A method, system, storage medium and program product for blast control of a wet desulphurization system

Abstract

The application provides a blast control method and system of a wet desulfurization system, a storage medium and a program product, relates to the technical field of wet desulfurization control, and the method comprises the following steps: multiple oxidation state monitoring points are arranged in a slurry pool of an absorption tower of the wet desulfurization system, slurry oxidation state parameters of each oxidation state monitoring point are collected in real time, operation condition parameters of the wet desulfurization system and fan control parameters of a magnetic suspension oxidation fan are obtained; during the operation of the wet desulfurization system, the fan control parameters, the slurry oxidation state parameters and the operation condition parameters are input into a prediction model, predicted slurry state parameters output by the prediction model are obtained, and the predicted slurry state parameters and the operation condition parameters are used to determine predicted air volume; a deviation feedback model is used to perform air volume correction operation on the slurry oxidation state parameters, and corrected air volume is obtained; target control air volume is generated according to the predicted air volume and the corrected air volume, and operation regulation and control operation is performed on the magnetic suspension oxidation fan by using the target control air volume.

Description

Technical Field

[0001] This application relates to the field of wet desulfurization control technology, and in particular to a blower control method, system, storage medium and program product for a wet desulfurization system. Background Technology

[0002] With increasingly stringent environmental protection requirements and continuously improving energy efficiency standards, wet desulfurization systems, as the core component of flue gas treatment in thermal power plants, have seen their operational efficiency and energy consumption control become key factors affecting the economic viability and environmental compliance of power plants. Particularly in the control of oxidation fans, precisely regulating the oxidation air volume to ensure the complete oxidation of calcium sulfite to gypsum while avoiding energy waste caused by excessive air supply has become an important issue for optimizing the operation of wet desulfurization systems.

[0003] In related technologies, wet desulfurization systems typically use Roots blowers, single-stage centrifugal blowers, or multi-stage high-speed centrifugal blowers as oxidation fans, supplying air to the absorber slurry pool via branch lines from a main pipe. Operational control primarily relies on single-point installed pH and density measuring instruments. Laboratory tests are used to assess the oxidation level of gypsum, and the results are reported periodically to guide airflow adjustments. When test results indicate insufficient calcium sulfite oxidation, operators manually increase the blower output; once the gypsum quality meets standards, they attempt to reduce the blower output to save energy. To ensure oxidation effectiveness under various operating conditions, this control method typically sets a large airflow margin, with the blower operating in an over-supply state for extended periods.

[0004] However, when using the above control method, because the operators rely solely on single-point sampling and delayed laboratory test results, they cannot accurately grasp the real-time oxidation demand of the absorption tower slurry pool. In order to avoid insufficient oxidation affecting the gypsum quality, they can only adopt a conservative strategy to maintain a high air supply, which in turn leads to poor air supply control effect of the oxidation fan in the wet desulfurization system in related technologies. Summary of the Invention

[0005] This application provides a blower control method, system, storage medium, and program product for a wet desulfurization system, which is used to improve the air supply control effect of the oxidation blower in the wet desulfurization system.

[0006] In a first aspect, this application provides a blower control method for a wet desulfurization system, applied to the aforementioned blower control system. The method includes: setting multiple oxidation state monitoring points in the slurry pool of the absorber tower of the wet desulfurization system to collect slurry oxidation state parameters at each monitoring point in real time, and obtaining the operating condition parameters of the wet desulfurization system and the blower control parameters of the magnetic levitation oxidation blower; during the operation of the wet desulfurization system, inputting the blower control parameters, slurry oxidation state parameters, and operating condition parameters into a DTW-CNN-LSTM prediction model to obtain predicted slurry state parameters output by the DTW-CNN-LSTM prediction model after performing a target airflow prediction operation on the blower control parameters, slurry oxidation state parameters, and operating condition parameters; determining the predicted airflow based on the predicted slurry state parameters and operating condition parameters; performing an airflow correction operation on the slurry oxidation state parameters using a DO-ORP deviation feedback model to obtain a corrected airflow; generating a target control airflow based on the predicted airflow and the corrected airflow; and using the target control airflow to perform operation control operation on the magnetic levitation oxidation blower.

