Method and system for drying lignite by using flue gas waste heat

Abstract

This application provides a method and system for drying semi-coke using waste heat from flue gas, relating to the technical field of semi-coke drying. The method includes: real-time detection of the moisture content of semi-coke at the outlet of the semi-coke dryer; when the moisture content is greater than a preset moisture content threshold, acquiring the drying data for the current batch; calculating a first theoretical moisture content based on a drying kinetic model of the current batch of drying data; generating perturbation data by applying a small disturbance to the drying data of the current batch, and calculating a second theoretical moisture content based on the perturbation data of the drying kinetic model of the drying kinetic model of the second batch of drying data; determining the perturbation difference based on the first theoretical moisture content and the second theoretical moisture content, and determining the perturbation coefficient based on the perturbation difference and the small perturbation; and determining parameter adjustment data based on the perturbation coefficient and a preset optimization objective function. This method, to a certain extent, solves the problem in the prior art where static prediction models cannot adapt to dynamic changes in the drying process conditions, leading to a decrease in control accuracy.

Description

Technical Field

[0001] This application relates to the technical field of drying semi-coke, and in particular to a method and system for drying semi-coke using waste heat from flue gas. Background Technology

[0002] Semi-coke, a clean solid fuel with high fixed carbon and high activity, typically contains high moisture content after production or wet quenching, requiring drying to meet the quality requirements for subsequent storage, transportation, and application. Rotary drum dryers are commonly used for drying semi-coke. Their working principle involves direct contact between hot flue gas and wet semi-coke within a rotating drum, evaporating moisture through heat exchange. During the drying process, the output moisture content is a core indicator of product quality; excessively high moisture content can lead to mold growth during storage and transportation, while excessively low moisture content results in energy waste and material pulverization. Therefore, how to dynamically adjust the dryer's operating parameters based on real-time output moisture content to stably control the product moisture within the target range is a continuously relevant technical issue in this field.

[0003] Chinese patent application CN113048777A, published on June 29, 2021, discloses a dryer control method, device, and computer equipment based on the discharge moisture content. This solution acquires the target moisture value of the material to be dried, formula information, inlet material moisture value, material mass flow rate, moisture content of the air entering the drying zone, moisture content of the exhaust air, and dryer operating parameters; selects a corresponding discharge moisture content prediction model based on the formula information; determines the increase in humidity in the drying zone based on the inlet and outlet air moisture content; inputs the humidity increase, inlet material moisture value, and material mass flow rate into the prediction model to obtain the predicted discharge material moisture value; when the difference between the predicted value and the target value does not meet the requirements, an optimized target value for the operating parameters is obtained based on this difference and the current operating parameters using a gradient descent method, and the dryer is adjusted accordingly. The advantage of this patent is that it achieves closed-loop control based on discharge moisture prediction, reducing reliance on manual experience. However, the optimization and adjustment of this scheme relies entirely on a pre-trained static prediction model. The model parameters are fixed after offline training and cannot be adaptively adjusted according to changes in material characteristics, environmental conditions, or equipment status during real-time production. When the actual operating conditions deviate from the data distribution during model training, the model's prediction accuracy decreases, leading to inaccurate optimization adjustments calculated based on these predictions, and the control effect deteriorates over time. Summary of the Invention

[0004] To address the problem that existing static prediction models cannot adapt to dynamic changes in the drying process conditions, leading to a decrease in control accuracy, this application provides a method and system for drying semi-coke using waste heat from flue gas.

[0005] Firstly, this application provides a method for drying semi-coke using waste heat from flue gas, employing the following technical solution: A method for drying semi-coke using waste heat from flue gas includes: real-time detection of the moisture content of semi-coke at the outlet of a semi-coke dryer; when the moisture content is greater than a preset moisture content threshold, acquiring drying data for the current batch; calculating a first theoretical moisture content based on a drying kinetic model of the drying data of the current batch; generating perturbed data by applying a small disturbance to the drying data of the current batch, and calculating a second theoretical moisture content based on the perturbed data of the drying kinetic model of the drying kinetic model of the perturbed data of the current batch; determining a perturbed difference based on the first theoretical moisture content and the second theoretical moisture content, and determining a perturbed coefficient based on the perturbed difference and the small perturbed; and determining parameter adjustment data based on the perturbed coefficient and a preset optimization objective function.

[0006] This application achieves closed-loop monitoring of the drying process by real-time detection of the moisture content of the semi-coke at the discharge port, enabling timely perception of product quality. When the moisture content exceeds a preset threshold, a control mechanism is triggered. By constructing a drying kinetic model to calculate the theoretical moisture content, this application combines physical mechanisms with real-time data, providing a more accurate reflection of the material's drying behavior under specific operating conditions compared to relying solely on experience or static models. Furthermore, this application calculates a second theoretical moisture content by slightly perturbing the drying data of the current batch. Based on the difference in moisture content before and after the perturbation, a perturbation coefficient is determined, quantifying the sensitivity of each operating parameter to the drying effect and providing scientific gradient information for parameter adjustment. Based on the perturbation coefficient and a preset optimization objective function, parameter adjustment data is determined. This application achieves the function of dynamically calculating optimal operating parameters according to real-time operating conditions, enabling the dryer to adaptively adjust its operating state according to changes in material characteristics and environmental conditions. This effectively alleviates the problem of the deterioration of static model prediction accuracy over time in the prior art, improving the long-term stability and adaptability of the control system.

