A flue gas cooling medium dynamic selection method based on water source state monitoring

Abstract

This invention relates to the field of equipment control technology, and more particularly to a dynamic selection method for flue gas cooling media based on water source status monitoring. This method collects and aligns multi-source operating condition time-series data, inputs it into a heat exchange prediction model to output predicted heat exchange efficiency and predicted equipment loss probability, and generates an initial allocation strategy through a heuristic optimization algorithm. It then combines electrical operating boundary data for safety calibration to generate a safe allocation scheme that drives the target execution unit to switch media. Furthermore, it extracts water quality characteristic parameters and calculates the prior scaling rate and transient thermal resistance increment of the heated surface through a material physical degradation mechanism model. Based on this, it performs prior evolution correction on the heat transfer coefficient in the heat exchange prediction model and simultaneously narrows the search space of feasible solutions for the heuristic optimization algorithm in the next control cycle. This invention achieves forward-looking adjustment of the control strategy, effectively alleviating the lag of traditional post-feedback control and improving the real-time performance of online control and the safety of equipment operation.

Description

Technical Field

[0001] This invention relates to the field of equipment control technology, and in particular to a method for dynamically selecting flue gas cooling media based on water source status monitoring. Background Technology

[0002] In large-scale industrial flue gas treatment and heat recovery systems for power, metallurgy, and chemical industries, using cooling media (such as fresh water, circulating wastewater, and reverse osmosis concentrate) to spray or exchange heat with high-temperature flue gas is a crucial step in ensuring the safe operation of subsequent flue gas purification equipment and improving the overall energy-saving and emission-reduction benefits of the system. Traditional flue gas cooling media formulation mainly relies on preset fixed empirical ratios or threshold switching control based on threshold rules. In recent years, with the improvement of industrial automation, some solutions have begun to introduce temperature feedback-based control algorithms or simple data-driven predictive models, passively adjusting the proportions of various cooling media and valve openings based on real-time fluctuations in flue gas outlet temperature.

[0003] However, existing technologies have significant technical shortcomings in complex industrial applications. On the one hand, control and regulation suffer from severe lag. Due to the non-steady-state fluctuations in industrial flue gas conditions and source water quality, existing passive feedback mechanisms based on outlet temperature residuals often only begin to adjust after the flue gas overheats or scale buildup on the heated surfaces causes temperature runaway. This makes it difficult to cope with transient shocks and can easily lead to cooling failures or equipment overload. On the other hand, there is a lack of awareness of the physical mechanisms of source water quality degradation. Different water sources (especially high-hardness concentrated water) have vastly different ion concentrations, which can easily lead to crystallization and scale buildup on heated surfaces and a sharp increase in thermal resistance over long-term operation. Existing pure black-box prediction models, lacking material degradation and thermodynamic constraints, cannot detect the decrease in thermal conductivity caused by scale buildup, resulting in severely distorted model prediction performance. Summary of the Invention

[0004] To overcome the above deficiencies, this invention provides a dynamic selection method for flue gas cooling media based on water source status monitoring, aiming to improve the technical problems of severe lag in control and regulation and lack of perception of the physical mechanism of water quality deterioration in existing flue gas cooling and distribution technologies.

[0005] This invention provides the following technical solution: a method for dynamically selecting flue gas cooling media based on water source status monitoring, comprising:

[0006] S1. Collect and align water source status data, flue gas operating condition data and environmental data of the target area to obtain multi-source time series data;

[0007] S2. Input the multi-source time-series data into the heat exchange prediction model and output the predicted heat exchange efficiency and predicted equipment loss probability of each candidate cooling medium.

[0008] S3. Input the predicted heat exchange efficiency and the predicted equipment loss probability into the heuristic optimization algorithm. The heuristic optimization algorithm performs optimization calculations under the preset global energy consumption, cooling target and equipment loss constraints, and outputs the initial allocation strategy of the cooling medium.

[0009] S4. Extract the electrical operating boundary data of the target execution unit, evaluate the expected electrical load corresponding to the initial allocation strategy, and selectively adjust the initial allocation strategy by linear attenuation based on the comparison result of the expected electrical load and the preset safety margin to generate a safe allocation scheme.

[0010] S5. Drive the target execution unit to switch the cooling medium according to the safety allocation scheme, and extract the water quality characteristic parameters of the source water in the multi-source time series data. Calculate the a priori scaling rate and transient thermal resistance increment of the heating surface corresponding to the current working condition through the material physical degradation mechanism model.

[0011] S6. Based on the prior scaling rate of the heated surface and the transient thermal resistance increment, the thermal conductivity coefficient in the heat exchange prediction model is modified by prior evolution, and the search space of the feasible solution of the heuristic optimization algorithm in the next control cycle is reduced.

[0012] Preferably, in S1, obtaining multi-source time-series data includes:

[0013] Anomaly detection is performed on the collected water source status data, flue gas operating condition data, and environmental data. Outliers are removed to obtain the cleaned operating data.

[0014] The cleaned running data is time-series aligned, and missing items generated during the time-series alignment process are filled in to obtain the multi-source time-series data.

[0015] Preferably, in S2, the heat exchange prediction model consists of a backbone model and a heat exchange loss function, wherein the backbone model is an LSTM model, and the heat exchange loss function includes a data error term and a physical constraint regularization term.

[0016] The data error term is used to characterize the mean square error between the predicted heat exchange efficiency and the actual heat exchange efficiency of each candidate cooling medium.

[0017] The physical constraint regularization term is constructed based on the law of conservation of energy in fluid thermodynamics and is used to characterize the degree of deviation of the predicted heat exchange efficiency from the preset heat balance equation between flue gas and cooling medium.

[0018] Preferably, in S3, the heuristic optimization algorithm is one of genetic algorithm, particle swarm optimization, simulated annealing algorithm, or ant colony optimization algorithm, and the initial formulation strategy of the output cooling medium includes:

[0019] Based on the preset global energy consumption, cooling target, and equipment loss constraints, a comprehensive optimization objective function is constructed.

[0020] Initialize the candidate solution space, which includes the type and flow rate of each candidate cooling medium.