[0007] By adopting the above technical solution, multiple oxidation state monitoring points set in the slurry pool of the absorption tower can form a spatially distributed monitoring network. The slurry oxidation state parameters collected in real time by each monitoring point can comprehensively reflect the differences in oxidation state at different locations in the slurry pool. The synchronously acquired operating condition parameters can characterize the real-time operating load of the wet desulfurization system, and the acquired fan control parameters can characterize the current operating status of the magnetic levitation oxidation fan. After inputting the fan control parameters, slurry oxidation state parameters, and operating condition parameters into the DTW-CNN-LSTM prediction model, the model can comprehensively consider the coupling relationship of the three factors: fan operating status, actual slurry oxidation state, and system operating load, and output the prediction result of the future slurry state. Compared with the method of prediction based solely on operating condition parameters, multi-source data fusion prediction can more accurately capture the dynamic response characteristics of the system. When determining the predicted air volume based on the predicted slurry state parameters and operating condition parameters, the predicted slurry state parameters are compared with the preset slurry oxidation target parameters. This allows for a more precise determination of the airflow required to achieve the ideal oxidation effect. The target parameters for slurry oxidation include a preset dissolved oxygen threshold and a preset oxidation-reduction potential threshold. These thresholds are target control values ​​for the slurry oxidation state, pre-set according to the desulfurization process requirements. This predicted airflow, as a feedforward control quantity, can respond in advance to changes in operating conditions. When the DO-ORP deviation feedback model performs airflow correction operations on the slurry oxidation state parameters, it determines the corrected airflow by comparing the deviation between the slurry oxidation state parameters and the target parameters. This corrected airflow, as a feedback control quantity, can compensate for prediction errors. The target control airflow generated by combining the predicted and corrected airflow comprehensively considers both feedforward prediction and feedback correction factors. When performing operation control operations on the magnetic levitation oxidation fan, the fan's high-speed response characteristics enable the actual output airflow to quickly and accurately reach the target control airflow, thereby achieving precise supply of oxidation airflow. This ensures sufficient slurry oxidation while avoiding energy waste caused by excessive air supply. This solves the technical problem of poor control effect of the oxidation fan air supply in wet desulfurization systems, and achieves the technical effect of improving the control effect of the oxidation fan air supply in wet desulfurization systems.

[0008] Secondly, embodiments of this application provide a blower control system, which includes: one or more processors and a memory; the memory is coupled to one or more processors, and the memory is used to store computer program code, the computer program code including computer instructions, and the one or more processors call the computer instructions to cause the blower control system to perform the method described in the first aspect and any possible implementation thereof.

[0009] Thirdly, embodiments of this application provide a computer program product containing instructions that, when the computer program product is run on a blower control system, cause the blower control system to perform the method described in the first aspect and any possible implementation thereof.

[0010] Fourthly, embodiments of this application provide a computer-readable storage medium including program instructions that, when executed on a blower control system, cause the blower control system to perform the method described in the first aspect and any possible implementation thereof. Attached Figure Description

[0011] Figure 1 This is a flowchart illustrating the blower control method of the wet desulfurization system in an embodiment of this application;

[0012] Figure 2 This is a schematic diagram of the deployment of a multi-point sampling and detection device for the slurry pool of the absorption tower in this application embodiment;

[0013] Figure 3 This is a schematic diagram of a structure of an adjustable front guide vane energy-saving integrated magnetic levitation oxidation fan in the embodiments of this application;

[0014] Figure 4 This is a schematic diagram of the architecture of the DTW-CNN-LSTM prediction model in the embodiments of this application. Figure 1 ;

[0015] Figure 5 This is a schematic diagram of the architecture of the DTW-CNN-LSTM prediction model in the embodiments of this application. Figure 2 ;

[0016] Figure 6 This is a schematic diagram illustrating the principle of DTW timing sequence alignment in an embodiment of this application;

[0017] Figure 7 This is a schematic diagram of a CNN local feature extraction module in an embodiment of this application;

[0018] Figure 8 This is a schematic diagram of a structure of the LSTM long-term dependency capture module in an embodiment of this application;

[0019] Figure 9 This is a schematic diagram of the internal structure of an LSTM cell unit in an embodiment of this application;

[0020] Figure 10 This is a schematic diagram of the physical device structure of a blower control system in an embodiment of this application. Detailed Implementation

[0021] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification and appended claims of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.

