Online monitoring system for slurry of flue gas desulfurization tower

Abstract

The invention provides an on-line monitoring system for slurry of a flue gas desulfurization tower. The on-line monitoring system comprises a defoaming module used for defoaming the slurry from the bottom of an absorption tower; the density analysis module is used for measuring the density of the slurry in real time through a density probe; the pH analysis module is used for measuring the pH value of the slurry in real time through a pH probe; the multi-component analysis module is used for analyzing the contents of suspended matters, chloride ions and sulfate radicals in the slurry in real time through a multi-component analysis probe; the flow cell module is used for accommodating slurry and is provided with a pH probe, a density probe and a multi-component analysis probe; the control valve module is used for controlling flow velocity and flow of slurry flowing into the flow cell module from the bottom of the absorption tower; the purging module is used for generating pulse airflow through compressed air and intermittently purging the surfaces of the pH probe, the density probe and the multi-component analysis probe; the upper computer module is used for displaying measurement data of the pH analysis module, the density analysis module and the multi-component analysis module, so that the problems of measurement misalignment and frequent maintenance are avoided.

Description

Technical Field

[0001] This invention relates to the field of industrial flue gas pollution control and online monitoring technology, and in particular to an online monitoring system for flue gas desulfurization tower slurry. Background Technology

[0002] Currently, coal-fired power units commonly use limestone-gypsum wet desulfurization technology to control sulfur dioxide emissions. In this process, limestone slurry in the absorption tower reacts with flue gas to generate calcium sulfite, which is further oxidized to calcium sulfate (gypsum). During this process, parameters such as slurry density, pH value, chloride ion content, sulfate content, and suspended solids content directly affect desulfurization efficiency, equipment corrosion level, and the quality of the byproduct gypsum.

[0003] Traditional methods rely primarily on manual periodic sampling and offline testing, which suffers from problems such as detection lag and large human errors, making it difficult to achieve real-time optimization and precise control of the process. In recent years, some power plants have attempted to adopt online monitoring devices, but due to the complex composition, corrosiveness, and tendency to scale in desulfurization slurry, the detection probes often become clogged or contaminated, leading to inaccurate measurements. This necessitates frequent manual cleaning and maintenance, increasing operation and maintenance costs and affecting the continuity and reliability of monitoring. Existing technologies lack intelligent judgment of probe scaling mechanisms and adaptive cleaning strategies, and also fail to deeply integrate multi-parameter monitoring data with real-time control of the desulfurization process, thus failing to fundamentally improve the economy and stability of desulfurization system operation. Summary of the Invention

[0004] This invention provides an online monitoring system for flue gas desulfurization tower slurry, which solves the technical problems of inaccurate measurement and frequent maintenance caused by slurry scaling and contamination in existing monitoring devices.

[0005] On one hand, the present invention provides an online monitoring system for flue gas desulfurization tower slurry, comprising:

[0006] The defoaming module is used to defoam the slurry from the bottom of the absorption tower.

[0007] The density analysis module is used to measure the density of the slurry in real time using a density probe;

[0008] The pH analysis module is used to measure the pH value of the slurry in real time using a pH probe;

[0009] The multi-component analysis module is used to analyze the content of suspended solids, chloride ions and sulfate ions in slurry in real time through a multi-component analysis probe.

[0010] The flow cell module is used to contain the slurry and install pH probes, density probes, and multi-component analysis probes.

[0011] The control valve module is used to control the flow rate and volume of the slurry flowing from the bottom of the absorption tower into the flow tank module;

[0012] The purging module is used to generate pulsed airflow using compressed air to intermittently purge the surfaces of pH probes, density probes, and multi-component analysis probes;

[0013] The host computer module is used to display the measurement data from the pH analysis module, density analysis module, and multi-component analysis module.

[0014] Optionally, the defoaming module includes a rotating separation unit and a stationary cavity unit;

[0015] The rotating separation unit is located inside the stationary cavity unit and forms a separation gap with the stationary cavity unit;

[0016] When the rotary separation unit rotates, the gas carrying foam in the slurry is separated into gas and liquid under the action of centrifugal force, so that the separated liquid is discharged through the separation gap and the separated gas is discharged through the exhaust port at the top of the stationary cavity unit.

[0017] Optionally, the multi-component analysis module employs a Transformer model, which is trained in the following manner:

[0018] Using training set sample data, the input spectral features were sorted through SHAP analysis to screen out the key wavelength variables that contribute the most to the prediction of suspended solids, chloride ions, and sulfate content.

[0019] Based on key wavelength variables, the Transformer model is trained using training set sample data, and the weight distribution of the query, key, and value matrices of the self-attention mechanism in the Transformer model is dynamically adjusted to enhance the model's ability to perceive key spectral features.

[0020] The root mean square error (RMSE) during training is monitored using validation set data. If the RMS error of the validation set data does not decrease for several consecutive training cycles, training is terminated.

[0021] Optionally, the defoaming module is also used for;

[0022] Real-time acquisition of the operating current of the drive motor of the rotary separation unit;

[0023] Compare the operating current with the preset current threshold range;

[0024] When the operating current is consistently higher than the upper limit of the current threshold range, the drive motor is controlled to increase its speed.

[0025] When the operating current remains below the lower limit of the current threshold range, the drive motor is controlled to reduce its speed.

[0026] The current threshold range is determined in advance using the following method:

[0027] Measure the motor operating current corresponding to slurry with different foaming levels, and calibrate the current threshold range required for defoaming.

[0028] Optionally, the defoaming module is also used for:

[0029] Continuously monitor the operating current of the drive motor and calculate its average value and trend slope within a preset sliding time window;

[0030] When the average value is consistently higher than the upper limit of the current threshold range and the trend slope is positive, it is determined that the foaming degree of the slurry is intensifying, and the drive motor is controlled to increase its speed in a stepwise manner.

[0031] Optionally, the host computer module is also used for:

[0032] Cluster analysis was performed on historical and real-time collected slurry multi-component data to obtain different slurry state clusters; each slurry state cluster corresponds to a component combination and operating condition.

[0033] Within each slurry state cluster, historical data records that triggered cleaning alarms were statistically analyzed to identify the most significant factors causing probe contamination or measurement inaccuracies within that slurry state cluster, which were then used as cleaning hotspots for that slurry state cluster.

[0034] Based on the cleaning hotspots, dynamically configure differentiated performance evaluation thresholds and cleaning trigger thresholds for different slurry state clusters;

[0035] Based on each slurry state cluster and its corresponding cleaning hotspot, a targeted cleaning strategy is generated to match it.

[0036] The current slurry state is matched with the clustering model in real time, and the corresponding cleaning strategy for that state cluster is executed.

[0037] Optionally, the purging module is also used for:

[0038] The first stability index of pH measurement data was coupled with the second stability index of chloride ion content measurement data for analysis.

[0039] When the product of the fluctuation coefficients of the first stability index and the second stability index exceeds the preset coupled fluctuation threshold, and the fluctuation coefficients of each index do not reach their respective preset independent thresholds, probe contamination is determined to exist, and preventive purging is initiated.

[0040] The stability index is calculated by the ratio of the standard deviation to the mean over multiple consecutive measurement periods.

[0041] Optionally, the multi-component analysis module is also used for:

[0042] Real-time monitoring of the ratio of scattering spectral intensity to absorption spectral intensity in the near-infrared spectrum;

[0043] When this ratio changes in a trend, it is determined that a dirt layer with specific optical properties has formed on the probe surface;

[0044] The host computer module will suggest that the operator switch to the maintenance mode based on ultrasonic cleaning and automatically generate the corresponding control commands.