[0007] Optionally, the calculation of the first theoretical moisture content and the calculation of the second theoretical moisture content are performed using a drying kinetic model constructed based on the thermal-mass coupling theory of porous media. The drying kinetic model includes a correlation function between the effective moisture diffusion coefficient and temperature, gas velocity, and material layer thickness.

[0008] By employing the thermo-mass coupling theory of porous media to construct a drying kinetic model, this application considers the physical mechanism of moisture migration within semi-coke and the coupling effect of heat transfer and mass transfer, enabling the model to more realistically describe the essential laws of the drying process. By introducing a correlation function between the effective moisture diffusion coefficient and temperature, gas velocity, and material layer thickness, this application establishes a quantitative relationship between operating parameters and drying rate, improving the model's predictive accuracy under different operating conditions, making the calculated theoretical moisture content closer to actual drying behavior, and thus enhancing the reliability of parameter adjustments.

[0009] Optionally, the minute perturbation includes a single-parameter perturbation or a combined perturbation. The single-parameter perturbation is applying an increment or decrement to a single parameter, and the combined perturbation is applying perturbations of different magnitudes to multiple parameters simultaneously.

[0010] By supporting both single-parameter perturbation and combined perturbation, this application can flexibly adapt to different optimization needs. Single-parameter perturbation is suitable for analyzing the sensitivity when each parameter acts independently, while combined perturbation can capture the coupling effect between parameters and reveal the nonlinear influence of multi-parameter synergistic changes on drying results, thereby improving the comprehensiveness and applicability of perturbation analysis and making the optimization strategy more in line with the characteristics of multivariate coupling in actual industrial processes.

[0011] Optionally, determining the perturbation coefficient includes calculating the ratio of the perturbation difference to the minute perturbation, and normalizing the calculation result to eliminate dimensional differences.

[0012] By calculating the ratio of the perturbation difference to the minute perturbation and performing normalization, this application transforms parameter changes with different physical dimensions into comparable dimensionless coefficients, eliminating dimensional differences between different types of parameters such as temperature, pressure, and flow rate. This processing method makes the perturbation coefficients of each parameter comparable, avoiding optimization biases caused by different dimensions and improving the objectivity and scientific rigor of parameter adjustment data calculations.

[0013] Optionally, the preset optimization objective function includes an energy consumption objective sub-function, a quality objective sub-function, and a safety objective sub-function. The energy consumption objective sub-function aims to minimize the waste heat consumption of flue gas and the power consumption of the fan. The quality objective sub-function aims to minimize the deviation between the output moisture content and the target moisture content. The safety objective sub-function aims to maximize the safety margin between the surface temperature and the critical oxidation temperature of semi-coke.

[0014] By constructing a multi-objective optimization function encompassing energy consumption, quality, and safety, this application achieves comprehensive control of the drying process. The energy consumption objective sub-function prompts the system to minimize waste heat and electrical energy consumption while meeting quality requirements, thereby improving energy utilization efficiency. The quality objective sub-function ensures that the output moisture content remains stable within the target range, improving product quality consistency. The safety objective sub-function reduces the risk of high-temperature oxidation by maintaining a safety margin between the surface temperature of the semi-coke and the critical oxidation temperature. The synergistic effect of the three sub-functions enables the system to achieve a balance between energy saving, quality, and safety, avoiding the side effects that may result from single-objective optimization and enhancing the practicality and robustness of the control strategy.

[0015] Optionally, the method further includes: calculating the predicted deviation between the first theoretical moisture content and the real-time detected actual moisture content; when the predicted deviation exceeds a first preset threshold and continues for a first preset duration, determining that the drying kinetic model is inaccurate or the moisture content detection sensor is faulty; when the predicted deviation exceeds a second preset threshold and the parameter adjustment data has been adjusted multiple times but the predicted deviation shows no decreasing trend, determining that the actuator response is abnormal; wherein, the second preset threshold is greater than the first preset threshold.

[0016] By continuously monitoring the predicted deviation between theoretical and actual moisture content, this application achieves real-time diagnosis of the system status. When the deviation exceeds a first preset threshold and persists for a certain period, it indicates that the model's predictive ability may be declining or the sensor may be malfunctioning, prompting timely maintenance requests. When the deviation exceeds a larger second preset threshold and adjustments are ineffective, it is determined that the actuator may have a delayed response or failure. This hierarchical diagnostic mechanism enables the system to distinguish between three typical faults: model, sensor, and actuator, improving the accuracy of fault location, facilitating timely and targeted maintenance measures, reducing unplanned downtime, and enhancing the system's reliability and maintainability.