[0022] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0023] This application provides a blower control method for a wet desulfurization system, see reference. Figure 1 , Figure 1 This is a schematic flowchart of a blower control method for a wet desulfurization system in an embodiment of this application, including the following steps:

[0024] Step S101: Set up multiple oxidation state monitoring points in the slurry pool of the absorber tower of the wet desulfurization system to collect the slurry oxidation state parameters of each oxidation state monitoring point in real time, and obtain the operating condition parameters of the wet desulfurization system and the fan control parameters of the magnetic levitation oxidation fan of the wet desulfurization system.

[0025] Step S102: During the operation of the wet desulfurization system, the fan control parameters, slurry oxidation state parameters, and operating condition parameters are input into the DTW-CNN-LSTM prediction model to obtain the predicted slurry state parameters output by the DTW-CNN-LSTM prediction model after performing the target air volume prediction operation on the fan control parameters, slurry oxidation state parameters, and operating condition parameters, and the predicted air volume is determined based on the predicted slurry state parameters and operating condition parameters.

[0026] Step S103: Use the DO-ORP deviation feedback model to perform air volume correction operation on the slurry oxidation state parameters to obtain the corrected air volume;

[0027] Step S104: Generate a target control air volume based on the predicted air volume and the corrected air volume, and use the target control air volume to perform operation control operation on the magnetic levitation oxidation fan.

[0028] Among them, wet desulfurization system refers to an environmental protection treatment system that uses limestone-gypsum method to remove sulfur dioxide from flue gas from coal-fired power plants, such as desulfurization devices used to purify flue gas in thermal power plants; absorber slurry pool refers to the container used in wet desulfurization system to store and circulate desulfurization slurry, in which the slurry reacts chemically with sulfur dioxide in the flue gas; oxidation state monitoring point refers to the measurement location set in absorber slurry pool for real-time monitoring of slurry oxidation state, such as sensor installation points arranged at different depths and orientations in the slurry pool; slurry oxidation state parameters represent the measurement data of the degree to which calcium sulfite in the slurry is oxidized to calcium sulfate, including but not limited to dissolved oxygen value, oxidation-reduction potential value, etc.; operation... The operating parameters represent real-time data reflecting the current operating status of the wet desulfurization system, such as generator load, flue gas volume, and sulfur dioxide concentration; the fan control parameters represent control data reflecting the current operating status of the magnetic levitation oxidation fan, including motor speed and guide vane opening; the DTW-CNN-LSTM prediction model refers to a model built based on machine learning algorithms to predict the slurry oxidation state, such as a neural network model using a DTW-CNN-LSTM fusion architecture; the predicted slurry state parameters represent the predicted values ​​of the future slurry oxidation state output by the DTW-CNN-LSTM prediction model based on the fan control parameters, slurry oxidation state parameters, and operating condition parameters.

[0029] Among them, the slurry oxidation target parameter refers to the target control value of the slurry oxidation state preset according to the desulfurization process requirements, including the preset dissolved oxygen threshold and the preset oxidation-reduction potential threshold. The preset dissolved oxygen threshold is used to ensure that calcium sulfite is fully oxidized to calcium sulfate, and the preset oxidation-reduction potential threshold is used to characterize that the slurry has good oxidizing properties. The predicted air volume is used to represent the theoretically required air volume value determined based on the deviation between the predicted slurry state parameter and the slurry oxidation target parameter. The DO-ORP deviation feedback model is a control model that performs air volume correction operation based on the deviation between the slurry oxidation state parameter and the slurry oxidation target parameter. The corrected air volume represents the air volume adjustment amount determined based on the feedback of the actual oxidation state of the slurry. A positive value indicates that the air volume needs to be increased, and a negative value indicates that the air volume needs to be decreased. The target control air volume is the air volume value that the oxidation fan should finally output after combining the predicted air volume and the corrected air volume. The magnetic levitation oxidation fan is used to represent the oxidation fan equipment that adopts magnetic levitation bearing technology and high-speed motor direct drive technology, which has the characteristics of high efficiency, low noise, and oil-free lubrication.