[0045] Optionally, the host computer module is also used for pattern recognition of changes in spectral ratios:

[0046] Establish a dirt spectral fingerprint database based on historical data. The fingerprint database contains the variation patterns of the ratio between scattered light intensity and absorbed light intensity corresponding to various dirt types.

[0047] The real-time monitored trend of ratio changes is matched with patterns in the fingerprint database;

[0048] If the match is successful, the system will recommend switching to ultrasonic cleaning mode, display the specific type of dirt diagnosed on the interface, and recommend an ultrasonic cleaning frequency and duration that match the type of dirt.

[0049] Optionally, the purging module is used for:

[0050] By controlling the frequency, intensity, and duration of the pulsed airflow, the surfaces of pH probes, density probes, and multi-component analysis probes are intermittently purged to achieve cleanliness and avoid interference with the measurement environment.

[0051] This invention provides an online monitoring system for flue gas desulfurization tower slurry. It improves detection accuracy through defoaming pretreatment, achieves comprehensive parameter monitoring through multi-probe collaborative operation, and dynamically adjusts maintenance strategies by combining a purging module and data analysis. A host computer performs coupled analysis of multi-parameter fluctuations, enabling intelligent assessment of pollution levels and triggering of cleaning. Through this technical solution, simultaneous online detection of multiple slurry parameters is achieved, eliminating the influence of bubble interference on measurements, and extending the probe cleaning cycle through an intelligent purging mechanism. The system can reflect changes in slurry composition in real time, providing data support for desulfurization process adjustments, reducing the frequency of manual sampling and cleaning maintenance, and improving the continuity and reliability of the monitoring system operation. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0053] Figure 1 This is one of the structural schematic diagrams of the online monitoring system for flue gas desulfurization tower slurry provided in the embodiments of the present invention;

[0054] Figure 2 This is the second schematic diagram of the online monitoring system for flue gas desulfurization tower slurry provided in this embodiment of the invention;

[0055] Figure 3 This is a schematic flowchart of the online monitoring method for flue gas desulfurization tower slurry provided in an embodiment of the present invention;

[0056] Figure 4 This is a schematic diagram of the structure of the electronic device provided in an embodiment of the present invention. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0058] Figure 1 This is one of the structural schematic diagrams of the intelligent assessment system for the entire life cycle of power equipment provided in this embodiment of the invention.

[0059] See Figure 1 The online monitoring system 100 for flue gas desulfurization tower slurry includes a defoaming module 110, a density analysis module 120, a pH analysis module 130, a multi-component analysis module 140, a flow pool module 150, a control valve module 160, a purging module 170, and a host computer module 180.

[0060] The defoaming module 110 is used to defoam the slurry from the bottom of the absorption tower;

[0061] Specifically, a design combining a rotating device and a stationary chamber is adopted. The stirring shaft drives the rotating device to rotate, causing the gas carrying foam to undergo centrifugal separation within the rotating device. The relative motion between the rotating device and the stationary cylinder further facilitates gas-liquid separation, ultimately discharging the foam-free liquid, effectively reducing the interference of foam on probe measurements.

[0062] The density analysis module 120 is used to measure the density of the slurry in real time using a density probe;

[0063] Specifically, an ultrasonic density meter is used. The probe of the ultrasonic density meter is installed on the flow tank using a Pg13.5 threaded connection. As the ultrasonic waves pass through the slurry, the change in their propagation velocity is measured, and the density of the slurry is calculated using the instrument's built-in calculation program. This ensures real-time, continuous online measurement of density and stable operation of the desulfurization process. The density detector analyzes and processes the digital signal, calculates the liquid density using a specific algorithm, and can store, display, and transmit the data.

[0064] pH analysis module 130 is used to measure the pH value of slurry in real time via pH probe;

[0065] Specifically, a high-performance pH sensor suitable for harsh environments can be selected. It employs a strong alkali-resistant (HA) glass membrane suitable for harsh conditions. This membrane is resistant to strong oxidants, solvents, acids and alkalis, as well as poisoning substances, and remains stable under high temperature and pressure. It provides accurate measurement results even in harsh chemical environments, with a measurement range of 0–14. The pH meter electrode is directly inserted into the flow cell and connected using a Pg13.5 threaded connection to ensure stable sensor installation. Equipped with a VP cable connector, it offers excellent sealing and anti-interference capabilities, ensuring stable signal transmission and safe sensor operation. The electrode rod is made of glass, and the overall structure is robust and durable, capable of withstanding harsh environmental conditions such as high temperature and pressure, with a pressure resistance range of 0 to 12 bar and an operating temperature range of -5 to 130°C.

[0066] The multi-component analysis module 140 is used to analyze the content of suspended solids, chloride ions and sulfate ions in slurry in real time through a multi-component analysis probe;

[0067] Specifically, suspended matter can include unreacted limestone (CaCO3) particles, fly ash, and crystals of reaction intermediates or byproducts, such as calcium sulfite (CaSO3·1 / 2H2O) and gypsum (CaSO4·2H2O). Near-infrared spectroscopy combined with artificial intelligence algorithms enables online analysis of suspended matter, chloride ions, and sulfate ions. The hardware includes a fiber optic probe, a light source, a collimator, and a spectral detector. A low-pressure halogen tungsten lamp with strong light and high energy is used as the light source. After emission, the light is collimated to ensure a parallel optical path. The beam entering the monochromator through a slit or fiber optic cable is collimated using a concave mirror before illuminating the grating. The grating spectrum emitted from the grating is converged using another concave mirror, focusing the grating spectrum onto the focal plane. Array detector pixels are placed on this focal plane, and finally, an indium gallium arsenide detector converts the dispersed spectrum into an electrical signal.

[0068] To achieve real-time online multi-component analysis, an improved Transformer model was introduced. Global dependencies of features are extracted by stacking multiple layers of self-attention and feedforward networks. A single-layer encoder containing self-attention and feedforward networks is constructed using a Transformer model, and then multiple encoder layers are stacked to extract more complex features. The encoder output is mapped to the dimension of the target variable through a fully connected layer. In quantitative tasks, the encoder output is typically mapped to the target value through a fully connected layer. Finally, model training is performed. The training process for the Transformer model is similar to that of other neural networks. In each epoch, the training and validation sets are processed separately. During the training phase, forward propagation, loss calculation, backpropagation, and optimizer steps are performed. During the validation phase, the model performance (RMSE) is evaluated.

[0069] The multi-component online analysis subsystem and the main workflow for establishing the multi-component analysis method include: First, acquiring content and spectral data. Slurry samples are collected at different time points, and the contents of suspended solids, chloride ions, and sulfate ions are determined using a diaphragm vacuum pump and spectrophotometer according to standard methods. These values ​​are then used as the model's true values. Simultaneously, near-infrared spectral data of the slurry are acquired. Next, samples are randomly divided into training and test sets. An improved Transformer model is then constructed. The true values ​​and spectral data from the training set are input into the model for training. Important parameters such as the model's dimensionality, the number of heads in the multi-head self-attention system, the number of encoder layers, and the hidden layer dimension of the feedforward network are continuously optimized to reduce the RMSE between the model's predicted and actual values ​​until the requirements are met. Finally, the spectra of the test set samples are input into the multi-component model to obtain the model's prediction accuracy and robustness.

[0070] The flow cell module 150 is used to contain the slurry and to install a pH probe, a density probe, and a multi-component analysis probe.

[0071] The control valve module 160 is used to control the flow rate and volume of the slurry flowing from the bottom of the absorption tower into the flow pool module 150;

[0072] The purging module 170 is used to generate pulsed airflow using compressed air to intermittently purge the surfaces of the pH probe, density probe, and multi-component analysis probe;

[0073] Specifically, compressed air is used to generate pulsed airflow, which is intermittently blown onto the probe surface. By controlling the frequency, intensity, and duration of the pulses, the probe surface can be effectively cleaned while avoiding excessive gas blowing in and causing unnecessary impact on the measurement environment.

