A multi-modal sensing air preheater cold end regulation method and system

Abstract

The application provides a multi-modal perception air preheater cold end regulation method. It comprises: collecting multi-modal data including flue gas parameters, temperature field data and air preheater operation parameters, calculating dynamic safety margin based on acid dew point temperature, ammonia escape rate and sulfur content of coal into the furnace, and then constructing dynamic constraint boundary with corrosion prevention and control temperature threshold, upper limit of total hot air flow and upper limit of cold end average temperature as the core, and establishing an optimization objective function with the minimum total hot air consumption as the target; the rotor cold end is divided into 4-8 independent regulation regions along the circumference, the temperature deviation of each region is calculated, and the hot air flow distribution weight of each partition is determined according to the deviation; the particle swarm optimization algorithm population is initialized with the weight, the objective function is solved under the dynamic constraint boundary, the optimal hot air flow distribution scheme is obtained, and the regulation is executed. The application realizes the partition control of the cold end temperature, considers the corrosion prevention and control, energy consumption optimization and ash deposition inhibition, and effectively prolongs the service life of the heat accumulating element.

Description

Technical Field

[0001] This application relates to the field of air preheater technology, and in particular to a multimodal sensing method and system for cold-end control of air preheaters. Background Technology

[0002] Rotary air preheaters are heat exchange equipment in coal-fired power units. Due to their low wall temperature in the cold end region, when the temperature falls below the flue gas acid dew point, sulfuric acid vapor in the flue gas condenses on the surface of the heat storage elements, causing severe low-temperature corrosion and sticky ash accumulation. Especially for units using medium-to-high sulfur coal, severe corrosion can lead to large-area perforation and collapse of the heat storage elements, resulting in high replacement costs per unit. At the same time, ash accumulation and blockage can increase the differential pressure of the air preheater, increase the power consumption of the induced draft fan, and significantly reduce the economic efficiency of the unit's operation.

[0003] Existing technologies primarily raise the cold-end temperature through methods such as warm air heaters and hot air recirculation. However, their control largely relies on empirical formulas to calculate the acid dew point, resulting in significant deviations from the actual acid dew point under real-world operating conditions. This leads to either excessively large temperature margins causing excessive heat loss or insufficient control resulting in localized condensation. Furthermore, existing solutions often employ a global heating mode, which is difficult to adapt to complex operating conditions such as deep peak shaving and frequent coal type fluctuations. Control lag times are long during load changes, making them prone to periodic sulfuric acid condensation problems. Currently, there is a lack of multi-objective coordinated control technologies that simultaneously address corrosion prevention, energy consumption optimization, and ash accumulation suppression, becoming a common challenge restricting the safe and economical operation of generating units. Summary of the Invention

[0004] This application proposes a multimodal sensing method for controlling the cold end of an air preheater that simultaneously meets the requirements of corrosion prevention, low energy consumption, and dust accumulation under multiple operating conditions.

[0005] A multimodal sensing air preheater cold end control system is also proposed.

[0006] A multimodal sensing method for controlling the cold end of an air preheater includes the following steps: Step 1: Collect multimodal data, including flue gas parameters, temperature field data, and air preheater operating parameters. Based on the collected multimodal data, construct dynamic constraint boundaries and establish an objective function. Step 2: Divide the rotor cold end into circumferential zones and calculate the temperature deviation of each zone. Determine the hot air flow distribution weight of each zone according to the temperature deviation. Step 3: Initialize the initial population of the particle swarm optimization algorithm with the weights, solve the objective function under the dynamic constraint boundary, obtain the optimal hot air flow distribution scheme, and regulate the hot air flow of each zone accordingly.

[0007] A multimodal sensing air preheater cold end control system, comprising: The multimodal sensing module consists of an acid dew point meter, an SO3 analyzer, an ammonia slip monitor, an area array infrared thermal imager, a differential pressure transmitter, and an online coal quality analyzer. It is used to collect flue gas parameters, temperature field data, and air preheater operating parameters in real time. The targeted hot air execution module consists of 4 to 8 independently controlled hot air compartments, electric regulating dampers, and flow transmitters. The hot air source is the hot secondary air duct at the air preheater outlet, and it is used to independently deliver hot air to each zone according to control commands. The multi-objective optimization control module, equipped with the aforementioned control algorithm, receives real-time data from the multimodal sensing module, performs optimization solutions, and outputs control signals for each compartment's damper.