[0021] The heuristic optimization algorithm iteratively optimizes the comprehensive optimization objective function within the candidate solution space to determine the optimal target solution that makes the comprehensive optimization objective function reach its optimum, and uses the target solution as the initial allocation strategy for the cooling medium.

[0022] Preferably, the comprehensive optimization objective function includes a penalty term corresponding to the equipment loss constraint. The penalty term includes a valve action amplitude smoothing term to suppress frequent switching of each candidate cooling medium, and a positive incentive term to increase the proportion of concentrated water reuse in the cooling medium.

[0023] The valve action amplitude smoothing term is proportional to the square of the valve opening difference between two adjacent control cycles, and the positive excitation term is proportional to the percentage of concentrated water consumption to total medium consumption in the current control cycle.

[0024] Preferably, in S4, generating a secure allocation scheme includes:

[0025] Obtain the system's current electrical safety status parameters and preset safety restriction rules;

[0026] The initial allocation strategy is substituted into the preset safety restriction rules for safety verification to determine whether the initial allocation strategy meets electrical safety and operational boundary constraints.

[0027] If the conditions are met, the initial allocation strategy will be directly determined as the safe allocation scheme.

[0028] If the conditions are not met, a safety calibration branch is triggered, reducing the total flow ratio and valve opening of all candidate cooling media in the initial allocation strategy in a linear decay manner until the electrical safety and operational boundary constraints are met, and the corrected strategy is determined as the safety allocation scheme.

[0029] Preferably, in S5, the calculation of the a priori scaling rate and transient thermal resistance increment of the heated surface corresponding to the current operating condition through the material physical degradation mechanism model includes:

[0030] The water ion concentration, near-wall water temperature, and heated surface pipe wall temperature of the current cooling medium are obtained, and the critical supersaturation of the near-wall fluid is calculated based on the solubility product constant.

[0031] The crystallization deposition rate is calculated based on the critical supersaturation, and the fluid erosion rate is calculated based on the current wall shear stress of the fluid and the existing scale thickness.

[0032] The a priori scaling rate of the heated surface is calculated based on the difference between the crystallization deposition rate and the fluid erosion rate, combined with the preset scale density.

[0033] Based on the prior scaling rate of the heated surface and the thermal conductivity of the scale layer corresponding to the current cooling medium ratio, the transient thermal resistance increment within the current control cycle is calculated.

[0034] Preferably, in S6, the prior evolution correction of the heat transfer coefficient in the heat exchange prediction model includes:

[0035] The calculated transient thermal resistance increment is added sequentially to the historical cumulative thermal resistance of the previous control cycle to obtain the total cumulative thermal resistance of the current heated surface.

[0036] Based on the total cumulative thermal resistance, the initial thermal conductivity coefficient in the heat exchange prediction model is calculated by reciprocal decay to update the prior effective thermal conductivity coefficient under the current operating condition.

[0037] The prior effective thermal conductivity coefficient is injected into the physical mechanism constraint term of the heat exchange prediction model, and the energy conservation equation coefficient matrix in the heat exchange prediction model is updated.

[0038] Preferably, in S6, narrowing the search space of feasible solutions for the heuristic optimization algorithm in the next control cycle includes:

[0039] Based on the prior effective thermal conductivity coefficient and the preset target cooling range of flue gas, the extreme flow range of each candidate cooling medium that meets the lower limit of heat exchange efficiency is calculated.

[0040] By combining the valve opening status of the current control cycle with the preset maximum single-step action rate of the actuator, the dynamic reachable physical range of the valve opening of each cooling medium in the next control cycle is determined.

[0041] The intersection of the valve opening range mapped from the extreme flow range and the dynamically reachable physical range is calculated to eliminate invalid solution regions that do not meet the thermodynamic efficiency and physical action constraints.

[0042] The contracted space formed after finding the intersection is used as the population initialization search boundary and solution space constraint condition of the heuristic optimization algorithm in the next control cycle.

[0043] The present invention has the following beneficial effects:

[0044] 1. In this invention, the prior scaling rate and thermal resistance increment are calculated by the material physical degradation mechanism model, and the heat exchange prediction model is modified by prior evolution. This overcomes the serious lag of traditional ex-post temperature feedback regulation and significantly improves the accuracy of forward-looking adaptive cooling control under unsteady conditions.

[0045] 2. In this invention, by introducing valve action smoothing and concentrate reuse incentive terms into the optimization target, and combining them with electrical load safety margin for strategy calibration, the invention effectively suppresses actuator wear and electrical overload while improving the high hardness concentrate reuse ratio and energy-saving benefits, thus ensuring the safety of system operation.

[0046] 3. In this invention, by pruning the intersection of the extreme flow range and the valve's dynamically reachable range, invalid solution regions are eliminated and the search space for feasible solutions is reduced, which significantly reduces the computational overhead and iteration count of the heuristic optimization algorithm and significantly improves the real-time response speed of the system's online control. Attached Figure Description

[0047] Figure 1 This is a flowchart of a method for dynamically selecting flue gas cooling media based on water source status monitoring, as proposed in this invention.

[0048] Figure 2 A flowchart for generating a secure allocation scheme provided in an embodiment of the present invention;

[0049] Figure 3 A flowchart for calculating the a priori scaling rate and transient thermal resistance increment of the heated surface, provided in an embodiment of the present invention;

[0050] Figure 4 A flowchart illustrating the prior evolution correction of the thermal conductivity coefficient provided for an embodiment. Detailed Implementation

[0051] The technical solutions in 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.

[0052] Example 1

[0053] In a first embodiment of the present invention, the present invention provides a method for dynamically selecting flue gas cooling media based on water source status monitoring, such as... Figure 1 As shown, it includes the following steps:

[0054] S1. Collect and align water source status data, flue gas operating condition data and environmental data of the target area to obtain multi-source time series data;

[0055] Preferably, in S1, obtaining multi-source time-series data includes:

[0056] Anomaly detection is performed on the collected water source status data, flue gas operating condition data, and environmental data. Outliers are removed to obtain the cleaned operating data.