[0074] The host computer module 180 is used to display the measurement data of the pH analysis module 130, the density analysis module 120 and the multi-component analysis module 140;

[0075] Specifically, the host computer software is a Windows 10 system software, which connects to the pH online analysis subsystem, density online analysis subsystem, and multi-component online analysis subsystem via data cables. Its main functions are as follows:

[0076] I. Serial Port Settings: Automatically scans serial ports and displays available ones in the software. II. System Switch: Enables system on / off functionality; other functions are disabled when the system is off and enabled only when the system is on. III. Display Function: Real-time display of collected data on density, pH, suspended solids, chloride ion, and sulfate content.

[0077] In this embodiment, the defoaming module 110 refers to a device that eliminates air bubbles in the slurry through physical separation. It can be implemented using a combination structure of a rotating separation unit 1101 and a stationary cavity. The centrifugal force generated by the rotating separation unit 1101 causes gas-liquid stratification, preventing air bubbles from interfering with subsequent detection. The density probe is a sensor based on the principle of vibration, which can calculate the density value by measuring the change in the resonant frequency of the slurry. The multi-component analysis probe is a detection unit equipped with near-infrared spectroscopy analysis function, which can simultaneously analyze multiple ion concentrations. The flow cell is a sealed cavity with corrosion-resistant properties, which can be made of polytetrafluoroethylene (PTFE) material to provide a stable detection environment for the probe. The purging module 170 is a compressed air system controlled by a solenoid valve, which can generate intermittent pulsed airflow through a timing controller.

[0078] Specifically, the slurry from the bottom of the absorption tower is regulated by a control valve and then enters the defoaming module 110. The rotary separation unit 1101 separates the gas-containing foam into liquid slurry and waste gas. The pretreated slurry flows into a flow-through tank, where a density probe collects density data in real time, a pH probe monitors changes in acidity and alkalinity, and a multi-component probe obtains ion content through spectral analysis. The purging module 170 triggers pulsed airflow according to a preset cycle to remove deposits from the probe surface. The host computer integrates the data from each module, displays real-time parameter curves through a human-machine interface, and triggers cleaning and maintenance commands based on data fluctuation trends.

[0079] In this embodiment, defoaming pretreatment improves detection accuracy, multi-probe collaborative operation enables comprehensive parameter monitoring, and the purging module 170, combined with data analysis, dynamically adjusts maintenance strategies. A host computer performs coupled analysis of multi-parameter fluctuations, enabling intelligent assessment of contamination levels and triggering of cleaning. Through the above technical solutions, this embodiment achieves synchronous online detection of multiple slurry parameters, eliminates the influence of bubble interference on measurements, and extends the probe cleaning cycle through an intelligent purging mechanism. The system can reflect changes in slurry composition in real time, providing data support for desulfurization process adjustments, reducing the frequency of manual sampling and cleaning maintenance, and improving the continuity and reliability of the monitoring system operation.

[0080] In one embodiment of this specification, the defoaming module 110 includes a rotating separation unit 1101 and a stationary cavity unit 1102. The rotating separation unit 1101 is located inside the stationary cavity unit 1102 and forms a separation gap with the stationary cavity unit 1102. When the rotating separation unit 1101 rotates, the gas carrying foam in the slurry is separated into gas and liquid under the action of centrifugal force, so that the separated liquid is discharged through the separation gap and the separated gas is discharged through the exhaust port at the top of the stationary cavity unit 1102.

[0081] In this embodiment, the rotary separation unit 1101 refers to a device that generates centrifugal force through mechanical rotation, specifically a rotor structure with blades, whose rotation axis is driven by a motor. The stationary cavity unit 1102 refers to a fixed container structure, specifically a cylindrical or conical cavity structure, used to house the rotary separation unit 1101 and form a gas-liquid separation space. The separation gap refers to the annular channel between the outer wall of the rotary separation unit 1101 and the inner wall of the stationary cavity unit 1102, specifically achieved by adjusting the difference between the rotor diameter and the inner diameter of the cavity, used to control the cross-sectional area of ​​the liquid discharge path.

[0082] In this embodiment, specifically, after the slurry enters the static cavity unit 1102, the rotating separation unit 1101 rotates at high speed under the drive of a motor, causing the foam and liquid in the slurry to separate into layers under centrifugal force. The foam, due to its lower density, is thrown towards the center of rotation and discharged from the upper exhaust port with the airflow; the liquid, due to its higher density, is thrown to the outside, flows downward through the separation gap, and is discharged. During this process, the size of the separation gap must ensure smooth liquid discharge while preventing gas backflow. The rotation speed is dynamically adjusted according to the degree of slurry foaming; for example, when increased foaming is detected, the rotation speed is increased to enhance the centrifugal force.

[0083] In this embodiment, physical separation is achieved through centrifugal force, eliminating the need for chemical reagents. Furthermore, the clearance fit design between the rotating separation unit 1101 and the stationary cavity unit 1102 allows for adaptation to slurries of varying viscosities, avoiding the clogging risk associated with the fixed structure of traditional mechanical defoaming devices. This effectively separates foam and liquid from the slurry, reducing interference from bubbles on subsequent parameters such as density and pH. Simultaneously, dynamically adjusting the rotation speed adapts to defoaming requirements under different operating conditions, reducing the frequency of manual cleaning and ensuring the continuous and stable operation of the online monitoring system 100.

[0084] In one embodiment of this specification, the multi-component analysis module 140 employs...

[0085] The Transformer model is trained in the following way:

[0086] Using training set sample data, the input spectral features were sorted through SHAP analysis to screen out the key wavelength variables that contribute the most to the prediction of suspended solids, chloride ions, and sulfate content.

[0087] Based on key wavelength variables, the Transformer model is trained using training set sample data, and the weight distribution of the query, key, and value matrices of the self-attention mechanism in the Transformer model is dynamically adjusted to enhance the model's ability to perceive key spectral features.

[0088] The root mean square error (RMSE) during training is monitored using validation set data. If the RMS error of the validation set data does not decrease for several consecutive training cycles, training is terminated.

[0089] In this embodiment, SHAP analysis refers to a game theory-based machine learning model interpretation method. Specifically, it can be implemented by calculating the contribution of each spectral wavelength feature to the prediction result, used to identify the core spectral bands affecting the accuracy of component detection. Key wavelength variables refer to spectral bands sensitive to changes in target component concentration. This can be achieved by selecting the top 20% of wavelengths by contribution after ranking SHAP values, used to reduce the dimensionality of the model input and eliminate redundant information interference. Self-attention mechanism weight allocation refers to the process of adjusting the correlation strength between different features in a neural network. This can be achieved by dynamically optimizing the product of the query matrix and the key matrix using the backpropagation algorithm, used to strengthen the model's attention to key spectral features. Validation set root mean square error monitoring refers to the evaluation of generalization ability during model training. This can be achieved by using a sliding window to statistically analyze the convergence trend of the validation set prediction error, used to prevent the model from overfitting the training data.

[0090] In this embodiment, the training process first uses SHAP analysis to rank the spectral data by feature importance, selecting the spectral wavelengths most strongly correlated with changes in suspended matter, chloride ion, and sulfate concentrations. These key wavelengths are used as input features for the Transformer model. During model training, the self-attention mechanism dynamically adjusts weight allocation, enabling the neural network to focus on spectral regions with high discriminative power. The training process terminates early when the prediction error on the validation set does not show a decreasing trend over consecutive training cycles to prevent the model from over-relying on noise in the training data.