[0008] The technical advantages of this application are as follows: The accuracy of cold end temperature control has been greatly improved: This application obtains key parameters such as acid dew point, ammonia escape rate and coal quality in real time through multimodal sensing, and dynamically calculates the corrosion prevention and control temperature threshold. The temperature margin control accuracy has been improved from ±15℃ in the traditional method to within ±3℃, which solves the problem of local sulfuric acid condensation at the cold end. The heat loss of the hot air system is significantly reduced: Compared with the traditional global heating method, this application reduces the total hot air consumption by 30% to 50% through zoned targeted delivery and particle swarm optimization, and the power consumption of the induced draft fan is reduced accordingly. Significantly effective in suppressing ash accumulation: Improved uniformity of cold-end temperature field, reduced air preheater soot blowing frequency by 40%–60%, and reduced differential pressure growth rate by approximately 50%; The lifespan of the heat storage element is significantly extended: when operating under the temperature threshold constraint of corrosion control, the corrosion rate of the heat storage element is reduced by more than 60%, and the service life is extended from 1-2 years to 3-5 years; Strong adaptability to operating conditions: The dynamic constraint boundary is updated in real time with parameters such as load, coal type, and ammonia slip, which can adapt to complex operating conditions such as deep peak shaving of the unit and frequent switching of coal type. Detailed Implementation

[0009] The embodiments of the technical solution of this application will be described in detail below. The following embodiments are only used to illustrate the technical solution of this application more clearly, and are therefore only examples, and should not be used to limit the scope of protection of this application.

[0010] A multimodal sensing method for controlling the cold end of an air preheater includes the following steps: Step 1: Collect multimodal data, including flue gas parameters, temperature field data, and air preheater operating parameters. Based on the collected multimodal data, construct dynamic constraint boundaries and establish an objective function. Step 2: Divide the rotor cold end into circumferential zones and calculate the temperature deviation of each zone. Determine the hot air flow distribution weight of each zone according to the temperature deviation. Step 3: Initialize the initial population of the particle swarm optimization algorithm with the weights, solve the objective function under the dynamic constraint boundary, obtain the optimal hot air flow distribution scheme, and regulate the hot air flow of each zone accordingly.

[0011] In a preferred embodiment, the flue gas parameters include the air preheater inlet acid dew point temperature. T dp SO3 concentration C SO3 ammonia escape rate R NH3 Sulfur content of coal fed into the furnace S ar carbon content in fly ash C ash .

[0012] The temperature field data was acquired using a planar infrared thermal imager, and the temperature of each region is expressed as follows: T i,j ,in i For area code, j Number the measurement points within the area; The operating parameters include unit load. L Differential pressure of air preheater ΔP Hot air temperature T hot Flow rate of each hot air compartment Q k .

[0013] Step 1 sets the optimization objective as minimizing total hot air consumption, and sets constraints as follows: "Temperature in all areas must not fall below the corrosion control threshold, total hot air flow must not exceed the upper limit, and average cold end temperature must not exceed the upper limit." This forms a constrained single-objective optimization problem, where: Corrosion control temperature threshold (core constraint) Through collection R NH3 , S ar Parameter calculation of dynamic safety margin K : The unit is ℃; It is a rounding function. To obtain The larger of the values ​​in 0, i.e., when the ammonia slip rate When the concentration exceeds 3 ppm, an additional 2°C safety margin is added for every 1 ppm increase, reflecting the increasing risk of ammonium bisulfate (ABS) formation. This is an indicator function, which takes the value of 1 when the condition is met, and 0 otherwise. Sar > 1.5% means that when the sulfur content of the coal fed into the furnace exceeds 1.5%, an additional 5°C safety margin is added, reflecting the aggravating effect of the increased partial pressure of sulfuric acid vapor on the risk of cold-end corrosion under high-sulfur coal conditions.

[0014] Based on the dynamic safety margin K, calculate the corrosion control temperature threshold: T thre = , Temperature constraints for each region are as follows T thre ≤T i,j This means that the temperature in all areas of the cold end must not fall below this threshold. Maximum total hot air flow rate (efficiency constraint) Based on collected unit load L Hot air temperature T hot Determine the upper limit of the total hot air flow: Q total,max = 8% × ( L / L rated ) ×Q rated , in, L This represents the current unit load (MW). L rated This refers to the rated load (MW) of the unit. Q rated This refers to the total secondary air flow rate (Nm³ / h) under rated operating conditions. This upper limit controls the hot air intake to within 8% of the total secondary air volume, thus avoiding a significant impact on the boiler combustion conditions.