[0057] The cleaned running data is time-series aligned, and missing items generated during the time-series alignment process are filled in to obtain the multi-source time-series data.

[0058] Specifically, in step S1, the sensor network collects real-time data on water source status, flue gas operating conditions, and environmental data in the target area, with a sampling frequency set to 1 Hz. The water source status data includes calcium ion concentration, magnesium ion concentration, pH value, conductivity, and water temperature; the flue gas operating conditions data includes flue gas inlet flow rate, flue gas inlet temperature, flue gas inlet dust content, and system pressure; and the environmental data includes ambient temperature and ambient humidity.

[0059] In the specific implementation process of obtaining multi-source time series data, the following processing is performed:

[0060] First, outlier removal is performed. A sliding window with a window length of 60 sampling points is used to detect outliers in the data of each channel, and outliers exceeding 3 times the standard deviation are removed to obtain cleaned running data.

[0061] Subsequently, time series alignment and data completion were performed. A unified time reference grid with a time interval of 1 second was established. Missing items were filled in on the cleaned operational data using a linear interpolation algorithm. The multi-source sensor data were aligned to the same time axis to construct a multi-source time series data matrix. ,in, Represents a multi-source time-series data matrix; This represents the total number of sampling time points within a single control cycle, with a value of 3600, corresponding to a 1-hour control window. This represents the multi-source feature dimension, with a value of 12, corresponding to the 12 water source, flue gas, and environmental feature parameters collected above.

[0062] S2. Input the multi-source time-series data into the heat exchange prediction model and output the predicted heat exchange efficiency and predicted equipment loss probability of each candidate cooling medium.

[0063] Preferably, in S2, the heat exchange prediction model consists of a backbone model and a heat exchange loss function, wherein the backbone model is an LSTM model, and the heat exchange loss function includes a data error term and a physical constraint regularization term.

[0064] The data error term is used to characterize the mean square error between the predicted heat exchange efficiency and the actual heat exchange efficiency of each candidate cooling medium.

[0065] The physical constraint regularization term is constructed based on the law of conservation of energy in fluid thermodynamics and is used to characterize the degree of deviation of the predicted heat exchange efficiency from the preset heat balance equation between flue gas and cooling medium.

[0066] Specifically, in step S2, the multi-source time-series data matrix is ​​input into the heat exchange prediction model, which outputs the predicted heat exchange efficiency and predicted equipment failure probability of each candidate cooling medium. The heat exchange prediction model consists of a backbone model and a heat exchange loss function. The backbone model uses a two-layer long short-term memory network, i.e., an LSTM network, with a hidden layer dimension of 64. The backbone model extracts features from the multi-source time-series data, and the network ends in a fully connected output layer, which maps and outputs the predicted heat exchange efficiency and predicted equipment failure probability of each candidate cooling medium. The total number of candidate cooling media is 3, corresponding to fresh water, circulating water discharge, and reverse osmosis concentrate, respectively.

[0067] The heat exchange loss function is constructed by weighting the data error term and the physical constraint regularization term. The overall calculation formula is as follows: ;

[0068] In the formula, This represents the total loss value of the heat exchange prediction model; Indicates the data error term; Represents physical constraint regularization terms; This represents the physical constraint weight coefficient, which is 0.15 in actual implementation.

[0069] The data error term characterizes the mean square error between the predicted heat exchange efficiency and the actual heat exchange efficiency of each candidate cooling medium. The calculation formula is as follows:

[0070] ;

[0071] In the formula, This represents the total number of candidate cooling media, with a value of 3. Indicates the index number of the candidate cooling medium; Indicates the first Predicted heat exchange efficiency of several candidate cooling media; Indicates the first Measured values ​​of the actual heat exchange efficiency of candidate cooling media.

[0072] The physical constraint regularization term is constructed based on the law of conservation of energy in fluid thermodynamics. It is used to characterize the degree of deviation of the predicted heat exchange efficiency from the preset heat balance equation between flue gas and cooling medium. The calculation formula is as follows:

[0073] ;

[0074] In the formula, Indicates the flue gas mass flow rate; This indicates the specific heat capacity of flue gas at constant pressure. Indicates the flue gas inlet temperature; This indicates the target flue gas outlet temperature.

[0075] S3. Input the predicted heat exchange efficiency and the predicted equipment loss probability into the heuristic optimization algorithm. The heuristic optimization algorithm performs optimization calculations under the preset global energy consumption, cooling target and equipment loss constraints, and outputs the initial allocation strategy of the cooling medium.

[0076] Preferably, in S3, the heuristic optimization algorithm is one of genetic algorithm, particle swarm optimization, simulated annealing algorithm, or ant colony optimization algorithm, and the initial formulation strategy of the output cooling medium includes:

[0077] Based on the preset global energy consumption, cooling target, and equipment loss constraints, a comprehensive optimization objective function is constructed.

[0078] Initialize the candidate solution space, which includes the type and flow rate of each candidate cooling medium.

[0079] The heuristic optimization algorithm iteratively optimizes the comprehensive optimization objective function within the candidate solution space to determine the optimal target solution that makes the comprehensive optimization objective function reach its optimum, and uses the target solution as the initial allocation strategy for the cooling medium.

[0080] Preferably, the comprehensive optimization objective function includes a penalty term corresponding to the equipment loss constraint. The penalty term includes a valve action amplitude smoothing term to suppress frequent switching of each candidate cooling medium, and a positive incentive term to increase the proportion of concentrated water reuse in the cooling medium.

[0081] The valve action amplitude smoothing term is proportional to the square of the valve opening difference between two adjacent control cycles, and the positive excitation term is proportional to the percentage of concentrated water consumption to total medium consumption in the current control cycle.

[0082] Specifically, in step S3, the predicted heat exchange efficiency and predicted equipment loss probability of each candidate cooling medium obtained in step S2 are input into the heuristic optimization algorithm, and the initial allocation strategy of the cooling medium is output. This embodiment uses particle swarm optimization (PSO) as the specific implementation of the heuristic optimization algorithm. The population size of the PSO algorithm is set to 50, the maximum number of iterations is set to 100, the inertia weight is set to 0.7, and both the individual learning factor and the social learning factor are set to 1.5.