[0017] Optionally, the method further includes: monitoring the adjustment frequency and adjustment range of the parameter adjustment data for each drying parameter; when the adjustment range of the same parameter exceeds a preset range threshold multiple times consecutively and the change corresponding to the disturbance coefficient is less than a preset change threshold, determining that the actuator corresponding to the same parameter has a mechanical jamming or valve blockage fault; when the adjustment direction of multiple parameters shows periodic oscillation and the oscillation amplitude increases, determining that the drying system has a coupled oscillation fault.

[0018] By monitoring the relationship between the frequency and amplitude of parameter adjustments and changes in the disturbance coefficient, this application can identify mechanical faults in the actuator. When an adjustment command is issued but the change in the disturbance coefficient is weak, it indicates that the actuator may have mechanical jamming or valve blockage, failing to effectively respond to control commands. By monitoring the periodic oscillation patterns of multi-parameter adjustments, this application can identify coupled oscillation faults at the system level, indicating the presence of negative interactions between control loops. These fault diagnosis capabilities enable maintenance personnel to detect mechanical wear or control misalignment problems in advance, improving the effectiveness of preventative maintenance and contributing to maintaining long-term stable operation.

[0019] Optionally, the method further includes: calculating the statistical distance between the current batch drying data and the historical drying data under normal operating conditions; when the statistical distance exceeds a preset distance threshold, determining that the characteristics of the feed semi-coke have undergone abnormal changes or that there is a hidden thermal fault in the drying system, triggering an alarm and switching to a conservative control mode.

[0020] By calculating the statistical distance between the current operating conditions and historical normal operating conditions, this application can identify abnormal deviations in data distribution. When the statistical distance exceeds a threshold, it indicates that the physical properties of the feed semi-coke may undergo a sudden change, or that the system has a latent thermal fault that has not yet manifested as an obvious parameter deviation. At this time, an alarm is triggered and the system switches to a conservative control mode. This application can take protective measures at the fault initiation stage to avoid product quality accidents or equipment damage caused by the deterioration of abnormal operating conditions, thereby improving the robustness and risk resistance of the system.

[0021] Optionally, the method further includes: when it is determined that the drying kinetic model is inaccurate or the prediction deviation continues to exceed a third preset threshold, collecting the actual output moisture content after parameter adjustment, calculating the prediction error between the actual output moisture content and the first theoretical moisture content; updating the key parameters in the drying kinetic model online using the recursive least squares method, and recalculating the first theoretical moisture content and the second theoretical moisture content based on the updated model.

[0022] By triggering an online update mechanism when the model becomes inaccurate or the prediction deviation remains large, this application utilizes recursive least squares to correct key model parameters based on the latest collected actual data. This enables the model to adapt to system characteristic drift caused by factors such as material properties and equipment aging. The aforementioned adaptive correction function allows the drying kinetic model to maintain high prediction accuracy during long-term operation, effectively overcoming the shortcomings of static models deteriorating over time in the prior art, extending the effective service life of the control system, and reducing the maintenance costs of manual remodeling or frequent calibration.

[0023] Secondly, this application provides a system for drying semi-coke using waste heat from flue gas, employing the following technical solution: A system for drying semi-coke using waste heat from flue gas includes: a processor and a memory communicatively connected to the processor; the memory is provided with a computer-readable storage medium, on which a computer program is stored; when the processor processes the computer program stored on the computer-readable storage medium, it implements the method as described in the first aspect.

[0024] In summary, this application includes at least one of the following beneficial technical effects: This application combines real-time detection with disturbance analysis, and dynamically calculates parameter adjustment data based on the drying kinetic model and optimized objective function. This enables the drying system to adaptively adjust its operating parameters according to real-time operating conditions, effectively solving the problem that static prediction models cannot adapt to changes in operating conditions, and improving the long-term stability and adaptability of the control system.

[0025] This application constructs a multi-objective optimization function that comprehensively considers energy consumption, quality, and safety, enabling the system to ensure product quality and production safety while saving energy and reducing consumption. This avoids the side effects of single-objective optimization and enhances the practicality and robustness of the control strategy. This application achieves graded diagnosis and identification of model inaccuracies, sensor failures, actuator anomalies, and sudden changes in feed characteristics by monitoring prediction deviations, parameter adjustment responses, and statistical distances. This improves the system's fault early warning capability and maintainability, and helps reduce unplanned downtime. Attached Figure Description

[0026] Figure 1 This is a flowchart of Embodiment 1 of this application. Detailed Implementation

[0027] The following combination Figure 1 This application will be described in further detail.