[0015] Upper limit of average temperature at the cold end (dust accumulation suppression constraint) Based on the collected fly ash carbon content C ash Differential pressure of air preheater ΔP Determine the upper limit of the average temperature of the cold end: T avg,max = 160℃-5℃ × [ C ash > 8%)]; Specifically, when the carbon content (Cash) of fly ash exceeds 8%, the upper limit of the average cold-end temperature decreases by 5°C. This is because a high carbon content in fly ash indicates incomplete combustion. Unburned carbon particles deposited on the surface of the heat storage element may undergo exothermic oxidation or even localized sintering at higher temperatures, exacerbating ash adhesion. Appropriately lowering the upper limit of the average cold-end temperature can suppress this effect.

[0016] objective function The objective function is to minimize the total hot air consumption. ; in, Q k For the first k Hot air flow rate of each zone (unit: Nm³) 3 / h), N This represents the total number of partitions.

[0017] In summary, the constrained optimization model constructed in step 1 is as follows: Target:

[0018] Constraints: T thre ≤ T̄ k ( k = 1, 2, ..., N Corrosion control constraints). Σ NQ k ≤ Q total,max (Total hot air volume constraint); (1 / N ) Σ NT̄ k ≤ T avg,max (Dust accumulation suppression constraint); Q k ≥ 0 ( k = 1, 2, ..., N (non-negativity constraint). In a preferred embodiment, step 2, "dividing the rotor cold end into circumferential zones and calculating the temperature deviation of each zone, and determining the hot air flow distribution weight of each zone according to the temperature deviation," specifically includes the following steps: Step 21: Divide the cold end temperature field into N ( N Select 4 to 8 independent control zones, each corresponding to a hot air compartment. The number of zones is determined based on the air preheater diameter and the required control precision: for units with smaller diameters, [the following can be used]. N=4, for units with larger diameters, the following can be taken: N =6 or 8.

[0019] Step 22: Calculate the temperature deviation for each region: Δ T k = T thre - T̄ k ( k = 1, 2, ..., N ) in, T̄ k For the first k The arithmetic mean of the temperatures at all measuring points within the region: T̄ k = (1 / M k ) Σ j=1 M k T k,j in, M k For the first k The total number of measuring points within a given area.

[0020] When Δ T k When Δ > 0, it indicates that the temperature in the area is below the corrosion control threshold, and hot air needs to be supplied; when Δ T k When the temperature is ≤ 0, it means that the temperature in the area has reached the standard and no additional hot air needs to be allocated.

[0021] Step 23: For Δ T k For low-temperature regions where the temperature is >0, calculate the hot air flow distribution weights: w k = max(Δ T k , 0) w̃ k = w k / Σ i=1 Nw i (Normalized weights) Normalized weights w̃ k Satisfy Σ w̃ k= 1, used to generate the initial population center point for the particle swarm algorithm in step 3.

[0022] Step 3: Weight Initialization, Particle Swarm Optimization and Control Execution This application employs an improved particle swarm optimization (PSO) algorithm to solve the optimization model constructed in step 1. The difference from traditional PSO lies in that this application uses the weight distribution obtained in step 2 as the basis for generating the initial population, rather than completely random initialization, thereby significantly accelerating the convergence speed and improving the quality of the solution.

[0023] The following are the parameter settings for the PSO algorithm in this application:

[0024] Based on the normalized weights obtained in step 23 w̃ k Generate an initial traffic allocation scheme: Q k 0 = w̃ k × Q total,max ( k = 1, 2, ..., N ) The positions of each particle in the initial population are Q 0 The initial value of the hot air flow rate allocated to the k-th partition in the nearby random distribution, i.e., in the m-th candidate scheme, is as follows: Q k ( m ) (0) = Q k 0 × (1 + U (-0.3, 0.3)) ( m = 1, 2, ..., M ; k = 1, 2, ..., N ) in, U (-0.3, 0.3) is a uniformly distributed random number in the interval [-0.3, 0.3], ensuring that the initial search space covers ±30% of the initial scheme.