[0083] Based on preset global energy consumption, cooling targets, and equipment loss constraints, a comprehensive optimization objective function is constructed. The calculation formula is as follows:

[0084] ;

[0085] In the formula, This represents the value of the comprehensive optimization objective function; the smaller the value, the better the overall effectiveness of the strategy. Indicates the system's global energy consumption. This indicates the preset system power limit. This indicates the preset lower limit of system power. This represents the global energy consumption weighting coefficient, which is set to 0.4 in actual implementation. This indicates the preset target heat load for flue gas cooling. Indicates the first Predicted heat exchange efficiency of candidate cooling media The normalized reference represents the square of the maximum deviation of the heat load. This represents the weighting coefficient for the deviation from the cooling target, which is set to 0.5 in actual implementation. This represents the smoothness coefficient of valve actuation amplitude, which is taken as 0.2 in actual implementation. Indicates the current control cycle Inner The percentage of control valve opening corresponding to each candidate cooling medium; Indicates the previous control cycle Inner The percentage of control valve opening corresponding to each candidate cooling medium. The maximum normalized reference for the valve's actuation range; This represents the positive incentive coefficient for concentrate reuse, which is taken as 0.3 in actual implementation. Indicates the current control cycle The volume of concentrated water consumed. Indicates the current control cycle The total volume consumed by all candidate cooling media. The valve actuation amplitude smoothing term in the formula. Proportional to the square of the valve opening difference between two adjacent control cycles, this is used to increase the penalty cost for frequent and large valve adjustments during the optimization process, thus suppressing mechanical wear of the actuator; positive incentive term. It is proportional to the percentage of concentrate consumption in the total media consumption during the current control cycle. By existing as a subtraction term in the objective function, it provides positive guidance for strategies to increase the concentrate reuse ratio.

[0086] Initialize the candidate solution space for the particle swarm optimization algorithm. Each particle represents a combination of the type and flow rate of a candidate cooling medium, expressed as a three-dimensional vector. The particle position element's value range is limited to 0 to 100%, corresponding to the valve's fully closed to fully open state. The actual flow rate of the candidate cooling medium is determined by the valve opening. The value is determined by multiplying the value by the maximum rated flow rate of the corresponding pipeline. Fifty initial particle position vectors and corresponding initial velocity vectors are randomly generated in the particle swarm space to form the initial candidate solution space.

[0087] Within the initial candidate solution space, the comprehensive optimization objective function value corresponding to each particle's position is calculated sequentially, and this value is used as the particle's current fitness. The fitness of each particle is compared, and the individual's historical best position and the global best position are updated. The particle's velocity and position are updated based on its current velocity, individual best position, and global best position. In practical implementation, when the algorithm has iterated 100 times or the objective function value corresponding to the global best position has changed by less than 5 times consecutively... Stop iterating when the time is reached. Output the final global optimal position vector, and use the valve opening degree and corresponding flow ratio contained in it as the initial distribution strategy of the cooling medium.

[0088] S4. Extract the electrical operating boundary data of the target execution unit, evaluate the expected electrical load corresponding to the initial allocation strategy, and selectively adjust the initial allocation strategy by linear attenuation based on the comparison result of the expected electrical load and the preset safety margin to generate a safe allocation scheme.

[0089] Preferably, in S4, generating a secure allocation scheme includes:

[0090] Obtain the system's current electrical safety status parameters and preset safety restriction rules;

[0091] The initial allocation strategy is substituted into the preset safety restriction rules for safety verification to determine whether the initial allocation strategy meets electrical safety and operational boundary constraints.

[0092] If the conditions are met, the initial allocation strategy will be directly determined as the safe allocation scheme.

[0093] If the conditions are not met, a safety calibration branch is triggered, reducing the total flow ratio and valve opening of all candidate cooling media in the initial allocation strategy in a linear decay manner until the electrical safety and operational boundary constraints are met, and the corrected strategy is determined as the safety allocation scheme.

[0094] Specifically, in step S4, the current electrical safety status parameters of the system are obtained in real time from the field power distribution monitoring system, including the real-time operating current of the drive motors of each cooling medium delivery pump group, the bus voltage, and the inverter operating temperature. Simultaneously, preset safety limit rules are loaded, including the maximum rated allowable current limit of the bus. Electrical load safety margin factor In actual implementation, the value is taken as 0.90, which corresponds to leaving a 10% safety buffer space for inrush current.

[0095] Substituting the initial allocation strategy output in step S3, i.e., the opening degree of each cooling medium control valve, into the pump group electrical power mapping model, the expected total system electrical current when the initial allocation strategy is executed is calculated. The calculation formula is as follows:

[0096] ;

[0097] In the formula, This represents the expected total electrical current corresponding to the initial allocation strategy; Indicates the total number of candidate cooling media; Indicates the index number of the candidate cooling medium; Indicates the first The no-load foundation current of a cooling medium delivery pump unit; Indicates the first The opening degree-current conversion coefficient of the cooling medium delivery pump set is taken as 0.45 in actual implementation; Indicating the first in the initial allocation strategy The percentage of valve opening controlled by the cooling medium. Verify the expected total electrical current. Whether electrical safety and operational boundary constraints are met, i.e. .

[0098] If the calculated expected total electrical current If the above constraints are met, it indicates that the initial dispatching strategy is within the electrical safety load range, and the initial dispatching strategy is directly determined as the safe dispatching scheme. If the calculated expected total electrical current... If the above constraints are exceeded, a safety calibration branch is triggered. The linear decay coefficient is calculated using a linear decay method based on the penalty factor. And proportionally reduce the valve opening and total flow ratio of all candidate cooling media in the same direction, as calculated by the following formula:

[0099] ;

[0100] ;

[0101] In the formula, Represents the linear attenuation coefficient. Indicates the calibration correction of the first The percentage of valve opening controlled by the candidate cooling medium; Indicating the first in the initial allocation strategy The initial opening percentage of the valve controlled by the cooling medium is determined. After linear decay correction, the corrected valve opening is substituted into the mapping model to calculate the expected current, which is strictly limited to within the safe upper limit.