[0030] In the above embodiment, a wet desulfurization system for a 300MW thermal power generating unit is used as an example. The wet desulfurization system includes an absorption tower, a slurry circulation system, and an oxidation air supply system. The oxidation air supply system includes a magnetic levitation oxidation fan and its associated oxidation air duct. The blower control system is used to perform operation control operations on the magnetic levitation oxidation fan in the oxidation air supply system. In specific implementation, six oxidation state monitoring points are set in the slurry pool of the absorption tower. These six monitoring points are distributed in different spatial locations in the slurry pool, with three monitoring points located in the lower layer area 1 meter from the bottom of the slurry pool, and the other three monitoring points located in the upper layer area 2.5 meters from the bottom of the slurry pool. Each monitoring point collects slurry oxidation state parameters in real time through an integrated measuring device, including dissolved oxygen value, oxidation-reduction potential value, pH value, and slurry density data. For example, at a certain moment, the dissolved oxygen values ​​collected at six monitoring points were 1.6 mg / L, 1.8 mg / L, 1.5 mg / L, 2.0 mg / L, 1.9 mg / L, and 1.7 mg / L, respectively, and the oxidation-reduction potential values ​​were -165 mV, -155 mV, -170 mV, -145 mV, -150 mV, and -160 mV, respectively. Simultaneously, the blower control system acquired the operating parameters of the wet desulfurization system. At the aforementioned moment, the acquired generator load was 250 MW, the flue gas flow rate into the absorber slurry pool was 950,000 Nm³ / h, the sulfur dioxide inlet concentration was 450 mg / Nm³, and the sulfur dioxide outlet concentration was 28 mg / Nm³. Furthermore, the blower control system also acquired the fan control parameters of the magnetic levitation oxidation fan. At the aforementioned moment, the acquired current motor speed was 18,000 rpm, and the current guide vane opening was 65%.

[0031] In the above embodiment, during the operation of the wet desulfurization system, the blower control system inputs the aforementioned blower control parameters, slurry oxidation state parameters, and operating condition parameters into a pre-trained DTW-CNN-LSTM prediction model. The DTW-CNN-LSTM prediction model performs a target airflow prediction operation on the blower control parameters, slurry oxidation state parameters, and operating condition parameters. Based on historical time-series data from the past 30 minutes, it analyzes and processes the data to output predicted slurry state parameters, including a predicted total dissolved oxygen value of 1.85 mg / L, a predicted total oxidation-reduction potential value of -152 mV, and a predicted slurry pH value of 5.6. The blower control system compares the predicted slurry state parameters with preset slurry oxidation target parameters, where the preset dissolved oxygen threshold is 2.0 mg / L, the preset oxidation-reduction potential threshold is -140 mV, and the preset pH threshold is 5.5. Based on the state deviation between the predicted slurry state parameters and the slurry oxidation target parameters, combined with the operating condition parameters, the specific process for determining the predicted airflow is as follows: Calculate the dissolved oxygen deviation = preset dissolved oxygen threshold - predicted total dissolved oxygen value, i.e. The redox potential deviation is calculated as follows: Preset redox potential threshold - Predicted total redox potential threshold. Then, based on the oxygen demand relationship of the desulfurization process, the theoretical oxygen demand is calculated based on the flue gas volume and sulfur dioxide inlet concentration in the operating parameters. The theoretical oxygen demand = flue gas volume × sulfur dioxide inlet concentration × oxidation reaction coefficient, i.e. The oxidation reaction coefficient of 0.5 represents the stoichiometric ratio for the complete oxidation of calcium sulfite to calcium sulfate; based on the theoretical oxygen demand... The baseline air volume is calculated based on the volume fraction of oxygen in the air. Baseline air volume = Theoretical oxygen demand / (Air density × Oxygen volume fraction × Oxygen utilization rate). The baseline air volume is adjusted and compensated based on dissolved oxygen deviation and redox potential deviation to predict the air volume. Then, considering the oxygen dissolution efficiency and mass transfer loss in the slurry, a mass transfer efficiency correction coefficient is introduced. ,get ,in The airflow compensation coefficient for dissolved oxygen deviation. The air volume compensation coefficient is the redox potential deviation, and the above compensation coefficient is obtained by calibration based on historical operating data.

