the threshold constraints of the slagging risk characterization quantity.

7. The combustion air volume control system for a biomass boiler according to claim 5, characterized in that, The model prediction air distribution decision unit is further configured to introduce air supply branch coordination consistency constraints and consistency penalty terms in the rolling optimization, wherein the primary air branch and each air outlet branch of the secondary air are defined as nodes to be allocated and a branch coupling relationship is established, so as to ensure that the air volume allocation ratio corresponding to each node remains consistent with the target air distribution ratio vector determined by the fuel reactivity index, under the premise of satisfying the set value of the total primary air volume and the set value of the total secondary air volume. The consistency penalty term is used to apply a weighted penalty to the deviation in the air volume allocation ratio between the coupled branch nodes, and the health compensation factor output by the actuator health estimation unit is used as the basis for updating the weighting coefficient, so that the branch with deteriorated health status is automatically reduced in allocation weight in the rolling optimization and the branch with better health status bears the compensation air volume. This allows for coordinated convergence of air distribution across multiple air outlets under fluctuating fuel quality conditions, while simultaneously satisfying constraints on the range of flue gas oxygen content, the upper limit of carbon monoxide concentration, the range of furnace negative pressure, and the threshold constraints of slagging risk characterization.

Images