[0091] In this embodiment, key wavelengths are dynamically screened through SHAP analysis, and dynamic weight adjustment using a self-attention mechanism enables the model to automatically capture the core spectral features affecting component concentrations under different operating conditions. This overcomes the detection inaccuracy problem of fixed wavelength selection methods under fluctuating operating conditions. The multi-component detection model achieves adaptability to complex operating conditions, effectively improving the online detection accuracy of suspended solids, chloride ions, and sulfate ions. Simultaneously, a validation set error monitoring mechanism avoids model overfitting, ensuring the stability of the detection system under different operating conditions and reducing maintenance needs due to model failure.

[0092] In one embodiment of this specification, the defoaming module 110 is further configured to acquire the operating current of the drive motor of the rotary separation unit 1101 in real time, compare the operating current with a preset current threshold range, control the drive motor to increase its speed when the operating current is continuously higher than the upper limit of the current threshold range, and control the drive motor to decrease its speed when the operating current is continuously lower than the lower limit of the current threshold range. The current threshold range is pre-calibrated by measuring the motor operating current corresponding to slurries with different foaming degrees to determine the current threshold range required for defoaming.

[0093] In this embodiment, the operating current refers to the current signal generated by the drive motor during the operation of the rotating separation unit 1101. Specifically, it can be collected in real time by a current sensor and transmitted to the control unit. This parameter can indirectly reflect the load change of the separation unit caused by the degree of slurry foaming. The current threshold range refers to the motor operating current range pre-calibrated through experiments. Specifically, it can be determined by collecting motor operating data corresponding to slurry samples with different foam concentrations and verifying the defoaming effect. This range is used to determine whether the speed needs to be adjusted under the current operating conditions. Speed ​​control refers to dynamically adjusting the output power of the drive motor according to the current monitoring results. Specifically, it can be implemented using a frequency converter or a servo controller to maintain the stability of the defoaming efficiency by matching the actual degree of slurry foaming.

[0094] In this embodiment, specifically, the defoaming module 110 continuously monitors the operating load of the drive motor through a current sensor. When the foaming degree of the slurry increases, the resistance experienced by the rotating separation unit 1101 increases, causing the motor operating current to rise. If the current continues to exceed the preset upper limit, the system automatically increases the motor speed to enhance centrifugal force and achieve effective defoaming; if the current continues to be below the preset lower limit, the speed is reduced to avoid energy waste. The current threshold range is calibrated by simulating slurry samples under different working conditions in the laboratory. For example, slurries with foam volume ratios of 5%, 10%, and 15% are input into the defoaming module 110, the corresponding current values ​​are recorded, and the defoaming effect is verified to determine the optimal threshold range.

[0095] In this embodiment, closed-loop speed control is achieved through current monitoring and threshold judgment, ensuring defoaming efficiency under different operating conditions while avoiding ineffective power consumption of the equipment under low load. Through the above technical solution, this embodiment can sense changes in the degree of slurry foaming in real time and automatically adjust the defoaming intensity, solving the problems of incomplete defoaming or equipment overload caused by the lag in manual adjustment in traditional methods. The experimental calibration method for the current threshold range ensures the matching of the control logic with actual operating conditions, effectively extending the service life of the equipment and reducing maintenance frequency.

[0096] In one embodiment of this specification, the defoaming module 110 is also used to continuously monitor the operating current of the drive motor and calculate its average value and trend slope within a preset sliding time window; when the average value is continuously higher than the upper limit of the current threshold range and the trend slope is positive, it is determined that the degree of foaming of the slurry is intensifying, and the drive motor is controlled to increase its speed in a stepwise manner.

[0097] In this embodiment, the preset sliding time window refers to the time interval used for real-time monitoring of the changing trend of the drive motor's operating current. Specifically, it can be implemented using a fixed-duration or dynamically adjusted algorithm to capture the continuous characteristics of current changes. The average value and trend slope are statistical quantities calculated from the current data within the sliding window, specifically implemented using a moving average algorithm and a linear regression method, to quantify the amplitude and direction of current changes. Stepwise speed increase refers to gradually increasing the rotational speed of the drive motor in stages, specifically implemented using a discrete speed control strategy to avoid mechanical shocks to the equipment caused by sudden speed changes.

[0098] In this embodiment, specifically during the slurry defoaming process, the operating current of the drive motor is continuously collected and input into the calculation unit. By setting a sliding time window, for example, with a period of 30 seconds, the current data is calculated using a moving average, and the slope of the current change over time within the window is fitted using the least squares method. When the system detects that the average current value exceeds a preset upper limit and the slope remains positive, it indicates that the degree of slurry foaming is dynamically increasing. At this time, the defoaming module 110 increases the speed of the drive motor in stages according to preset speed step parameters, for example, increasing by 50 rpm each time and maintaining stable operation for 5 seconds before re-evaluating, thereby enhancing the centrifugal separation effect.

[0099] In this embodiment, by introducing trend analysis within a sliding window, the trend of slurry foaming intensification can be predicted in advance, and gradual speed adjustment can be implemented within the equipment's load-bearing capacity. This avoids equipment failure caused by sudden load impacts and achieves a balance between defoaming efficiency and equipment stability. Through the above technical solution, this application can initiate enhanced defoaming measures in advance before the slurry foaming degree reaches a critical state, effectively suppressing the interference of foam accumulation on subsequent detection modules. The stepped speed adjustment mechanism can prevent equipment vibration or overload caused by sudden speed changes and extend the service life of the drive motor. This solution significantly improves the working condition adaptability and system control accuracy of the defoaming module 110 by dynamically responding to changes in slurry state, and reduces the frequency of manual intervention.

[0100] In one embodiment of this specification, the host computer module 180 is further configured to:

[0101] Cluster analysis was performed on historical and real-time collected slurry multi-component data to obtain different slurry state clusters; each slurry state cluster corresponds to a component combination and operating condition.

[0102] Within each slurry state cluster, historical data records that triggered cleaning alarms were statistically analyzed to identify the most significant factors causing probe contamination or measurement inaccuracies within that slurry state cluster, which were then used as cleaning hotspots for that slurry state cluster.

[0103] Based on the cleaning hotspots, dynamically configure differentiated performance evaluation thresholds and cleaning trigger thresholds for different slurry state clusters;

[0104] Based on each slurry state cluster and its corresponding cleaning hotspot, a targeted cleaning strategy is generated to match it.

[0105] The current slurry state is matched with the clustering model in real time, and the corresponding cleaning strategy for that state cluster is executed.

[0106] In this embodiment, cluster analysis refers to classifying slurry data with similar component combinations and operating conditions into the same category using machine learning algorithms. Specifically, K-means or hierarchical clustering algorithms can be used to establish the correlation between slurry state and contamination risk. Cleaning hotspots refer to identifying the key factors most likely to cause probe contamination under specific slurry states by statistically analyzing historical alarm data. This can be achieved using frequency statistics or association rule mining methods to focus on the main contamination sources under different operating conditions. Dynamically configuring differentiated performance evaluation thresholds and cleaning trigger thresholds refers to adjusting the alarm trigger conditions for different slurry state clusters based on cleaning hotspots. Specifically, thresholds can be set based on the standard deviation multiple of the data distribution within a cluster to avoid false positives or false negatives caused by uniform thresholds. Targeted cleaning strategies refer to developing matching cleaning schemes for the contamination characteristics of different slurry state clusters. Specifically, decision tree models can be used to generate cleaning parameter combinations to improve the effectiveness of cleaning operations.