[0025] Each particle m In the k The dimension, that is, the first dimension k The speed of hot air flow in each zone v kThe position update formula is: v k ( t +1) = w · v k ( t ) + c 1· r 1·( p best,k - x k ( t )) + c 2. r 2·( g best - x k ( t )) x k ( t +1) = x k ( t ) + v k ( t +1) The following explains the physical meaning of each symbol in the formula within the context of this application: x k ( t ): No. k Each particle at time t The position vector, in this scheme, is the hot air flow distribution scheme for each zone represented by the particle; v k ( t ): No. k Each particle at time t The velocity vector represents the direction and magnitude of the adjustment of the particle's flux allocation scheme in the next iteration; p best,k : No. k The individual historical optimal position vector of a particle, that is, the flow allocation scheme with the minimum objective function value found by the particle in each iteration. g best The global optimal position vector is the flow distribution scheme with the minimum objective function value found by all particles in each iteration. w The inertia weight ranges from 0.4 to 0.9, decreasing linearly with the number of iterations. The specific formula is as follows: w( t ) = w start - ( w start - w end ) × t / T max ( w start =0.9, w end =0.4), used to balance global exploration and local development capabilities; c 1. c 2 represents the learning factor, all of which are set to 2.0; r 1. r 2 is a uniformly distributed random number in the interval [0,1], which is generated independently in each iteration; t represents the current iteration number.

[0026] This application employs a rejection method to handle constraint violations: for each particle's updated position, four constraints are checked sequentially (corrosion control constraint, total hot air volume constraint, ash accumulation suppression constraint, and non-negativity constraint). If any constraint is not satisfied, the position is discarded, the particle is reset to its individual historical best position, and a new position is randomly generated within the feasible region.

[0027] The algorithm stops when any of the following conditions are met: The global optimal objective function value no longer decreases after 10 consecutive iterations; Reaching the maximum number of iterations T max =100.

[0028] The algorithm outputs a globally optimal traffic allocation scheme Q* = [Q1*,Q2*, ..., Q... N *] Based on this, the multi-objective optimization control module generates opening control signals for the electric regulating dampers of each compartment, thereby regulating the hot air flow of each zone.

[0029] The following numerical example illustrates the specific implementation process of the method in this application.

[0030] The operating parameters are shown in the table below:

[0031] Build an optimization model: Dynamic safety margin: K = ⌈max(4 - 3, 0) × 2⌉ + [1.8% > 1.5%] × 5 = ⌈2⌉ + 1 × 5 = 7℃ Corrosion control temperature threshold: T thre = 118 + 7 = 125℃ Maximum total hot air flow rate: Q total,max = 8% × (240 / 300) × 500,000 = 32,000 Nm 3 / h = 32,000 Nm 3 / h Upper limit of average cold end temperature: T avg,max = 160 - 5 × [6% > 8%] = 160 - 0 = 160℃ Partition weight calculation The average temperature of each area is measured in real time by an infrared thermal imager.

[0032] (Note: Δ) T k Region ≤ 0 w k Set to 0; the normalized weights for regions 2, 3, and 5 are 0.22, 0.50, and 0.28, respectively. Particle Swarm Optimization Generate an initial scheme based on the weights: Q 0 = [0, 7040, 16000, 0, 8960, 0] Nm³ / h = [0, 0.70 million, 1.60 million, 0, 0.90 million, 0] Nm 3 / h Initial objective function value: Σ Q k = 32,000 Nm 3 / h = 32,000 Nm 3 / h.

[0033] Particle swarm optimization searches near the initial scheme (partial typical iterative process):

[0034] The solution converges in the 50th round, the global optimum no longer decreases, and the final solution is output: Q * = [0, 5300, 11200, 0, 6500, 0] Nm 3 / h Σ Q k * = 23,000 Nm 3 / h = 23,000 Nm 3 / h Constraint verification: Zone 1 temperature: 130℃ ≥ 125℃, meets the requirements; Zone 2 temperature: 125.5℃ ≥ 125℃, meets the requirement; Zone 3 temperature: 125.2℃ ≥ 125℃, meets the requirements; Zone 4 temperature: 128℃ ≥ 125℃, meets the requirements; Zone 5 temperature: 125.8℃ ≥ 125℃, meets the requirements; Zone 6 temperature: 131℃ ≥ 125℃, meets the requirements; Total hot air volume: 23,000 ≤ 32,000 Nm 3 / h / meets the requirements; Average cold end temperature: (130+125.5+125.2+128+125.8+131) / 6 = 127.6℃ ≤ 160℃, which meets the requirements; All constraints are satisfied, and the solution is feasible.

[0035] The following is an example of a 300MW coal-fired unit according to this application.