[0032] In the above embodiment, the blower control system utilizes the DO-ORP deviation feedback model to perform airflow correction operations on the slurry oxidation state parameters. The DO-ORP deviation feedback model calculates a corrected airflow of +12 Nm³ / min (a positive value indicates a need to increase airflow) based on the deviation between the dissolved oxygen and oxidation-reduction potential values ​​at each monitoring point and a preset threshold. The blower control system generates a target control airflow based on the predicted and corrected airflow. By adding the predicted airflow of 185 Nm³ / min to the corrected airflow of +12 Nm³ / min, the target control airflow is obtained as 197 Nm³ / min. The wet desulfurization system includes an absorption tower, a slurry circulation system, and an oxidation air supply system, wherein the oxidation air supply system includes a magnetic levitation oxidation blower and its associated oxidation air ducts. The blower control system is used to perform operational control operations on the magnetic levitation oxidation blower in the oxidation air supply system.

[0033] Through the above steps, multiple oxidation state monitoring points set in the slurry pool of the absorption tower can form a spatially distributed monitoring network. The slurry oxidation state parameters collected in real time by each monitoring point can comprehensively reflect the differences in oxidation state at different locations within the slurry pool. The synchronously acquired operating condition parameters can characterize the real-time operating load of the wet desulfurization system, and the acquired fan control parameters can characterize the current operating state of the magnetic levitation oxidation fan. After inputting the fan control parameters, slurry oxidation state parameters, and operating condition parameters into the DTW-CNN-LSTM prediction model, the model can comprehensively consider the coupling relationship between the fan operating state, the actual oxidation state of the slurry, and the system operating load, and output the prediction result of the future slurry state. Compared with the method of prediction based solely on operating condition parameters, multi-source data fusion prediction can more accurately capture the dynamic response characteristics of the system. When determining the predicted air volume based on the predicted slurry state parameters and operating condition parameters, the predicted slurry state parameters are compared with the preset slurry oxidation target parameters. To more accurately determine the air volume value required to achieve the ideal oxidation effect, the target parameters for slurry oxidation include a preset dissolved oxygen threshold and a preset oxidation-reduction potential threshold. These thresholds are target control values ​​for the slurry oxidation state set in advance according to the requirements of the desulfurization process. The predicted air volume, as a feedforward control quantity, can respond to changes in operating conditions in advance. When the DO-ORP deviation feedback model performs air volume correction operations on the slurry oxidation state parameters, it determines the corrected air volume by comparing the deviation between the slurry oxidation state parameters and the target parameters. This corrected air volume, as a feedback control quantity, can compensate for prediction errors. The target control air volume generated by combining the predicted and corrected air volumes comprehensively considers both feedforward prediction and feedback correction factors. When performing operation control operations on the magnetic levitation oxidation fan, the high-speed response characteristics of the fan enable the actual output air volume to quickly and accurately reach the target control air volume, thereby achieving precise supply of oxidation air volume, ensuring sufficient slurry oxidation while avoiding energy waste caused by excessive air supply. This solves the technical problem of poor control effect of the oxidation fan air supply in wet desulfurization systems, and achieves the technical effect of improving the control effect of the oxidation fan air supply in wet desulfurization systems.

[0034] The entity performing the above steps can be a system with blower control capability, such as a blower control system, or a device with blower control capability, or a controller or processor in the device or system, or a standalone controller or processor, or other processing devices or processing units with similar processing functions, but is not limited to these.