[0107] Specifically, multi-component data of the slurry is first input into a clustering model, forming multiple slurry state clusters through distance calculations in the feature vector space. Historical alarm data within each state cluster is statistically analyzed, and the primary factor causing probe failure in that cluster is determined by calculating the alarm trigger frequency ratio of each contaminant. Based on the characteristic distribution of this primary factor, the performance evaluation threshold for that cluster is recalculated. For example, when suspended solids are the main contaminant, the cleaning trigger threshold for suspended solids concentration is set to the 95th percentile of the concentration distribution of that cluster. The cleaning strategy generation module matches the corresponding physical cleaning method according to the contamination type. For example, high-frequency pulse purging is used for crystalline contaminants, and gradient pressure regulation is used for sedimentary contaminants. Real-time monitoring data is mapped to the established clustering model. When the current slurry state is detected to enter a specific state cluster, the corresponding cleaning parameters for that cluster are automatically invoked to execute the operation.

[0108] In this embodiment, data-driven dynamic threshold adjustment accurately identifies the contamination risk level corresponding to different component combinations, allowing for targeted matching of cleaning intensity and frequency. This ensures probe cleanliness while reducing unnecessary cleaning attempts. By establishing a correlation model between slurry state and contamination characteristics, dynamic optimization of the cleaning strategy is achieved, effectively reducing false alarms caused by unreasonable threshold settings and improving the accuracy of probe maintenance and the continuity of system operation.

[0109] In one embodiment of this specification, the purging module 170 performs coupled analysis on a first stability index of pH measurement data and a second stability index of chloride ion content measurement data. When the product of the fluctuation coefficients of the first stability index and the second stability index exceeds a preset coupled fluctuation threshold, and their respective standard fluctuation coefficients do not reach their respective preset independent thresholds, it is determined that there is slight probe contamination, and a low-intensity, high-frequency preventive purging is initiated. The stability index is calculated by the ratio of the standard deviation to the mean of multiple consecutive measurement cycles.

[0110] In this embodiment, the first stability index reflects the dispersion of pH measurement data over a continuous period. Specifically, it can be calculated by dividing the standard deviation of pH values ​​over ten consecutive measurement periods by their mean. This index characterizes whether the pH probe's measurement fluctuations exceed the normal range. The second stability index reflects the fluctuation characteristics of chloride ion content measurement data. Specifically, it can be calculated by dividing the standard deviation of chloride ion measurements over the same number of periods by their mean. This is used to determine whether the chloride ion probe experiences abnormal interference. The coupled fluctuation threshold is the critical judgment value of the product of the two indices. Specifically, it can be determined by analyzing historical data to identify the distribution range of the product of the two indices during periods of slight pollution. Its function is to identify potential pollution states where a single indicator does not exceed the standard but there is compound interference.

[0111] In this embodiment, specifically, when contaminants begin to adhere to the surfaces of the pH probe and the chloride ion probe, their measurement data will exhibit asynchronous, small fluctuations. By calculating the product of their stability indices in real time and comparing it with a preset threshold, the trend of combined contamination can be identified in advance before the contaminants cause significant inaccuracies in a single probe. For example, when the pH stability index is 0.08 and the chloride ion stability index is 0.12, if their product 0.0096 exceeds the preset threshold of 0.008, a preventative purging is triggered. At this time, the purging module 170 initiates low-intensity but high-frequency compressed air pulses to remove initial deposits without interrupting the measurement. This process continues until the product of the two indices falls back below the threshold, thereby maintaining the continuous availability of the probes.

[0112] In this embodiment, by coupling the stability indices of two key parameters, slight contamination can be detected earlier, avoiding measurement inaccuracies caused by contaminant accumulation. Existing technologies using high-frequency, fixed-cycle purging methods easily lead to energy waste and equipment wear, while this solution only initiates purging when composite fluctuations exceed limits, significantly optimizing maintenance efficiency. Through the above technical solution, this application can implement preventative maintenance in the early stages of probe contamination, effectively extending the interval between two manual cleanings. By coupling the analysis of multi-parameter fluctuation characteristics, misjudgments caused by single-parameter fluctuations are avoided, improving the accuracy of the cleaning triggering mechanism. While maintaining measurement accuracy, unnecessary purging actions are reduced, lowering compressed air consumption and equipment mechanical wear, ensuring the long-term stable operation of the monitoring system.

[0113] In one embodiment of this specification, the multi-component analysis module 140 is also used to monitor in real time the ratio of the scattering spectral intensity to the absorption spectral intensity of the near-infrared spectrum; when the ratio changes in a trend, it is determined that a dirt layer with specific optical properties may have formed on the probe surface; the host computer module 180 will suggest that the operator switch to the maintenance mode based on ultrasonic cleaning and automatically generate corresponding control commands.

[0114] In this embodiment, the ratio of the scattering spectral intensity to the absorption spectral intensity in the near-infrared spectrum refers to the numerical proportional relationship between the intensity of the scattered light signal and the intensity of the absorbed light signal obtained by the spectral sensor. Specifically, this can be achieved using a spectrophotometer in conjunction with a photodetector. Changes in this ratio can reflect changes in the optical properties of the slurry composition or the deposits on the probe surface. A dirt layer with specific optical properties refers to a deposit that produces a specific scattering or absorption effect on light waves. For example, the scattering rate of a gypsum crystal layer is higher than its absorption rate for near-infrared light. This can be achieved through pattern recognition by establishing a database of the optical characteristics of different dirt materials. The maintenance mode based on ultrasonic cleaning refers to a physical cleaning method that uses high-frequency mechanical vibration to peel off hard scale from the probe surface. Specifically, this can be achieved using a piezoelectric ceramic transducer to generate ultrasonic waves of 20-40 kHz.

[0115] Specifically, the multi-component analysis module 140 continuously collects near-infrared spectral data of the slurry and calculates the dynamic ratio of scattering intensity to absorption intensity. When this ratio shows a continuous upward or downward trend, it determines whether a specific type of fouling layer has formed on the probe surface, based on a preset optical characteristic database. For example, a thin layer of gypsum can significantly increase scattering intensity, while oily contaminants may cause an increase in absorption intensity. The host computer module 180 matches a corresponding cleaning scheme based on the optical characteristics of the fouling layer. When hard scale that is difficult to remove with the purging airflow is detected, the ultrasonic cleaning program is automatically triggered and an equipment switching command is generated, while maintenance suggestions are pushed to the operation interface.

[0116] In this embodiment, online diagnosis of dirt types is achieved through spectral feature analysis. The optimal cleaning mode is matched to different dirt characteristics, avoiding the problem of repeated maintenance caused by a single cleaning method being ineffective for specific dirt. Through the above technical solution, this application effectively solves the problem of cleaning strategy mismatch caused by the inability of traditional monitoring systems to identify probe contamination types. Accurately determining the nature of dirt through optical feature analysis and switching cleaning modes accordingly avoids equipment damage caused by high-intensity cleaning and prevents measurement errors caused by inefficient cleaning, significantly improving the accuracy of maintenance operations and the stability of system operation.

[0117] In one embodiment of this specification, this application further proposes an online monitoring system 100 for flue gas desulfurization tower slurry. The host computer module 180 is also used to perform pattern recognition on the changes in spectral ratios: establishing a fouling spectral fingerprint database based on historical data, the fingerprint database containing scattering / absorption ratio change patterns corresponding to various fouling types such as gypsum layer, carbon deposit layer, and oil stain layer; matching the real-time monitored ratio change trend with the patterns in the fingerprint database; if the match is successful, while suggesting switching to ultrasonic cleaning mode, displaying the diagnosed specific fouling type on the operation interface, and recommending an ultrasonic cleaning frequency and duration that match the fouling type.