[0102] S5. Drive the target execution unit to switch the cooling medium according to the safety allocation scheme, and extract the water quality characteristic parameters of the source water in the multi-source time series data. Calculate the a priori scaling rate and transient thermal resistance increment of the heating surface corresponding to the current working condition through the material physical degradation mechanism model.

[0103] Preferably, in S5, the calculation of the a priori scaling rate and transient thermal resistance increment of the heated surface corresponding to the current operating condition through the material physical degradation mechanism model includes:

[0104] The water ion concentration, near-wall water temperature, and heated surface pipe wall temperature of the current cooling medium are obtained, and the critical supersaturation of the near-wall fluid is calculated based on the solubility product constant.

[0105] The crystallization deposition rate is calculated based on the critical supersaturation, and the fluid erosion rate is calculated based on the current wall shear stress of the fluid and the existing scale thickness.

[0106] The a priori scaling rate of the heated surface is calculated based on the difference between the crystallization deposition rate and the fluid erosion rate, combined with the preset scale density.

[0107] Based on the prior scaling rate of the heated surface and the thermal conductivity of the scale layer corresponding to the current cooling medium ratio, the transient thermal resistance increment within the current control cycle is calculated.

[0108] Specifically, in step S5, the current water ion concentration of the cooling medium is acquired in real time, including calcium ion concentration and carbonate ion concentration. Simultaneously, the near-wall water temperature and the heating surface tube wall temperature are acquired. Based on the anomalous solubility characteristics, the solubility product constant is retrieved according to the heating surface tube wall temperature. The equilibrium saturation concentration of the fluid near the wall of the heated surface is calculated based on the solubility product constant. And combined with the actual near-wall water ion concentration Calculate the critical supersaturation .

[0109] Based on critical supersaturation Calculate the crystallization deposition rate The calculation formula is as follows: In the formula, Indicates the crystallization deposition rate; represents the mass transfer and reaction rate coefficients; n represents the crystallization reaction order. The fluid erosion rate is calculated by considering the wall shear stress of the current cooling medium and the existing scale thickness. The calculation formula is as follows: In the formula, Indicates the fluid erosion rate; Indicates the mechanical erosion coefficient; This represents the shear stress at the fluid wall. This indicates the current accumulated scale thickness on the heated surface.

[0110] Based on crystallization deposition rate With fluid erosion rate The difference, combined with the preset scale density, is used to calculate the a priori scaling rate of the heated surface. The calculation formula is as follows: In the formula, Indicates the a priori rate of scaling on the heated surface; This indicates the preset apparent density of the scale layer.

[0111] Based on the mixing volume ratio of the three candidate cooling media in the safety preparation plan, the thermal conductivity of the scale layer formed under the current ratio is calculated using a weighted average. The calculation formula is as follows: In the formula, This indicates the thermal conductivity of the scale layer corresponding to the mixing ratio of the cooling media; Indicates the first The volume percentage of each candidate cooling medium in the safe formulation; Indicates the first The characteristic thermal conductivity of a single cooling medium when it forms scale. Based on the a priori scaling rate of the heated surface. thermal conductivity of scale Calculate the transient thermal resistance increment within the current control cycle. The calculation formula is as follows: In the formula, This represents the transient thermal resistance increment within the current control cycle; Indicates the a priori rate of scaling on the heated surface; Indicates the duration of a single control cycle.

[0112] S6. Based on the prior scaling rate of the heated surface and the transient thermal resistance increment, the thermal conductivity coefficient in the heat exchange prediction model is modified by prior evolution, and the search space of the feasible solution of the heuristic optimization algorithm in the next control cycle is reduced.

[0113] Preferably, in S6, the prior evolution correction of the heat transfer coefficient in the heat exchange prediction model includes:

[0114] The calculated transient thermal resistance increment is added sequentially to the historical cumulative thermal resistance of the previous control cycle to obtain the total cumulative thermal resistance of the current heated surface.

[0115] Based on the total cumulative thermal resistance, the initial thermal conductivity coefficient in the heat exchange prediction model is calculated by reciprocal decay to update the prior effective thermal conductivity coefficient under the current operating condition.

[0116] The prior effective thermal conductivity coefficient is injected into the physical mechanism constraint term of the heat exchange prediction model, and the energy conservation equation coefficient matrix in the heat exchange prediction model is updated.

[0117] Specifically, in step S6, the transient thermal resistance increment calculated in step S5 is obtained. Simultaneously, the historical cumulative thermal resistance of the previous control cycle k-1 is extracted from the system's historical database. The current control cycle is calculated by summing the transient thermal resistance increment with the historical cumulative thermal resistance over time. Total cumulative thermal resistance of the inner heating surface In the formula, This represents the total cumulative thermal resistance of the heated surface within the current control cycle k, which is the additional amount of heat transfer resistance caused by surface fouling from the time the heated surface is put into operation until the end of the current control cycle.

[0118] Based on the principle of superposition of thermal conduction impedance, the prior effective thermal conduction coefficient in the heat exchange prediction model is updated by calculating the reciprocal decay of the initial thermal conduction coefficient according to the total cumulative thermal resistance, thus obtaining the a priori effective thermal conduction coefficient under the current operating condition. The calculation formula is as follows:

[0119] ;

[0120] In the formula, Indicates the current control cycle The prior effective thermal conductivity coefficient obtained from the next update; This represents the initial thermal conductivity of the heated surface in a clean state; This represents the inherent thermal resistance of the heated surface tube wall and fluid boundary layer under clean conditions.