[0035] In an optional embodiment, multiple oxidation state monitoring points are set in the slurry pool of the absorber tower of the wet desulfurization system to collect the slurry oxidation state parameters of each monitoring point in real time, and to obtain the operating condition parameters of the wet desulfurization system and the fan control parameters of the magnetic levitation oxidation fan of the wet desulfurization system. Specifically, this includes: arranging multiple integrated measuring devices at different depths and radial orientations in the absorber tower slurry pool to obtain multiple oxidation state monitoring points; using multiple integrated measuring devices to collect the dissolved oxygen value, oxidation-reduction potential value, and pH value of each oxidation state monitoring point in real time; determining the total dissolved oxygen value of the absorber tower slurry pool based on the dissolved oxygen value of each oxidation state monitoring point; and determining the total oxidation-reduction potential value of the absorber tower slurry pool based on the oxidation-reduction potential value of each oxidation state monitoring point. The total dissolved oxygen value is the average of the dissolved oxygen values ​​of multiple oxidation state monitoring points, and the total oxidation-reduction potential value is the average of the total dissolved oxygen values ​​of multiple oxidation state monitoring points. The average value of the oxidation-reduction potential (ORP) at each oxidation state monitoring point is calculated. A first deviation analysis is performed between the dissolved oxygen value at each oxidation state monitoring point and the total dissolved oxygen value to obtain the first dissolved oxygen deviation value for each oxidation state monitoring point. A second deviation analysis is performed between the ORP value at each oxidation state monitoring point and the total ORP value to obtain the first ORP deviation value for each oxidation state monitoring point. The total dissolved oxygen value, total ORP value, first dissolved oxygen deviation value, first ORP deviation value, and pH value are used as slurry oxidation state parameters. The sulfur dioxide inlet concentration is collected from the flue gas inlet of the absorption tower, and the sulfur dioxide outlet concentration is collected from the flue gas outlet of the absorption tower. The sulfur dioxide inlet concentration and sulfur dioxide outlet concentration are used as operating condition parameters. The current motor speed and current guide vane opening of the magnetic levitation oxidation fan are obtained and used as fan control parameters.

[0036] Among them, integrated measuring devices refer to integrated measuring equipment that integrates multiple sensors into the same measuring probe or measuring unit. For example, a dissolved oxygen sensor, a redox potential sensor, and a pH sensor are integrated into the same stainless steel probe. Different depth positions represent different height layers in the vertical direction from the bottom to the liquid surface of the absorber slurry pool, such as positions 0.5 meters, 1.5 meters, and 2.5 meters from the bottom of the slurry pool. Different radial orientations are used to represent different angular directions on the horizontal plane with the center of the absorber tower as the origin, such as four orientations: 0 degrees, 90 degrees, 180 degrees, and 270 degrees. Dissolved oxygen value refers to the concentration of dissolved oxygen in the slurry, measured in milligrams per liter (mg / L), reflecting the amount of oxygen available for oxidation reactions in the slurry; oxidation-reduction potential (ORP) value represents the oxidation-reduction capacity of the slurry, measured in millivolts (mV), with higher values ​​indicating stronger oxidizing properties; total dissolved oxygen value is the average of dissolved oxygen values ​​measured at all oxidation state monitoring points, used to characterize the overall dissolved oxygen level of the absorber slurry pool; total oxidation-reduction potential (ORP) value is the average of oxidation-reduction potential values ​​measured at all oxidation state monitoring points, used to characterize the overall oxidation-reduction state of the absorber slurry pool; first deviation analysis. This section describes a data processing operation that compares the dissolved oxygen values ​​at each monitoring point with the total dissolved oxygen value to obtain the deviation. The first dissolved oxygen deviation value represents the difference between the dissolved oxygen values ​​at each monitoring point and the total dissolved oxygen value, reflecting the unevenness of dissolved oxygen distribution within the slurry tank. The second deviation analysis section describes a data processing operation that compares the oxidation-reduction potential values ​​at each monitoring point with the total oxidation-reduction potential value to obtain the deviation. The first oxidation-reduction potential deviation value represents the difference between the oxidation-reduction potential values ​​at each monitoring point and the total oxidation-reduction potential value, reflecting the unevenness of oxidation state distribution within the slurry tank. The sulfur inlet concentration indicates the concentration of sulfur dioxide in the flue gas before entering the absorption tower, expressed in milligrams per standard cubic meter (mg / Nm³). The sulfur dioxide outlet concentration refers to the concentration of sulfur dioxide in the flue gas discharged from the absorption tower after desulfurization treatment, also expressed in milligrams per standard cubic meter (mg / Nm³). The current motor speed refers to the actual speed of the magnetic levitation oxidation fan motor at the current moment, expressed in revolutions per minute (rpm). The current guide vane opening refers to the current degree of opening of the guide vanes at the inlet of the magnetic levitation oxidation fan, expressed as a percentage, used to adjust the airflow direction and flow rate into the fan.