[0118] In this embodiment, the dirt spectral fingerprint database refers to a database accumulated through historical operational data. Specifically, it can be implemented by using a spectrometer to collect optical characteristic data of different dirt layers, storing the variation patterns of the scattering and absorption spectral ratios corresponding to typical dirt layers such as gypsum layers and carbon deposits. The scattering / absorption ratio variation pattern refers to the differences in optical response of different substances in the near-infrared band, which can be achieved by measuring the optical parameters of dirt layer samples using a Fourier transform spectrometer. Its variation trend reflects the physicochemical properties of the dirt. Pattern matching refers to calculating the similarity between real-time monitoring data and historical features, which can be implemented using dynamic time warping algorithms or correlation coefficient analysis methods, used to identify the correlation between the current dirt type and known dirt.

[0119] Specifically, when the multi-component analysis module 140 detects a trend change in the ratio of scattering spectral intensity to absorption spectral intensity, the host computer module 180 compares the real-time data with pre-stored patterns in the dirt spectral fingerprint database. For example, gypsum layer dirt exhibits a linear change with a negative slope, where scattering intensity continuously increases while absorption intensity decreases, within a specific wavelength range; carbon deposits, on the other hand, show a step-like increase in scattering intensity. After identifying the current dirt type through a dynamic matching algorithm, the operation interface synchronously displays the diagnostic results and pushes the corresponding cleaning parameter combination. For example, high-frequency short-time ultrasonic cleaning is recommended for gypsum layers, while low-frequency long-time cleaning is suitable for oily layers.

[0120] In some specific implementations, the establishment of a dirt spectral fingerprint database can be achieved by collecting spectral data of deposits on the probe surface under different operating conditions, combining this data with offline chemical analysis to determine the dirt composition, and then constructing a database containing time-series variation characteristics. During the matching process, a sliding window algorithm can be used to extract trend features from real-time data, sort them by similarity with patterns in the fingerprint database, and select the dirt type with the highest matching degree as the diagnostic result.

[0121] In this embodiment, spectral ratio pattern recognition can accurately determine the type of dirt, thereby adapting differentiated cleaning parameters. Through the above technical solution, this application solves the problem of low cleaning efficiency in existing technologies due to the inability to identify specific dirt types. By establishing a dirt characteristic database and real-time pattern matching, the type of deposits on the probe surface can be accurately identified, and targeted cleaning solutions can be recommended. For example, gypsum layers require high-frequency ultrasonic waves to disrupt the crystal structure, while oily layers require extended cleaning time to dissolve organic matter, thereby significantly improving cleaning effectiveness and reducing maintenance frequency.

[0122] In one embodiment of this specification, the purging module 170 is used to control the frequency, intensity, and duration of the pulsed airflow to intermittently purge the surfaces of the pH probe, density probe, and multi-component analysis probe to achieve cleaning and avoid interference with the measurement environment.

[0123] In some other embodiments of this specification, the host computer module 180 is used for source tracing analysis of abnormal slurry states, including establishing a time-series evolution map of state clusters, recording and analyzing the time-series transformation relationships between different slurry state clusters, and constructing a dynamic evolution map with state clusters as nodes and transformation probabilities and conditions as edges; identifying and marking abnormal transformation paths, marking key abnormal transformation paths that ultimately require shutdown for cleaning or performance degradation in the evolution map based on historical data; real-time source tracing and root cause inference, when the slurry state is detected to enter a high-risk cluster or is about to trigger a cleaning alarm, automatically tracing back the recent state cluster transformation sequence and matching the sequence with the abnormal transformation path library; if an abnormal path is matched, highlighting the initial root cause cluster and its inducing factors most likely to cause the current state on the interface based on historical correlation information; the host computer module 180 provides a source tracing report to guide operators to make adjustments from the source of the process.

[0124] In this embodiment, the state cluster time-series evolution map refers to a network structure formed by modeling the slurry state transition patterns using machine learning algorithms. Specifically, it can be implemented using a Markov chain model combined with graph database technology to reveal the potential patterns of slurry state evolution under different operating conditions. Anomaly transition paths refer to state transition sequences with high fault correlation verified by historical data. Specifically, they can be identified using association rule mining algorithms to quickly locate high-risk operating condition evolution patterns. Real-time source tracing and root cause inference refers to a state tracking mechanism based on real-time data streams. Specifically, it can be implemented using time series pattern matching algorithms to accurately identify the origin of problems in complex operating condition changes.

[0125] Specifically, the system continuously records the transition process of slurry state clusters and constructs a state transition network using a probabilistic model. When an abnormal state is detected, it automatically traces back the complete transition chain before entering the current state and compares it with a pre-stored high-risk path library for similarity. Upon successful matching, the system triggers a root cause analysis module, combining fault cause data from similar paths in a historical case library to locate the earliest state cluster node that caused the anomaly and its corresponding process parameter combination. For example, when the system identifies a persistent "high chloride ion - low pH" state leading to gypsum supersaturation, the interface will automatically display the complete evolution process of this abnormal path and corresponding absorption tower slurry supply adjustment suggestions.

[0126] In this embodiment, by constructing a dynamic evolution map and anomaly path library, automated tracking and pattern recognition of the slurry state deterioration process are achieved, overcoming the technical bottlenecks of low efficiency and high misjudgment rate in manual analysis. Existing technologies can only issue alarms based on current measurements, while this solution can trace back to the source of process fluctuations several hours ago, significantly improving the timeliness and accuracy of fault diagnosis. Through the above technical solution, this application achieves intelligent source tracing analysis of abnormal states in desulfurization slurry, enabling operators to quickly locate initial deviations in process control. By revealing the root causes of abnormal states, the system guides adjustments to key process parameters such as the absorber slurry supply rate and oxidation air volume, effectively avoiding the superficial treatment of simply cleaning probes or temporarily adjusting parameters. This technical approach reduces the number of unplanned shutdowns caused by misjudgments, improving the operational stability of the desulfurization system and the quality control capability of gypsum by-products.

[0127] In some other embodiments of this specification, the host computer module 180 is also used for source tracing analysis of abnormal slurry states: establishing a temporal evolution map of state clusters, the system records and analyzes the temporal transition relationship between different slurry state clusters, and constructs a dynamic evolution map with state clusters as nodes and transition probabilities and conditions as edges; identifying and marking abnormal transition paths, in the evolution map, marking the key abnormal transition paths that ultimately require shutdown for cleaning or performance deterioration based on historical data; real-time source tracing and root cause inference, when the system detects that the slurry state has entered a high-risk cluster or is about to trigger a cleaning alarm, it automatically traces back its recent state cluster transition sequence; matching the transition sequence with the abnormal transition path library; if an abnormal path is matched, the initial root cause cluster and its inducing factors most likely to cause the current state are highlighted on the host computer interface based on the historical association information of the path; the host computer module 180 provides a source tracing report to guide operators to make adjustments from the source of the process, rather than just dealing with the current state.

[0128] In this embodiment, the state cluster time-series evolution map refers to a network structure formed by analyzing the transformation patterns over time by recording the clustering results of different slurry component combinations and operating conditions. Specifically, it can be implemented by storing each state cluster node and its transformation edge weights in a graph database to reveal the potential patterns of slurry state changes. Anomaly transformation paths refer to high-probability paths from normal state clusters to state clusters requiring shutdown for cleaning or experiencing performance degradation. Specifically, this can be achieved by annotating the state transformation sequences before historical alarm events to identify high-risk operating condition evolution trends in advance. Real-time source tracing and root cause inference refers to tracing the initial causes leading to the current slurry state based on its position in the evolution map. Specifically, it can be implemented by matching a graph traversal algorithm with a historical anomaly path database to accurately locate key nodes in process adjustments.