[0121] The calculated prior effective thermal conductivity coefficient The physical constraint regularization term of the pre-set heat exchange prediction model in step S2 is injected to dynamically update the coefficient matrix of the energy conservation equation. In the heat exchange prediction model, the theoretical heat exchange capacity equation of the heated surface is expressed as:

[0122] ;

[0123] In the formula, Indicates the first The theoretical limit of heat exchange capacity of a cooling medium under the current scaling condition; Indicates the a priori effective thermal conductivity; This represents the total heat transfer area of ​​the heat exchanger's heating surface. This represents the logarithmic mean temperature difference. The theoretical limit heat exchange capacity is injected as a coefficient adjustment factor into the physical constraint regularization term of the loss function, resulting in the updated physical constraint energy conservation coefficient matrix. Expressed as:

[0124] ;

[0125] In the formula, Indicates the current control cycle The updated coefficient matrix of the energy conservation equation Indicates the flue gas mass flow rate; This represents the specific heat capacity of flue gas at constant pressure. This is achieved by updating the coefficient matrix. By replacing the static energy conservation coefficient in the original heat exchange prediction model, the heat exchange prediction model can detect the thermal efficiency degradation caused by fouling in advance during the forward prediction process, thus realizing the prior evolution correction of the prediction model.

[0126] Preferably, in S6, narrowing the search space of feasible solutions for the heuristic optimization algorithm in the next control cycle includes:

[0127] Based on the prior effective thermal conductivity coefficient and the preset target cooling range of flue gas, the extreme flow range of each candidate cooling medium that meets the lower limit of heat exchange efficiency is calculated.

[0128] By combining the valve opening status of the current control cycle with the preset maximum single-step action rate of the actuator, the dynamic reachable physical range of the valve opening of each cooling medium in the next control cycle is determined.

[0129] The intersection of the valve opening range mapped from the extreme flow range and the dynamically reachable physical range is calculated to eliminate invalid solution regions that do not meet the thermodynamic efficiency and physical action constraints.

[0130] The contracted space formed after finding the intersection is used as the initial search boundary and solution space constraint condition of the heuristic optimization algorithm in the next control cycle.

[0131] Specifically, in step S6, the preset target temperature reduction of the flue gas is obtained from the system settings. Combined with flue gas mass flow rate flue gas specific heat capacity at constant pressure Based on the target temperature reduction of the flue gas, calculate the minimum theoretical heat exchange required for the next control cycle. The calculation formula is as follows:

[0132] ;

[0133] In the formula, Indicates the minimum theoretical heat exchange rate; This represents the thermal equilibrium safety factor. Combined with the previously updated a priori effective thermal conductivity... Based on the energy conservation equation of heat transfer, the first... The extreme flow rate limit of candidate cooling media when satisfying the lower limit of heat exchange efficiency The calculation formula is as follows:

[0134] ;

[0135] In the formula, This represents the extreme flow rate limit that satisfies the lower limit of heat exchange efficiency. Indicates the minimum theoretical heat exchange rate; Indicates the first Specific heat capacity of a cooling medium; Indicates the first The maximum permissible outlet water temperature after the cooling medium absorbs heat; Indicates the first The initial inlet water temperature of the cooling medium. The calculated lower limit of the extreme flow rate. and the rated maximum flow rate of the pipeline Mapped to the lower limit of the corresponding control valve opening. and upper limit The extreme flow opening interval is thus constructed.

[0136] Extract the actual opening status of each candidate cooling medium control valve within the current control cycle k. Based on the preset maximum single-step action rate and control cycle duration of the actuator, determine the physical action limit of the actuator in the next control cycle k+1. Calculate the maximum physical opening variation that the valve controlling the m-th candidate cooling medium can achieve in the next control cycle. In the formula, This indicates the maximum percentage change in valve opening allowed in a single step; This indicates the maximum rate of valve operation controlled by the actuator; This indicates the control adjustment cycle of the algorithm. In this embodiment, the maximum allowable change in valve opening per step is 15%. This is based on the current valve opening state. The lower limit of the dynamically reachable physical range is determined by the maximum opening change. and upper limit The calculation formula is as follows:

[0137] ;

[0138] ;

[0139] In practice, we assume the current opening degree is 60%. This represents the maximum change in opening, with a value of 15%. The calculated dynamic reachable physical range is 45% to 75%.

[0140] The intersection of the extreme flow rate range and the dynamically reachable physical range is calculated to eliminate invalid solution regions that do not meet the thermodynamic heat transfer efficiency constraints or the physical action rate constraints of the actuator. The calculation formula is as follows:

[0141] ;

[0142] ;

[0143] ;

[0144] In the formula, This represents the contracted feasible solution space formed after finding the intersection; Represents the first in the contracted space The final lower bound for the search of the opening degree of the candidate cooling medium; Represents the first in the contracted space The final upper bound for the search of the opening degree of the candidate cooling medium.

[0145] The contracted space formed after finding the intersection In the heuristic optimization algorithm injected into the next control cycle k+1, it serves as the boundary constraint condition for initializing the population or candidate solution set. When starting optimization in the next control cycle k+1, the decision variable vectors representing the valve openings of each medium in each candidate solution are uniformly and randomly sampled within the corresponding contraction space boundary. The initialization formula for the decision variables is as follows:

[0146] ;

[0147] In the formula, Indicates the first The candidate solution is initialized at the _th ... The numerical values ​​of the opening degree decision variables corresponding to the type of cooling medium; This represents a random sampling function that follows a uniform distribution between the lower and upper bounds.

[0148] Example 2

[0149] In the flue gas treatment system of the desulfurization reaction tower, high-temperature flue gas enters from the top or bottom of the tower. A high-speed rotating atomizer atomizes the cooling medium and limestone slurry into micron-sized droplets for spray evaporation and cooling, providing the optimal temperature range for subsequent desulfurization reactions. This embodiment uses the method of the present invention to achieve dynamic intelligent allocation of the cooling medium. The specific execution process is as follows:

[0150] Step S1: The system collects operating data in real time at a sampling frequency of 1 Hz through a sensor network distributed on the desulfurization reaction tower and water supply pipeline.

[0151] Water source status data: including calcium ion concentration, suspended solids content, pH value, conductivity and water temperature of fresh water, process recycled water and limestone slurry supernatant;

[0152] Flue gas operating data: including flue gas flow rate at the inlet of the reaction tower (set to 400,000 cubic meters per hour), flue gas temperature at the inlet (set to 180 degrees Celsius), dust content in the inlet flue gas, and micro-negative pressure value inside the tower;

[0153] Environmental data: including ambient temperature and ambient humidity.