[0037] In the above embodiment, a medium-sized wet desulfurization system with an absorption tower slurry pool diameter of 15 meters and an effective liquid level height of 4 meters is used as an example. Six integrated measuring devices are arranged at different depths and radial orientations within the absorption tower slurry pool, specifically as follows: Vertically, the system is divided into two depth layers: a lower layer 1 meter from the bottom of the slurry pool and an upper layer 2.5 meters from the bottom. Horizontally, with the center of the absorption tower as the origin, one measuring device is arranged at each of the three radial orientations: 0 degrees, 120 degrees, and 240 degrees. Thus, a total of six oxidation state monitoring points are obtained at the three orientations of the lower layer and the three orientations of the upper layer. Each integrated measuring device includes a dissolved oxygen sensor (using the fluorescence quenching principle, range 0-10 mg / L, accuracy ±0.1 mg / L), a redox potential sensor (platinum electrode method, range -500~+500 mV, accuracy ±2 mV), and a pH sensor (glass electrode method, range 0-14, accuracy ±0.02). Each integrated measuring device collects dissolved oxygen, oxidation-reduction potential, and pH values ​​at the corresponding oxidation state monitoring point in real time at a frequency of once per minute.

[0038] In the above embodiment, at a certain sampling time, the collected data from the six oxidation state monitoring points are as follows: Monitoring point 1 (located at a lower 0-degree position): dissolved oxygen 1.6 mg / L, oxidation-reduction potential -165 mV, pH 5.2; Monitoring point 2 (located at a lower 120-degree position): dissolved oxygen 1.8 mg / L, oxidation-reduction potential -155 mV, pH 5.1; Monitoring point 3 (located at a lower 240-degree position): dissolved oxygen 1.5 mg / L, oxidation-reduction potential -165 mV, pH 5.1; The dissolved oxygen value at monitoring point 4 (located at 0 degrees above the surface) was -170 mV, and the pH value was 5.3. The dissolved oxygen value at monitoring point 5 (located at 120 degrees above the surface) was 1.9 mg / L, with a redox potential of -150 mV and a pH value of 5.1. The dissolved oxygen value at monitoring point 6 (located at 240 degrees above the surface) was 1.7 mg / L, with a redox potential of -160 mV and a pH value of 5.2. The total dissolved oxygen value in the absorption tower slurry tank was determined based on the dissolved oxygen values ​​at each oxidation state monitoring point. The collected data underwent a first deviation analysis, yielding the first dissolved oxygen deviation values ​​for each oxidation state monitoring point: First dissolved oxygen deviation value for monitoring point 1 = 1.6 - 1.75 = -0.15 mg / L; First dissolved oxygen deviation value for monitoring point 2 = 1.8 - 1.75 = +0.05 mg / L; First dissolved oxygen deviation value for monitoring point 3 = 1.5 - 1.75 = -0.25 mg / L; First dissolved oxygen deviation value for monitoring point 4 = 2.0 - 1.75 = +0.25 mg / L; First dissolved oxygen deviation value for monitoring point 5 = 1.9 - 1.75 = +0.15 mg / L; First dissolved oxygen deviation value for monitoring point 6 = 1.7 - 1.75 = -0.05 mg / L.

[0039] Similarly, in the above embodiments, the total oxidation-reduction potential value of the absorber slurry pool is determined based on the oxidation-reduction potential value of each oxidation state monitoring point. A second deviation analysis was performed on the collected data to obtain the first oxidation-reduction potential deviation values ​​for each oxidation state monitoring point: First oxidation-reduction potential deviation value for monitoring point 1 = -165 - (-157.5) = -7.5mV; First oxidation-reduction potential deviation value for monitoring point 2 = -155 - (-157.5) = +2.5mV; First oxidation-reduction potential deviation value for monitoring point 3 = -170 - (-157.5) = -12.5mV; First oxidation-reduction potential deviation value for monitoring point 4 = -145 - (-157.5) = +12.5mV; First oxidation-reduction potential deviation value for monitoring point 5 = -150 - (-157.5) = +7.5mV; First oxidation-reduction potential deviation value for monitoring point 6 = -160 - (-157.5) = -2.5mV. The total dissolved oxygen value, total oxidation-reduction potential value, first dissolved oxygen deviation value, first oxidation-reduction potential deviation value, and pH value are collectively used as parameters for the slurry oxidation state. Simultaneously, the blower control system collects the sulfur dioxide inlet concentration from the flue gas inlet of the absorption tower, which is measured to be 450 mg / Nm³ using a flue gas analyzer; the sulfur dioxide outlet concentration is collected from the flue gas outlet of the absorption tower, measured to be 28 mg / Nm³, and these sulfur dioxide inlet and outlet concentrations are used as operating condition parameters. Furthermore, the blower control system obtains the current motor speed of the magnetic levitation oxidation blower as 18000 rpm and the current guide vane opening as 65%, and uses these current motor speed and guide vane opening as blower control parameters.