[0129] In this embodiment, specifically, the host computer module 180 uses a clustering algorithm to divide historical and real-time slurry data into multiple state clusters, each cluster representing a specific combination of components and operating conditions. The temporal evolution map of the state clusters forms a dynamic network structure by recording the transition frequencies and conditions between each cluster. When the current slurry is detected to have entered a high-risk state cluster, the system automatically traces back its state transition sequence before entering that cluster and matches it with a predefined abnormal path library. If the transition sequence is found to highly match a certain historical abnormal path, the initial cause is highlighted on the host computer interface based on the root cause cluster information recorded in that path. For example, when the system detects an abnormal increase in the current gypsum supersaturation, it may trace back to a process problem where the slurry entered a high chloride ion-low pH state cluster three hours ago and was not adjusted in time, thereby guiding operators to prioritize adjusting the amount of limestone added or the amount of wastewater discharged.

[0130] In this embodiment, by constructing a state cluster evolution map, the evolution path of the slurry state can be dynamically tracked, and the root causes leading to performance degradation can be identified. For example, traditional methods only trigger a cleaning command when the probe measurement is inaccurate, while this solution can further analyze the abnormal transformation process of the slurry components before the inaccuracy, guiding adjustments at the source of the process to avoid repeated contamination. Through the above technical solution, this application can achieve dynamic tracing of abnormal slurry states, reducing misjudgments caused by fluctuations in local parameters. Operators can adjust the desulfurization process parameters in a targeted manner based on the root cause cluster information prompted by the system, avoiding the increased maintenance costs caused by relying solely on frequent probe cleaning. At the same time, by identifying abnormal transformation paths, early warnings can be given before the slurry state deteriorates, improving the continuity and stability of system operation.

[0131] like Figure 2 As shown, the slurry flows out from the bottom of the absorption tower 1, and after the flow rate is regulated by the control valve 2, it enters the flow tank 3. In the flow tank 3, it is detected in real time by the pH meter probe 6, the density meter probe 5, and the fiber optic probe 4 in sequence. The detection signals are processed by the pH meter processor 12, the density processor 13, and the spectral detector 14 and sent to the circuit board 11. The results are displayed on the display screen 19. The motor 7 on the system housing 8 drives the defoaming device. The defoaming device switch 16, the purging device switch 17, and the detection switch 18 control the start and stop of defoaming, purging, and detection, respectively. The heat dissipation window 9 dissipates heat from the equipment, and the power supply 10 supplies power to the entire system through the main power switch 15.

[0132] The specific process is as follows: Open control valve 2 to allow the slurry from the absorption tower to enter the flow tank 3. Turn on the main power switch 15 of the system. First, defoam the slurry to be tested by turning on the defoaming device switch 16. After the motor 7 of the defoaming device starts, the design combines the rotating device with the stationary chamber. The stirring shaft drives the rotating device to rotate, causing the gas carrying foam to undergo centrifugal separation in the rotating device. Utilizing the relative motion between the rotating device and the stationary cylinder, further gas-liquid separation is achieved, and finally, the foam-free liquid is discharged, effectively reducing the interference of foam on the probe measurement. The defoamed slurry passes sequentially through the pH meter probe 6, the density meter probe 5, and the fiber optic probe 4. Then, turn on the detection switch 18.

[0133] The working process of the pH online analysis subsystem mainly involves the pH meter electrode being directly inserted into the flow cell 3 of the absorption tower. Its basic principle is to use a glass membrane with specific selectivity for H+ to isolate the internal electrolyte from the external fermentation broth, thereby generating a potential on the electrode. This electrode potential is linearly related to the logarithm of the hydrogen ion concentration in the fermentation broth. The analysis is centrally monitored by the pH meter processor 12, and the results are displayed on the screen 19. To prevent scaling on the pH meter electrode surface, which could lead to decreased sensitivity, the purging device switch 17 is periodically turned on to flush the electrode. The purging device starts the motor 7, which uses compressed air to generate pulsed airflow, intermittently blowing it onto the surface of the pH meter probe 6. When the pulsed airflow impacts the liquid, it creates a localized pressure change on the surface of the pH meter probe 6. By controlling the frequency, intensity, and duration of the pulses, the surface of the pH meter probe 6 can be effectively cleaned while avoiding excessive gas blowing that could unnecessarily affect the measurement environment.

[0134] The workflow of the online density analysis subsystem mainly involves the density meter probe 5, typically made of piezoelectric ceramic, which converts electrical energy into ultrasonic mechanical vibrations and emits them. It can also receive ultrasonic signals and convert them back into electrical signals. The density processor 13 converts the analog signals into digital signals for analysis and processing, enabling data storage, display, and transmission.

[0135] The workflow of the multi-component online analysis subsystem mainly involves the fiber optic probe 4 acquiring the spectral information of the sample via transmission. A low-pressure halogen tungsten lamp with strong light and high energy is used as the light source. After emission, the light is collimated to ensure parallelism of the optical path. The beam entering the monochromator from the fiber optic cable is collimated by a concave mirror before the irradiation grating of the spectral detector 14. The grating spectrum emitted from the grating is converged by another concave mirror, focusing the grating spectrum onto the focal plane. The array detector pixels are placed on this focal plane, and finally, the indium gallium arsenide detector converts the dispersed spectrum into an electrical signal. The main workflow for establishing the multi-component analysis method includes: first, acquiring content and spectral data. Slurry samples are collected at different time points, and the contents of suspended solids, chloride ions, and sulfate ions are determined using a diaphragm vacuum pump and spectrophotometer according to standard methods, serving as the true values ​​for the model. Simultaneously, near-infrared spectral data of the slurry are acquired. Then, a random partitioning method is used to divide the samples into training and test sets. An improved Transformer model is then constructed. The training set's true values ​​and spectral data are input into the model for training. Key parameters such as model dimensionality, the number of heads in the multi-head self-attention network, the number of encoder layers, and the hidden layer dimension of the feedforward network are continuously optimized to reduce the RMSE (Real-Time Sequence) between the model's predicted and true values ​​until the requirements are met. Finally, the spectra of the test set are input into the multi-component model to obtain the model's prediction accuracy and robustness. A single-layer encoder, including self-attention and feedforward networks, is constructed using a Transformer. Multiple encoders are then stacked to extract more complex features, and the encoder output is mapped to the dimension of the target variable through a fully connected layer. In quantitative tasks, the encoder output is typically mapped to the target value through a fully connected layer. Finally, the model is trained. The training process for the Transformer model is similar to other neural networks. In each epoch, the training and validation sets are processed separately. During the training phase, forward propagation, loss calculation, backpropagation, and optimizer steps are performed. During the validation phase, the model's RMSE performance is evaluated. This simplifies the process of traditional modeling methods. SHAP analysis is a powerful tool that can deeply reveal the specific contribution of each feature to the model's prediction results, greatly enhancing the model's interpretability. SHAP allows us to understand how the model makes predictions based on input features, which is crucial for improving the model's credibility and transparency. Therefore, the improved Transformer model utilizes SHAP analysis to obtain important parameters for optimal model building. After model construction, the real-time spectral information of the slurry is input into the model, and the software processes it to obtain real-time information on the suspended solids, chloride ions, and sulfate ions in the slurry, which is then displayed on screen 19.

[0136] The density meter probe 5 and the fiber optic probe 4 require regular cleaning of their surfaces to prevent clogging and contamination.

[0137] Based on the same general inventive concept, this invention also protects an online monitoring method for flue gas desulfurization tower slurry, such as... Figure 3 As shown, Figure 3 This is a schematic flowchart of the online monitoring method for flue gas desulfurization tower slurry provided in an embodiment of the present invention. The online monitoring method for flue gas desulfurization tower slurry provided by the present invention will be described below. The online monitoring method for flue gas desulfurization tower slurry described below can be referred to in correspondence with the online monitoring system for flue gas desulfurization tower slurry described above. The online monitoring method for flue gas desulfurization tower slurry can be applied to the online monitoring system for flue gas desulfurization tower slurry in any of the above embodiments.