[0028] Example 1: This example discloses a method for controlling the drying of semi-coke using waste heat from flue gas based on porous media thermo-mass coupling and multi-objective optimization, referring to... Figure 1 The method includes: S1 moisture content detection and data acquisition, S2 calculation of the first theoretical moisture content based on a porous media model, S3 calculation of the second theoretical moisture content using combined perturbation, S4 determination of normalized perturbation coefficients, and S5 determination of multi-objective parameters. The execution process of each step in this embodiment is as follows: S1 Moisture Content Detection and Data Acquisition: A microwave online moisture content detector installed at the outlet of the semi-coke dryer detects the moisture content of the discharged semi-coke in real time with a sampling period of 30 seconds. The microwave online moisture content detector is based on the microwave resonant cavity perturbation principle. When water-containing semi-coke passes through the resonant cavity, the high dielectric constant of water molecules causes a shift in the resonant frequency. The moisture content is calculated by measuring the change in the quality factor, featuring non-contact and fast response. The real-time detected moisture content is compared with a preset moisture content threshold (e.g., a mass fraction of 12%). If the real-time moisture content is greater than the preset threshold, it indicates that the current batch of semi-coke is not sufficiently dried, and the drying parameters need to be adjusted. At this time, the drying data for the current batch is acquired. The current batch is defined as: the material entering the dryer within a complete drying residence time (e.g., 15 minutes) tracing back from the current control cycle as the end point. The drying data for the current batch includes: flue gas inlet temperature, flue gas inlet velocity, drum rotation speed, material layer thickness, ambient humidity, and the initial moisture content of the semi-coke at the current moment. The material layer thickness is obtained by detecting an ultrasonic level sensor installed at the feed end of the drum. This sensor emits ultrasonic pulses and receives echo signals, calculates the material level height based on the time difference, and then converts it into the material layer thickness. The ambient humidity is used to calculate the equilibrium moisture content as a boundary condition parameter of the drying kinetic model. These data are collected by a sensor network distributed at various key locations in the dryer, including thermocouple temperature sensors, Pitot tube flow rate sensors, encoder speed sensors, etc.

[0029] If the real-time moisture content is not greater than the preset moisture content threshold, the current batch of semi-coke is deemed to be dry and the existing operating parameters are maintained without further processing.

[0030] S2 First Theoretical Moisture Content Calculation: Input the current batch drying data obtained in S1 into the drying kinetic model constructed based on the porous medium thermo-mass coupling theory, calculate the theoretical moisture content that the semi-coke should reach after the drying process under the current working conditions, and record it as the first theoretical moisture content.

[0031] The drying kinetics model is constructed based on the thermo-mass coupling theory of porous media, comprehensively considering the coupling effect of heat transfer and moisture migration. The model treats the semi-coke particles as porous media, with internal moisture migrating in the form of liquid diffusion and water vapor. In engineering applications, a lumped parameter model with spatial average along the material thickness direction (radial) is adopted to transform the partial differential equations into ordinary differential equations for solution, balancing the requirements of computational accuracy and real-time performance.

[0032] First theoretical moisture content The calculation formula is as follows:

[0033] in, The initial moisture content (%) of the current batch of semi-coke is obtained by a microwave moisture content meter at the feed inlet. The average residence time (in minutes) of the material inside the drum is calculated from the drum length, inclination angle, and rotational speed. The calculation formula is as follows: ,in For the length of the roller, It is the angle of inclination. For rotational speed, The diameter of the drum. Angle of repose of the material; The space-average drying rate (% / min) is calculated by the drying kinetics model based on the current drying data.

[0034] The space average drying rate The calculation is based on the effective water diffusion coefficient. The correlation function between this coefficient and temperature, gas velocity, and material layer thickness is as follows:

[0035] in, The effective diffusion coefficient of water (m² / s) reflects the ability of water to migrate within a porous medium. The pre-exponential factor (m² / s) was obtained through laboratory isothermal drying experiments. The activation energy for water migration (J / mol) reflects the binding strength between water and the semi-coke matrix. The universal gas constant is taken as 8.314 J / (mol·K); The absolute temperature is (K). The actual velocity of the flue gas (m / s); The reference flow velocity (m / s) is usually taken as the standard flow velocity under design conditions, which is 2 m / s. The velocity influence index ranges from 0.3 to 0.8. The actual layer thickness of the material (m); The reference layer thickness (m) is typically taken as 10% of the roller diameter; The thickness effect index ranges from -0.5 to -0.2. A negative sign indicates that increased thickness will inhibit the diffusion of internal moisture.

[0036] Based on the above effective moisture diffusion coefficient, the drying kinetic differential equation is established as follows:

[0037] in, Instantaneous moisture content (%); Time (s); The coordinates (m) represent the internal position of the material, with the material layer thickness as the coordinate axis, the origin located on the material surface, and the positive direction pointing inwards. The surface evaporation coefficient (1 / s) is related to the flue gas convective heat transfer coefficient. To balance the moisture content (%), it is determined from the ambient humidity using a hygroscopic isotherm.

[0038] The above partial differential equations are solved using the finite difference method or the finite element method. Combined with the drying data of the current batch (flue gas inlet temperature, flow rate, and material layer thickness), the first theoretical moisture content is calculated. .

[0039] S3 involves generating perturbation data and calculating the second theoretical moisture content by performing a small perturbation process on the current batch drying data obtained in S1. This small perturbation refers to applying a small increment or decrease to the original parameter values ​​to analyze the sensitivity of the effect of parameter changes on the drying effect. The amplitude of the small perturbation is set to 1% to 3% of the current parameter values.