[0154] After data acquisition, outliers of three standard deviations were detected and removed using a sliding window containing 60 sampling points to eliminate sensor noise. Subsequently, a unified time grid with 1-second intervals was established, and missing items were filled in using a linear interpolation algorithm to construct a unified time-series aligned multi-source time-series data matrix.

[0155] Step S2: Input the multi-source time-series data matrix into a pre-trained two-layer long short-term memory (LSTM) network. After the backbone network extracts the multi-source features, it outputs the predicted heat exchange efficiency under atomization spray conditions for each of the three candidate cooling media: fresh water, process recycled water, and limestone slurry supernatant, as well as the predicted scaling and clogging probability of the rotating atomizer spray disc and pipeline.

[0156] During the prediction process, a physical constraint regularization term based on the law of conservation of energy in fluid thermodynamics is injected into the loss function to constrain the predicted heat exchange efficiency with the theoretical thermal envelope calculated based on flue gas flow rate, flue gas specific heat capacity and internal and external temperature difference. This prevents the pure data-driven model from generating non-physical prediction biases when flue gas conditions fluctuate drastically.

[0157] Step S3: Using the target flue gas temperature at the reaction tower outlet (set to 130 degrees Celsius) as the control objective, construct a comprehensive optimization objective function. The objective function integrates the system's global energy consumption (power of each feedwater pump and atomizer motor), a cooling target deviation term, a smoothing term to suppress frequent adjustments of the atomizer frequency converter and control valves, and a positive excitation term to increase the proportion of industrial wastewater reuse. All sub-terms are dimensionless and mapped to the same order of magnitude using a maximum-minimum normalization method.

[0158] A heuristic optimization algorithm is employed to iteratively optimize within the candidate solution space. By evaluating the overall fitness of different media flow mixing ratios and atomizer rotation speed combinations, the algorithm converges after multiple iterations, determining the solution with the minimum overall optimization objective function value, and using this solution as the initial cooling media mixing strategy for the current control cycle.

[0159] Step S4: Obtain the current operating current, bus voltage, and module temperature of the rotary atomizer drive motor and the water pump inverter. Substitute the initial allocation strategy output from S3 into the electrical load mapping rule between the pump group and the atomizing motor, and calculate the expected total electrical current when executing this strategy.

[0160] The expected total current is compared and verified with the preset maximum allowable current limit of the bus and the 90% safety margin factor:

[0161] If the expected total current is lower than the upper limit of the safety margin, it indicates that the electrical load of the system is safe, and the initial dispatch strategy is directly determined as the safe dispatch scheme.

[0162] If the expected total current exceeds the safety margin limit, a safety calibration branch is triggered. The algorithm calculates a linear attenuation coefficient that makes the current just meet the safety margin, and proportionally reduces the opening of all candidate medium water supply valves and the rotation speed of the rotary atomizer according to this coefficient. The corrected parameters are then determined as the safety allocation scheme and sent to the actuator.

[0163] Step S5: While the atomizer is running under the safety control scheme, extract the current water ion concentration and the near-wall temperature of the atomizer spray plate / reaction tower heating surface.

[0164] Based on the anomalous solubility characteristics of calcium carbonate and calcium sulfate at high temperatures, the critical supersaturation of the fluid near the wall is calculated. The deposition rate of crystals on the spray plate and pipe wall is calculated based on the critical supersaturation. Simultaneously, the fluid shear erosion rate is calculated by combining the fluid wall shear stress generated by high-speed atomization and the existing scale thickness.

[0165] The a priori scaling rate of the heated surface is calculated by subtracting the deposition rate from the erosion rate and combining this with the preset apparent scale density. Then, based on the weighted thermal conductivity of the scale under the current volume ratio of the three media, the transient thermal resistance increment caused by micro-scaling within the current control cycle is calculated.

[0166] Step S6: The transient thermal resistance increment calculated in S5 is added sequentially with the historical cumulative thermal resistance to obtain the current total cumulative thermal resistance of the heated surface. The initial thermal conductivity coefficient under clean conditions is reciprocally decayed using the principle of thermal resistance superposition to update the prior effective thermal conductivity coefficient. The updated coefficient is then injected into the energy conservation coefficient matrix in the loss function of the heat exchange prediction model in S2, enabling the prediction model to automatically detect the decay of the heat transfer efficiency of the heated surface during the next cycle's forward calculation.

[0167] Simultaneously, based on the prior effective heat transfer coefficient and the target cooling range, the extreme water supply required to meet the minimum cooling heat load is calculated, and the extreme valve opening range is mapped to obtain the extreme valve opening range. Combining the current valve opening and the maximum single-step action rate of the rotary atomizer inverter, the dynamic reachable physical range in the next control cycle is calculated.

[0168] The intersection of the extreme value opening interval and the dynamically reachable physical interval is calculated, and invalid solution regions that do not meet the cooling capability and exceed the speed limit capability are eliminated. The significantly reduced shrinkage space after the intersection is used as the boundary constraint for the initialization of the heuristic optimization algorithm population in the next control cycle, realizing efficient pruning of the algorithm's search range and ensuring that the system converges and responds quickly within the control cycle.

[0169] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for dynamically selecting flue gas cooling media based on water source status monitoring, characterized in that, include: S1. Collect and align water source status data, flue gas operating condition data and environmental data of the target area to obtain multi-source time series data; S2. Input the multi-source time-series data into the heat exchange prediction model and output the predicted heat exchange efficiency and predicted equipment loss probability of each candidate cooling medium. S3. Input the predicted heat exchange efficiency and the predicted equipment loss probability into the heuristic optimization algorithm. The heuristic optimization algorithm performs optimization calculations under the preset global energy consumption, cooling target and equipment loss constraints, and outputs the initial allocation strategy of the cooling medium. S4. Extract the electrical operating boundary data of the target execution unit, evaluate the expected electrical load corresponding to the initial allocation strategy, and selectively adjust the initial allocation strategy by linear attenuation based on the comparison result of the expected electrical load and the preset safety margin to generate a safe allocation scheme. S5. Drive the target execution unit to switch the cooling medium according to the safety allocation scheme, and extract the water quality characteristic parameters of the source water in the multi-source time series data. Calculate the a priori scaling rate and transient thermal resistance increment of the heating surface corresponding to the current working condition through the material physical degradation mechanism model. S6. Based on the prior scaling rate of the heated surface and the transient thermal resistance increment, the thermal conductivity coefficient in the heat exchange prediction model is modified by prior evolution, and the search space of the feasible solution of the heuristic optimization algorithm in the next control cycle is reduced.