[0036] Perception layer configuration: One conductivity acid dew point meter with a range of 60 to 160℃ and an accuracy of ±1℃ is installed in the inlet flue of the air preheater. One SO3 analyzer is installed in the inlet flue of the air preheater, with a range of 0 to 50 ppm and an accuracy of ±0.5 ppm; One ammonia slip monitor is installed in the air preheater outlet flue. It is based on tunable diode laser absorption spectroscopy (TDLAS) technology, with a range of 0 to 10 ppm and an accuracy of ±0.3 ppm. One area array infrared thermal imager with a resolution of 128×128, a temperature measurement range of 0~200℃, and an accuracy of ±2℃ is installed on the cold end flue gas outlet side. One high-precision differential pressure transmitter is installed in each of the air preheater inlet and outlet flues, with a range of 0–3000 Pa and an accuracy of ±0.5%. One online coal quality analyzer is installed at the outlet of the coal feeder, which can detect parameters such as sulfur content and ash content of the coal entering the furnace in real time.

[0037] Execution layer configuration: Six independent hot air compartments are evenly arranged along the circumference of the cold end of the air preheater, with each compartment covering a 60° arc. Each compartment is equipped with one electric regulating damper and one vortex flow transmitter. Hot air is drawn from the secondary air duct and introduced into each hot air compartment via a bypass pipe. The rated total flow rate is 8% of the total secondary air volume.

[0038] Control layer configuration: It employs an edge computing controller and is equipped with an improved particle swarm optimization algorithm, with a control cycle of 30 seconds. The system connects to the unit's DCS system via Modbus TCP / IP protocol to achieve bidirectional transmission of operating parameters.

[0039] Typical operational data after implementing the proposed solution are as follows:

[0040] The above data shows that the proposed solution has achieved significant improvements in three dimensions: corrosion control, energy consumption optimization, and dust accumulation suppression.

[0041] The hot air source for this application is the hot secondary air duct at the air preheater outlet. It should be noted that this solution does not draw all the hot secondary air from the air preheater outlet back to the cold end for reheating. Instead, it draws only a very small proportion (no more than 8% of the total secondary air under rated operating conditions) as compensating hot air, precisely supplementing the energy gap in areas with insufficient cold-end temperature through zoned targeted distribution. The impact of this drawn portion on the boiler's total air volume and combustion conditions is negligible, but its benefits in controlling cold-end corrosion far outweigh the increased flue gas heat loss caused by drawing a small amount of hot air. Furthermore, compared to traditional air heater solutions (which require additional steam consumption), this solution directly utilizes the waste heat from the air preheater's own outlet, without increasing external energy consumption, resulting in a significant energy efficiency advantage.

[0042] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A method for regulating the cold end of a multi-modal sensing air preheater, characterized by, Includes the following steps: Step 1: Collect multimodal data, including flue gas parameters, temperature field data, and air preheater operating parameters. Based on the collected multimodal data, construct dynamic constraint boundaries and establish an objective function. Step 2: Divide the rotor cold end into circumferential zones and calculate the temperature deviation of each zone. Determine the hot air flow distribution weight of each zone according to the temperature deviation. Step 3: Initialize the initial population of the particle swarm optimization algorithm with the weights, solve the objective function under the dynamic constraint boundary, obtain the optimal hot air flow distribution scheme, and regulate the hot air flow of each zone accordingly.

2. The multimodal sensing air preheater cold end control method as described in claim 1, characterized in that, The flue gas parameters include the air preheater inlet acid dew point temperature. T dp SO3 concentration C SO3 ammonia escape rate R NH3 Sulfur content of coal fed into the furnace S ar carbon content in fly ash C ash The temperature field data was acquired using an area array infrared thermal imager, and the temperature of each region is expressed as follows: T i,j ,in i For area code, j The measurement points within the area are numbered; the operating parameters include unit load. L Δ air preheater differential pressure P Hot air temperature T hot Flow rate of each hot air compartment Q k .

3. The multimodal sensing air preheater cold end control method as described in claim 1, characterized in that, The dynamic constraint boundary mentioned in step 1 includes the corrosion control temperature threshold. T thre Maximum total hot air flow rate Q total,max and upper limit of average temperature of cold end T avg,max ; Among them, the corrosion control temperature threshold T thre Calculate using the following steps: Step 11: Based on the collected ammonia slip rate R NH3 and the sulfur content of the coal fed into the furnace S ar Calculate dynamic safety margin K : The unit is ℃; in, It is a rounding function. This is an indicator function that takes the value 1 when the condition is true and 0 otherwise. Step 12: Calculate the corrosion control temperature threshold: T thre = T dp + K Temperature constraints for each region are as follows: T thre ≤ T i,j .