[0040] The small perturbations include two modes: single-parameter perturbations and combined perturbations. Single-parameter perturbation mode: An increment or decrement is applied to a single operating parameter while keeping other parameters constant. For example, only the flue gas inlet temperature is affected. Apply or only for flue gas velocity Apply Single-parameter perturbations are suitable for analyzing the sensitivity of each parameter acting independently, as well as for testing the response characteristics of equipment.

[0041] Combined perturbation mode: This mode applies perturbations of different amplitudes to multiple parameters simultaneously to capture the coupling effects between them. For example, when there is a strong coupling between flue gas temperature and flow velocity (increasing both temperature and flow velocity can enhance heat transfer but does not significantly increase the risk of surface hardening), perturbations can be applied simultaneously to the flue gas temperature. Applying pressure to the drum rotation speed Combined perturbations can reveal the nonlinear effects of coordinated changes in multiple parameters on drying results.

[0042] Based on flue gas inlet temperature For example, small perturbations Typically, 2% of the current temperature value is taken. For example, if the current temperature is 300°C, then... Set the temperature to 6°C. Generate disturbance data. Other parameters remain unchanged. It should be noted that the material layer thickness... As a state variable rather than a direct control variable, it remains consistent with the current detection value in disturbance analysis and does not actively apply disturbances.

[0043] The disturbance data were input into the same drying kinetic model, and the same calculation process as S2 was used to calculate the theoretical moisture content under the disturbance condition, which was denoted as the second theoretical moisture content. .

[0044] The S4 disturbance coefficient is determined, and the difference between the first theoretical moisture content and the second theoretical moisture content is calculated and denoted as the disturbance difference. :

[0045] Based on disturbance difference and tiny perturbations (here) This represents the amount of disturbance to a certain controllable parameter, such as... Determine the sensitivity coefficient of this parameter. :

[0046] The sensitivity coefficient The physical meaning of is: the change in moisture content caused by a unit change in control parameter, reflecting the sensitivity of the parameter to the drying effect. The larger the absolute value of , the more significant the effect of this parameter on the moisture content.

[0047] To eliminate differences in the dimensions of different control parameters, the sensitivity coefficient is normalized. The selection criterion for the normalization method is: if the historical data sample is sufficient (sample size...). If the data distribution is approximately normal, Z-score standardization is used; if the data sample is finite or there are obvious boundary constraints, Min-Max normalization is used. The specific calculation formula is as follows: When using Z-score standardization:

[0048] When using Min-Max normalization:

[0049] in, The normalized perturbation coefficient; and These are the mean and standard deviation of the sensitivity coefficient of this parameter in historical data, respectively; and These represent the maximum and minimum values ​​of the sensitivity coefficients for each parameter within the current control cycle. The normalized disturbance coefficients. The value ranges from [0,1]. The larger the value, the more sensitive the parameter is to the moisture content.

[0050] The parameter adjustment data for S5 is determined based on the disturbance coefficients and preset optimization objective function determined in S4, using a multi-objective optimization algorithm. This parameter adjustment data includes: flue gas temperature adjustment, flue gas velocity adjustment, and drum rotation speed adjustment. It should be noted that the material layer thickness is indirectly controlled by adjusting the matching relationship between the feed rate and drum rotation speed. It is not used as an independent optimization variable, but rather by maintaining the ratio of feed rate to rotation speed to keep the material layer thickness within the range of 8% to 15% of the drum diameter.

[0051] Preset optimization objective function By using a weighted summation method, the multi-objective optimization problem is transformed into a single-objective optimization problem for solution:

[0052] in, The energy consumption objective subfunction aims to minimize the waste heat consumption of flue gas and the power consumption of the fan. The quality objective subfunction aims to minimize the deviation between the output moisture content and the target moisture content (e.g., 10%). The safety objective subfunction aims to maximize the safety margin between the surface temperature and the critical oxidation temperature of semi-coke (typically around 350°C). , , The weighting coefficients for each sub-function are set according to production priority.

[0053] The weighting coefficients are dynamically adjusted based on real-time operating conditions. Normal operating conditions (actual moisture content deviation from target <3%, and temperature >50°C): Setting Prioritize product quality; High load conditions (processing capacity > 90% of rated value): Setting Prioritize energy conservation; Critical operating condition (temperature margin <30°C or moisture content deviation >5%): Forced setting Safety should be the top priority.

[0054] The constrained sequential quadratic programming (SQP) method is used to solve the above optimization problem. The optimization variables are the adjustment amounts of each operating parameter. The constraints include: equipment capacity constraints (such as flue gas temperature not exceeding 400°C and drum speed not exceeding 5 rpm), process requirement constraints (such as discharge moisture content not less than 6% to prevent over-drying), and material layer thickness constraints (maintained within a reasonable range by adjusting the feed rate).