2. The method for dynamic selection of flue gas cooling medium based on water source status monitoring according to claim 1, characterized in that, In S1, obtaining multi-source time-series data includes: Anomaly detection is performed on the collected water source status data, flue gas operating condition data, and environmental data. Outliers are removed to obtain the cleaned operating data. The cleaned running data is time-series aligned, and missing items generated during the time-series alignment process are filled in to obtain the multi-source time-series data.

3. The method for dynamic selection of flue gas cooling medium based on water source status monitoring according to claim 1, characterized in that, In S2, the heat exchange prediction model consists of a backbone model and a heat exchange loss function. The backbone model is an LSTM model, and the heat exchange loss function includes a data error term and a physical constraint regularization term. The data error term is used to characterize the mean square error between the predicted heat exchange efficiency and the actual heat exchange efficiency of each candidate cooling medium. The physical constraint regularization term is constructed based on the law of conservation of energy in fluid thermodynamics and is used to characterize the degree of deviation of the predicted heat exchange efficiency from the preset heat balance equation between flue gas and cooling medium.

4. The method for dynamic selection of flue gas cooling medium based on water source status monitoring according to claim 1, characterized in that, In S3, the heuristic optimization algorithm is one of genetic algorithm, particle swarm optimization, simulated annealing algorithm, or ant colony optimization algorithm, and the initial formulation strategy for the output cooling medium includes: Based on the preset global energy consumption, cooling target, and equipment loss constraints, a comprehensive optimization objective function is constructed. Initialize the candidate solution space, which includes the type and flow rate of each candidate cooling medium. The heuristic optimization algorithm iteratively optimizes the comprehensive optimization objective function within the candidate solution space to determine the optimal target solution that makes the comprehensive optimization objective function reach its optimum, and uses the target solution as the initial allocation strategy for the cooling medium.

5. The method for dynamic selection of flue gas cooling medium based on water source status monitoring according to claim 4, characterized in that, The comprehensive optimization objective function includes a penalty term corresponding to the equipment loss constraint. The penalty term includes a valve action amplitude smoothing term to suppress frequent switching of each candidate cooling medium, and a positive incentive term to increase the proportion of concentrated water reuse in the cooling medium. The valve action amplitude smoothing term is proportional to the square of the valve opening difference between two adjacent control cycles, and the positive excitation term is proportional to the percentage of concentrated water consumption to total medium consumption in the current control cycle.

6. The method for dynamic selection of flue gas cooling medium based on water source status monitoring according to claim 1, characterized in that, In S4, the generation of the security allocation scheme includes: Obtain the system's current electrical safety status parameters and preset safety restriction rules; The initial allocation strategy is substituted into the preset safety restriction rules for safety verification to determine whether the initial allocation strategy meets electrical safety and operational boundary constraints. If the conditions are met, the initial allocation strategy will be directly determined as the safe allocation scheme. If the conditions are not met, a safety calibration branch is triggered, reducing the total flow ratio and valve opening of all candidate cooling media in the initial allocation strategy in a linear decay manner until the electrical safety and operational boundary constraints are met, and the corrected strategy is determined as the safety allocation scheme.

7. The method for dynamic selection of flue gas cooling medium based on water source status monitoring according to claim 1, characterized in that, In S5, the calculation of the a priori scaling rate and transient thermal resistance increment of the heated surface corresponding to the current operating condition through the material physical degradation mechanism model includes: The water ion concentration, near-wall water temperature, and heated surface pipe wall temperature of the current cooling medium are obtained, and the critical supersaturation of the near-wall fluid is calculated based on the solubility product constant. The crystallization deposition rate is calculated based on the critical supersaturation, and the fluid erosion rate is calculated based on the current wall shear stress of the fluid and the existing scale thickness. The a priori scaling rate of the heated surface is calculated based on the difference between the crystallization deposition rate and the fluid erosion rate, combined with the preset scale density. Based on the prior scaling rate of the heated surface and the thermal conductivity of the scale layer corresponding to the current cooling medium ratio, the transient thermal resistance increment within the current control cycle is calculated.

8. The method for dynamic selection of flue gas cooling medium based on water source status monitoring according to claim 1, characterized in that, In S6, the prior evolution correction of the heat transfer coefficient in the heat exchange prediction model includes: The calculated transient thermal resistance increment is added sequentially to the historical cumulative thermal resistance of the previous control cycle to obtain the total cumulative thermal resistance of the current heated surface. Based on the total cumulative thermal resistance, the initial thermal conductivity coefficient in the heat exchange prediction model is calculated by reciprocal decay to update the prior effective thermal conductivity coefficient under the current operating condition. The prior effective thermal conductivity coefficient is injected into the physical mechanism constraint term of the heat exchange prediction model, and the energy conservation equation coefficient matrix in the heat exchange prediction model is updated.

9. The method for dynamic selection of flue gas cooling medium based on water source status monitoring according to claim 8, characterized in that, In S6, narrowing the search space of feasible solutions for the heuristic optimization algorithm in the next control cycle includes: Based on the prior effective thermal conductivity coefficient and the preset target cooling range of flue gas, the extreme flow range of each candidate cooling medium that meets the lower limit of heat exchange efficiency is calculated. By combining the valve opening status of the current control cycle with the preset maximum single-step action rate of the actuator, the dynamic reachable physical range of the valve opening of each cooling medium in the next control cycle is determined. The intersection of the valve opening range mapped from the extreme flow range and the dynamically reachable physical range is calculated to eliminate invalid solution regions that do not meet the thermodynamic efficiency and physical action constraints. The contracted space formed after finding the intersection is used as the population initialization search boundary and solution space constraint condition of the heuristic optimization algorithm in the next control cycle.

Images