4. The multimodal sensing air preheater cold end control method as described in claim 1, characterized in that, The upper limit of total hot air flow rate mentioned in step 1 Q total,max Calculated using the following formula: Q total,max = 8% × ( L / L rated ) × Q rated in, L This represents the current unit load (MW). L rated This refers to the rated load (MW) of the unit. Q rated The total secondary air flow rate under rated operating conditions (Nm³) 3 / h).

5. The multimodal sensing air preheater cold end control method as described in claim 1, characterized in that, The upper limit of the average cold end temperature mentioned in step 1 T avg,max Calculated using the following formula: T avg,max = 160 - 5 × [ C ash > 8%] in, C ash The carbon content of fly ash, [ C ash [>8%] is an indicator function; when the carbon content of fly ash exceeds 8%, the upper limit of the average temperature of the cold end is reduced by 5°C to avoid unburned carbon causing high-temperature sintering on the surface of the heat storage element.

6. The multimodal sensing air preheater cold end control method as described in claim 1, characterized in that, The objective function described in step 1 is to minimize the total hot air consumption: minutes F = S NQ k in, Q k For the first k Hot air flow rate of each zone N This represents the total number of partitions.

7. The multimodal sensing air preheater cold end control method as described in claim 1, characterized in that, Step 2 specifically includes the following sub-steps: Step 21: Divide the cold end temperature field into 4 to 8 independent control zones, with each zone corresponding to a hot air compartment; Step 22: Calculate the temperature deviation for each region: D T k = T thre - T̄ k in, T̄ k For the first k The average temperature of each region; Step 23: For Δ T k In the low-temperature region > 0, hot air flow weights are assigned. w k ,and w k With Δ T k Positive correlation.

8. The multimodal sensing air preheater cold end control method as described in claim 1, characterized in that, The particle swarm optimization algorithm described in step 3 specifically includes: M particles are randomly generated to form an initial population, and each particle represents a set of hot air flow distribution scheme vectors for each zone. An initial flow allocation scheme is generated proportionally based on the weights obtained in step 2. Q 0 Distribute the particles in the initial population in Q 0 Nearby, search area is Q 0 ±30%; The objective function value is calculated for each particle and a constraint check is performed. Particles that do not meet the constraints are eliminated and regenerated within the feasible region. The particle updates its velocity and position based on its individual optimal position and the global optimal position, iterating until the convergence condition is met, and then outputs the global optimal flow allocation scheme.

9. The multimodal sensing air preheater cold end control method as described in claim 8, characterized in that, The formulas for updating the velocity and position of the particles are as follows: v k ( t +1) = w · v k ( t ) + c 1· r 1·( p best,k - x k ( t )) + c 2· r 2·( g best - x k ( t )) x k ( t +1) = x k ( t ) + v k ( t +1) The following explains the physical meaning of each symbol in the formula within the context of this application: x k ( t ): No. k Each particle at time t The position vector, in this scheme, is the hot air flow distribution scheme for each zone represented by the particle; v k ( t ): No. k Each particle at time t The velocity vector represents the direction and magnitude of the adjustment of the particle's flux allocation scheme in the next iteration; p best,k : No. k The individual historical optimal position vector of a particle, that is, the flow allocation scheme with the minimum objective function value found by the particle in each iteration. g best The global optimal position vector is the flow distribution scheme with the minimum objective function value found by all particles in each iteration. w This is an inertial weight, with a value ranging from 0.4 to 0.9, which decreases linearly with the number of iterations. It is used to balance global exploration and local development capabilities. c 1. c 2 represents the learning factor, all of which are set to 2.0; r 1. r 2 is a uniformly distributed random number in the interval [0,1], which is generated independently in each iteration; t represents the current iteration number.

10. A multimodal sensing air preheater cold-end control system, characterized in that, include: The multimodal sensing module consists of an acid dew point meter, an SO3 analyzer, an ammonia slip monitor, an area array infrared thermal imager, a differential pressure transmitter, and an online coal quality analyzer. It is used to collect flue gas parameters, temperature field data, and air preheater operating parameters. The targeted hot air execution module consists of 4 to 8 independently controlled hot air compartments and their matching electric regulating dampers and flow transmitters. The hot air source is the hot secondary air duct at the air preheater outlet, which is used to independently deliver hot air to each zone according to control commands. The multi-objective optimization control module is equipped with the control algorithm in any one of claims 1 to 9, and is used to receive data from the multimodal sensing module, perform optimization solutions, and output control signals for each compartment damper.

Images