[0055] By iterative calculation, we find the objective function. Minimize the optimal adjustment amount This refers to the parameter adjustment data. (The rest of the text appears to be incomplete and requires further context.) This is converted into specific control commands, such as adjusting the opening degree of the flue gas valve. This enables closed-loop optimized control of the drying process.

[0056] In a specific example, Factory A uses the above method to control a rotary drum dryer with a diameter of 3 meters and a length of 12 meters. The control cycle is set to 30 seconds, and the preset moisture content threshold is 12%. When the output moisture content is detected to be 14%, the following data for the current batch is obtained: flue gas inlet temperature 300°C, flow velocity 2.5 m / s, drum rotation speed 3 rpm, material layer thickness 0.25 m, ambient humidity 45%, and initial moisture content 22%.

[0057] Calculation of the first theoretical moisture content based on a porous media model: First, the effective water diffusion coefficient is calculated. ,in m² / s, J / mol, K, m / s, m / s, , m, m, Substituting into the formula, we get... m² / s. Solving the differential equation of drying kinetics, the first theoretical moisture content is obtained. %.

[0058] A small perturbation of +2% (i.e., +6°C) is applied to the flue gas inlet temperature to calculate the second theoretical moisture content. %. Calculate the disturbance difference. %, sensitivity coefficient % / °C. The perturbation coefficient is obtained after Min-Max normalization. .

[0059] Multi-objective optimization is employed, assuming the current condition is normal, with weights... The solution yielded a flue gas temperature adjustment of +15°C, a flow rate adjustment of +0.2 m / s, and a rotational speed adjustment of -0.2 rpm. The actuator was adjusted accordingly, and after 30 seconds, the moisture content was detected to have decreased to 12.5%, gradually approaching the target value.

[0060] Example 2: The difference between this example and Example 1 is that it adds a fault identification and handling function for the drying system based on multi-level diagnosis: Continuously calculate the first theoretical moisture content Compared with the actual moisture content detected in real time Prediction deviation between :

[0061] The prediction deviation reflects the degree of deviation between the predicted values ​​of the drying kinetic model and the actual measured values, and is an important basis for judging the accuracy of the model and the operating status of the system.

[0062] Two threshold levels are set for graded diagnosis: First preset threshold (e.g., 2%) and the second preset threshold (e.g., 5%), and The judgment logic adopts the "continuous exceedance" principle, that is, the judgment is triggered only if the condition is met in multiple consecutive control cycles, so as to avoid misjudgment caused by instantaneous fluctuations to a certain extent.

[0063] when And continuously for the first preset duration If the time interval is 5 minutes (i.e., 10 consecutive control cycles), it indicates that the model's predictive ability may have decreased or the sensor may be malfunctioning, indicating that the drying kinetics model is inaccurate or the moisture content detection sensor is faulty. At this time, the model calibration procedure is triggered to check the sensor calibration status or prompt for recalibration of the model parameters.

[0064] when Furthermore, the parameter adjustment data determined in S5 has been adjusted more than three times consecutively (i.e., adjustments have been performed for three consecutive cycles). When there is no decreasing trend (i.e., the current deviation is not less than 90% of the deviation 3 cycles ago), it indicates that the adjustment command has been issued but the process parameters have not improved accordingly. It is determined that the actuator response is abnormal, such as valve jamming, motor stalling or transmission mechanism slippage.

[0065] Monitoring parameter adjustment data on the frequency of adjustment of each drying parameter and adjustment range Establish diagnostic rules based on response characteristics: When the adjustment range of the same parameter exceeds the preset range threshold 5 times consecutively. (e.g., 20% of full scale), and the normalized perturbation coefficient corresponding to this parameter. Change When an adjustment command is issued but the model predicts almost no change in sensitivity, it indicates that the actuator has received a significant adjustment command but the actual process response is weak. This suggests that the actuator corresponding to the same parameter has a mechanical jamming or valve blockage fault. Here, "the change in the disturbance coefficient" specifically refers to the difference between the theoretically calculated sensitivity coefficient and the actually observed parameter-moisture content response under the current operating conditions.

[0066] When the adjustment direction of multiple parameters (two or more) exhibits periodic oscillations, i.e., repeated increases and decreases within adjacent control cycles (e.g., increase in the first cycle, decrease in the second cycle, increase in the third cycle), and the oscillation period is less than three control cycles, while the oscillation amplitude increases progressively, it is determined that the drying system has a coupled oscillation fault. This indicates a negative interaction between control loops or mechanical resonance in the equipment, requiring adjustment of the control strategy or inspection of the equipment's mechanical connections.

[0067] Calculate the drying data for the current batch. Compared with historical normal operating condition benchmark dataset Statistical distance The statistical distance is used to quantify the degree of deviation between the current operating conditions and historical normal operating conditions, and to identify sudden changes in feed characteristics or hidden thermal faults.

[0068] First, establish a historical baseline dataset based on normal operating conditions: collect drying data under stable operating conditions over the past 30 days, with one sample every 10 minutes, and calculate the mean vector. Covariance Matrix The baseline dataset is updated every 24 hours, and outlier historical data exceeding 3 standard deviations are removed.