[0138] Online monitoring system methods for flue gas desulfurization tower slurry include:

[0139] Step 301: Defoaming treatment is performed on the slurry from the bottom of the absorption tower using a defoaming module;

[0140] Step 302: Measure the density of the slurry in real time using the density probe via the density analysis module;

[0141] Step 303: Measure the pH value of the slurry in real time using a pH probe via a pH analysis module;

[0142] Step 304: Analyze the content of suspended solids, chloride ions, and sulfate ions in the slurry in real time using the multi-component analysis module and multi-component analysis probe;

[0143] Step 305: Contain the slurry through the flow cell module and install the pH probe, density probe and multi-component analysis probe;

[0144] Step 306: Control the flow rate and volume of the slurry flowing from the bottom of the absorption tower into the flow pool module through the control valve module;

[0145] Step 307: The purging module generates pulsed airflow using compressed air to intermittently purge the surfaces of the pH probe, density probe, and multi-component analysis probe.

[0146] Step 308: Display the measurement data of the pH analysis module, density analysis module, and multi-component analysis module through the host computer module.

[0147] Figure 4 This is a schematic diagram of the structure of the electronic device provided in an embodiment of the present invention.

[0148] like Figure 4As shown, the electronic device may include a processor 410, a communication interface 420, a memory 430, and a communication bus 440. The processor 410, communication interface 420, and memory 430 communicate with each other via the communication bus 440. The processor 410 can call logical instructions from the memory 430 to execute an online monitoring method for the flue gas desulfurization tower slurry.

[0149] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0150] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the online monitoring method for flue gas desulfurization tower slurry provided by the above methods.

[0151] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the online monitoring method for flue gas desulfurization tower slurry provided by the methods described above.

[0152] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0153] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0154] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An online monitoring system for flue gas desulfurization tower slurry, characterized in that, include: The defoaming module is used to defoam the slurry from the bottom of the absorption tower. The density analysis module is used to measure the density of the slurry in real time using a density probe; The pH analysis module is used to measure the pH value of the slurry in real time using a pH probe; The multi-component analysis module is used to analyze the content of suspended solids, chloride ions and sulfate ions in slurry in real time through a multi-component analysis probe. A flow cell module is used to contain the slurry and to install the pH probe, the density probe, and the multi-component analysis probe. The control valve module is used to control the flow rate and volume of the slurry flowing from the bottom of the absorption tower into the flow pool module; The purging module is used to generate pulsed airflow using compressed air to intermittently purge the surfaces of the pH probe, the density probe, and the multi-component analysis probe; The host computer module is used to display the measurement data from the pH analysis module, the density analysis module, and the multi-component analysis module.

2. The online monitoring system for flue gas desulfurization tower slurry according to claim 1, characterized in that, The defoaming module includes a rotating separation unit and a stationary cavity unit; The rotating separation unit is located inside the stationary cavity unit and forms a separation gap with the stationary cavity unit; When the rotary separation unit rotates, the gas carrying foam in the slurry is separated into gas and liquid under the action of centrifugal force, so that the separated liquid is discharged through the separation gap and the separated gas is discharged through the exhaust port at the top of the stationary cavity unit.

3. The online monitoring system for flue gas desulfurization tower slurry according to claim 1, characterized in that, The multi-component analysis module employs a Transformer model, which is trained in the following manner: Using training set sample data, the input spectral features were sorted through SHAP analysis to screen out the key wavelength variables that contribute the most to the prediction of suspended solids, chloride ions, and sulfate content. Based on the key wavelength variables, the Transformer model is trained using the training set sample data, and the weight distribution of the query, key, and value matrices of the self-attention mechanism in the Transformer model is dynamically adjusted to enhance the ability to perceive key spectral features. The root mean square error (RMSE) during training is monitored using validation set data. If the RMS error of the validation set data does not decrease for several consecutive training cycles, training is terminated.

4. The online monitoring system for flue gas desulfurization tower slurry according to claim 2, characterized in that, The defoaming module is also used for; The operating current of the drive motor of the rotary separation unit is acquired in real time. The operating current is compared with a preset current threshold range; When the operating current is continuously higher than the upper limit of the current threshold range, the drive motor is controlled to increase its speed; When the operating current remains below the lower limit of the current threshold range, the drive motor is controlled to reduce its speed. The current threshold range is predetermined by the following method: Measure the motor operating current corresponding to slurry with different foaming levels, and calibrate the current threshold range required for defoaming.

5. The online monitoring system for flue gas desulfurization tower slurry according to claim 4, characterized in that, The defoaming module is also used for: The operating current of the drive motor is continuously monitored, and its average value and trend slope within a preset sliding time window are calculated. When the average value is consistently higher than the upper limit of the current threshold range and the trend slope is positive, it is determined that the foaming degree of the slurry is intensifying, and the drive motor is controlled to increase its speed in a stepwise manner.

6. The online monitoring system for flue gas desulfurization tower slurry according to claim 1, characterized in that, The host computer module is also used for: Cluster analysis was performed on historical and real-time collected slurry multi-component data to obtain different slurry state clusters; each slurry state cluster corresponds to a component combination and operating condition. Within each slurry state cluster, historical data records that triggered cleaning alarms were statistically analyzed to identify the most significant factors causing probe contamination or measurement inaccuracies within that slurry state cluster, which were then used as cleaning hotspots for that slurry state cluster. Based on the cleaning hotspots, dynamically configure differentiated performance evaluation thresholds and cleaning trigger thresholds for different slurry state clusters; Based on each slurry state cluster and its corresponding cleaning hotspot, a targeted cleaning strategy is generated to match it. The current slurry state is matched with the clustering model in real time, and the corresponding cleaning strategy for that state cluster is executed.

7. The online monitoring system for flue gas desulfurization tower slurry according to claim 1, characterized in that, The purging module is also used for: The first stability index of pH measurement data was coupled with the second stability index of chloride ion content measurement data for analysis. When the product of the fluctuation coefficients of the first stability index and the second stability index exceeds the preset coupled fluctuation threshold, and the fluctuation coefficients of each index do not reach their respective preset independent thresholds, probe contamination is determined to exist, and preventive purging is initiated. The first stability index and the second stability index are both calculated by the ratio of the standard deviation to the mean of their respective consecutive measurement periods.

8. The online monitoring system for flue gas desulfurization tower slurry according to claim 1, characterized in that, The multi-component analysis module is also used for: Real-time monitoring of the ratio of scattering spectral intensity to absorption spectral intensity in the near-infrared spectrum; When this ratio shows a trend change, it is determined that a dirt layer has formed on the probe surface; The host computer module will suggest that the operator switch to the maintenance mode based on ultrasonic cleaning and automatically generate corresponding control commands.

9. The online monitoring system for flue gas desulfurization tower slurry according to claim 8, characterized in that, The host computer module is also used to perform pattern recognition on the changes in the spectral ratio: Establish a dirt spectral fingerprint database based on historical data; wherein the fingerprint database contains the variation patterns of the ratio between scattered light intensity and absorbed light intensity corresponding to various dirt types; The real-time monitored trend of ratio changes is matched with the patterns in the fingerprint database; If the match is successful, the system will recommend switching to ultrasonic cleaning mode, display the specific type of dirt diagnosed on the interface, and recommend an ultrasonic cleaning frequency and duration that match the type of dirt.

10. The online monitoring system for flue gas desulfurization tower slurry according to claim 1, characterized in that, The purging module is used for: The frequency, intensity, and duration of the pulsed airflow are controlled to intermittently purge the surfaces of the pH probe, the density probe, and the multi-component analysis probe, thereby achieving cleanliness and avoiding interference with the measurement environment.

Images