[0069] Statistical distance is calculated using Mahalanobis distance:

[0070] in, This is the drying data vector for the current batch, containing parameters such as temperature, flow rate, rotation speed, and material layer thickness. This is a vector of historical means. This is the covariance matrix, reflecting the correlation between the parameters; superscript Indicates matrix transpose; superscript This represents finding the inverse of a matrix.

[0071] when When a preset distance threshold (e.g., Mahalanobis distance 3.0) is reached, it is determined that the characteristics of the feed semi-coke have undergone an abnormal change (e.g., the initial moisture content suddenly increases from 18% to 25%) or there is a hidden thermal fault in the drying system (e.g., abnormal heat loss due to damage to the drum insulation layer). At this time, an audible and visual alarm is triggered and the system automatically switches to conservative control mode: reducing the flue gas temperature by 10% and the drum speed by 15% to reduce the production load and ensure equipment safety while awaiting manual confirmation.

[0072] In a specific example, during continuous operation, the drying system of Factory B experienced prediction deviations. If the deviation exceeds 2% for six consecutive minutes, the S1 diagnostic is triggered. The system automatically checks the sensor status and detects a drift in the calibration coefficient of the microwave moisture content meter, prompting the operator to perform calibration. After calibration, the prediction deviation returns to normal.

[0073] In another case, Factory C detected that the flue gas temperature parameter was adjusted by more than 20% of full scale five times consecutively, but the corresponding change in the disturbance coefficient was only 0.05, far less than the threshold of 0.1. The system determined that the flue gas valve was mechanically stuck and issued a maintenance alarm, prompting the inspection of the valve actuator. Maintenance personnel found that the valve bearing was worn, and after replacement, the fault was resolved.

[0074] In the case of Factory D, the calculated Mahalanobis distance between the current batch data and the historical baseline was 4.2, exceeding the threshold of 3.0. The system determined that the feed characteristics were abnormal and switched to conservative mode. Manual inspection revealed an anomaly in the upstream wet quenching process, causing the feed moisture content to surge from the normal 20% to 28%. Because the system switched to conservative mode in time, a large number of defective products were avoided.

[0075] Example 3: The difference between this example and Example 1 is that an online update function for the drying kinetics model based on the recursive least squares method is added: When the drying kinetics model is determined to be inaccurate or the prediction is biased in Example 2 Exceeding the third preset threshold for 5 consecutive control cycles When the accuracy drops to 3%, the online update process for model parameters is triggered. The third preset threshold is between the first and second preset thresholds and is used to determine whether the model prediction accuracy has decreased to the point where adaptive correction is needed but has not reached the level of a serious fault.

[0076] Actual output moisture content after parameter adjustment Calculate the moisture content compared to the first theoretical moisture content. Prediction error :

[0077] Due to the key parameters in the drying kinetics model (such as the pre-exponential factor in the effective moisture diffusion coefficient) and activation energy Since it cannot be measured directly, an indirect parameter identification method is used.

[0078] An observational equation for the drying rate is established. Based on the model in Example 1, at discrete time steps... The observed average drying rate can be approximated as:

[0079] Taking the logarithm of the expression for the effective water diffusion coefficient in Example 1, a linear regression model is established:

[0080] in, Let be the error term. , , The model is then modeled in standard linear form. ,in This is the parameter vector to be updated.

[0081] The parameter vector is updated online using the recursive least squares (RLS) method. :

[0082]

[0083]

[0084] in, for The parameter estimates at time 1; The gain matrix determines the step size for parameter correction. The covariance matrix reflects the uncertainty of parameter estimation, with initial values... Take 100 times the size of the identity matrix; The forgetting factor, with a value ranging from 0.95 to 0.99, is used to reduce the weight of historical data, enhance the sensitivity to new data, and enable the model to quickly track system characteristic drift caused by changes in material properties or equipment aging. It is the identity matrix; for The observed value at time; for The regression vector at time step.

[0085] The parameter vector is updated in real time through the above recursive calculation. This updates the pre-exponential factor. and activation energy Based on the updated model parameters, the first theoretical moisture content was recalculated. Second theoretical moisture content This update allows the drying kinetics model to adapt to changes in material properties or system characteristic drift caused by equipment aging. After the update, the continuous exceedance counter is reset to zero, restoring normal control procedures.

[0086] In one specific case, after three months of operation, the drying system at Plant E experienced a change in the source of semi-coke raw materials, which altered the material's pore structure and caused the original model parameters to become inaccurate, with prediction deviations exceeding 3% for six consecutive cycles. This triggered the S1 update process.

[0087] Collect current operating condition data: K, m / s, m, actual output moisture content %, initial moisture content %, stay time min. Calculate the observed drying rate. % / min. Let: .

[0088] Regression vector .

[0089] Update using recursive least squares method, assuming the current... (correspond , ), For a diagonal matrix, the forgetting factor is... Calculate the gain matrix. Updated .

[0090] After the update m² / s, J / mol. The first theoretical moisture content was recalculated based on the new parameters, and the deviation from the measured value decreased to 1.2%, restoring model accuracy, and the system continued to operate normally. Subsequently, the model accuracy was automatically checked every 24 hours to ensure long-term stable operation to a certain extent.

[0091] Example 4: This example discloses a system for drying semi-coke using waste heat from flue gas. The system includes a memory and a processor. The memory contains a computer-readable storage medium; When the processor processes a computer program stored on the computer-readable storage medium, it implements the method.

[0092] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A method for drying semi-coke using waste heat from flue gas, applied to a semi-coke dryer, characterized in that, include: The moisture content of the semi-coke at the outlet of the semi-coke dryer is detected in real time. When the moisture content is greater than the preset moisture content threshold, the drying data of the current batch is obtained. The first theoretical moisture content is calculated based on the drying data of the current batch processed using a drying kinetics model. The drying data of the current batch is slightly perturbed to generate perturbed data, and the perturbed data is processed based on the drying kinetics model to calculate the second theoretical moisture content; The disturbance difference is determined based on the first theoretical moisture content and the second theoretical moisture content, and the disturbance coefficient is determined based on the disturbance difference and the small disturbance. The parameter adjustment data is determined based on the perturbation coefficient and the preset optimization objective function.

2. The method for drying semi-coke using waste heat from flue gas according to claim 1, characterized in that, The first theoretical moisture content and the second theoretical moisture content are calculated using a drying kinetic model constructed based on the thermal-mass coupling theory of porous media. The drying kinetic model includes a correlation function between the effective moisture diffusion coefficient and temperature, gas velocity, and material layer thickness.

3. The method for drying semi-coke using waste heat from flue gas according to claim 1, characterized in that, The small perturbation includes single-parameter perturbation or combined perturbation. The single-parameter perturbation is to apply an increment or decrement to a single parameter, and the combined perturbation is to apply perturbations of different magnitudes to multiple parameters simultaneously.

4. The method for drying semi-coke using waste heat from flue gas according to claim 1, characterized in that, The determination of the perturbation coefficient includes calculating the ratio of the perturbation difference to the minute perturbation, and normalizing the calculation result to eliminate dimensional differences.

5. The method for drying semi-coke using waste heat from flue gas according to claim 1, characterized in that, The preset optimization objective function includes an energy consumption objective sub-function, a quality objective sub-function, and a safety objective sub-function. The energy consumption objective sub-function aims to minimize the waste heat consumption of flue gas and the power consumption of the fan. The quality objective sub-function aims to minimize the deviation between the output moisture content and the target moisture content. The safety objective sub-function aims to maximize the safety margin between the surface temperature and the critical oxidation temperature of semi-coke.

6. The method for drying semi-coke using waste heat from flue gas according to any one of claims 1-5, further comprising: Calculate the predicted deviation between the first theoretical moisture content and the actual moisture content detected in real time; When the prediction deviation exceeds the first preset threshold and continues for the first preset duration, it is determined that the drying kinetics model is inaccurate or the moisture content detection sensor is faulty. When the prediction deviation exceeds the second preset threshold and the parameter adjustment data has been adjusted multiple times but the prediction deviation shows no decreasing trend, the actuator response is determined to be abnormal. Wherein, the second preset threshold is greater than the first preset threshold.

7. The method for drying semi-coke using waste heat from flue gas according to claim 6, characterized in that, Also includes: Monitor the frequency and magnitude of adjustments to each drying parameter based on the parameter adjustment data; When the adjustment range of the same parameter exceeds the preset range threshold multiple times consecutively and the change corresponding to the disturbance coefficient is less than the preset change threshold, it is determined that the actuator corresponding to the same parameter has a mechanical jamming or valve blockage fault. When the adjustment direction of multiple parameters exhibits periodic oscillations and the oscillation amplitude increases, it is determined that there is a coupled oscillation fault in the drying system.

8. The method for drying semi-coke using waste heat from flue gas according to claim 7, characterized in that, Also includes: Calculate the statistical distance between the current batch drying data and historical drying data under normal operating conditions; When the statistical distance exceeds the preset distance threshold, it is determined that the characteristics of the feed semi-coke have undergone abnormal changes or that there is a hidden thermal fault in the drying system, triggering an alarm and switching to conservative control mode.

9. The method for drying semi-coke using waste heat from flue gas according to claim 8, characterized in that, Also includes: When it is determined that the drying kinetics model is inaccurate or the prediction deviation continues to exceed the third preset threshold, the actual output moisture content after parameter adjustment is collected, and the prediction error between the actual output moisture content and the first theoretical moisture content is calculated. The key parameters in the drying kinetics model are updated online using the recursive least squares method, and the first theoretical moisture content and the second theoretical moisture content are recalculated based on the updated model.

10. A system for drying semi-coke using waste heat from flue gas, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory is provided with a computer-readable storage medium, and a computer program is stored on the computer-readable storage medium. When the processor processes a computer program stored on the computer-readable storage medium, it implements the method as described in any one of claims 1-9.

Images