[0040] In the above embodiments, the input features of the DTW-CNN-LSTM prediction model are determined using a dual criterion of mechanistic screening and statistical validation. In the mechanistic screening stage, the kinetic equations of the wet desulfurization oxidation reaction are used... Parameters directly related to the oxidation reaction were selected as candidate features, including fan control parameters (current motor speed, current guide vane opening), slurry oxidation state parameters (total dissolved oxygen, total oxidation-reduction potential, first dissolved oxygen deviation, first oxidation-reduction potential deviation, pH value), and operating condition parameters (sulfur dioxide inlet concentration, sulfur dioxide outlet concentration), ensuring a physical correlation between the features and the predicted slurry state parameters. In the statistical validation phase, Pearson correlation coefficient and mutual information were used to quantitatively evaluate the importance of the candidate features. Redundant features with an absolute Pearson correlation coefficient less than 0.3 and mutual information less than 0.1 were eliminated, ultimately retaining 8 core features for model input.

[0041] In an optional embodiment, during the operation of the wet desulfurization system, the fan control parameters, slurry oxidation state parameters, and operating condition parameters are input into the DTW-CNN-LSTM prediction model to obtain the predicted slurry state parameters output by the DTW-CNN-LSTM prediction model after performing a target airflow prediction operation on the fan control parameters, slurry oxidation state parameters, and operating condition parameters. Specifically, this includes: inputting the fan control parameters, slurry oxidation state parameters, and operating condition parameters into the DTW-CNN-LSTM prediction model so that the DTW-CNN-LSTM prediction model performs the following target airflow prediction operation on the fan control parameters, slurry oxidation state parameters, and operating condition parameters: the DTW-CNN-LSTM prediction model converts the fan control parameters, slurry oxidation state parameters, and operating condition parameters into a three-dimensional feature matrix through the input layer; the DTW-CNN-LSTM prediction model performs time-series alignment processing on the three-dimensional feature matrix through the DTW time-series alignment module to obtain a time-synchronized sequence, the time-synchronized sequence being the one after the fan control parameters have been eliminated. The time-aligned sequence of response lag times between control parameters and slurry oxidation state parameters is used. The DTW-CNN-LSTM prediction model extracts correlation features from the time-synchronized sequence through the CNN local feature extraction module to obtain short-term local correlation features. These short-term local correlation features represent the short-term correlation response pattern between the adjustment changes of the fan control parameters and the response changes of the slurry oxidation state parameters within a first preset time period. The DTW-CNN-LSTM prediction model extracts temporal features from the short-term local correlation features through the LSTM long-term dependency capture module to obtain long-term cumulative trend features. These long-term cumulative trend features represent the long-term response characteristics of the slurry oxidation state parameters to the fan control parameters within a second preset time period, which is longer than the first preset time period. The DTW-CNN-LSTM prediction model performs feature mapping processing on the long-term cumulative trend features through a fully connected output layer to obtain predicted slurry state parameters, including predicted regional dissolved oxygen values, predicted regional redox potential values, and predicted outlet pressure values.

[0042] Among them, the three-dimensional feature matrix refers to a data structure that organizes the fan control parameters, slurry oxidation state parameters, and operating condition parameters according to three dimensions: time step, feature category, and feature dimension, and is used to uniformly represent multi-source heterogeneous input data; the DTW time sequence alignment module refers to a neural network module that uses a dynamic time warping algorithm to eliminate time offsets between time series; the time synchronization sequence represents the synchronized time series after DTW alignment processing, which has eliminated the time lag between the fan control parameters and the slurry oxidation state parameters; the response lag time is used to represent the change in the slurry oxidation state parameters (such as the current motor speed, the current guide vane opening) after the control parameters of the magnetic levitation oxidation fan (such as the current motor speed, the current guide vane opening) change. The time delay required for changes in total dissolved oxygen (TOO) and total redox potential (COD) can be 1-2 minutes, etc.; the CNN local feature extraction module refers to a feature extraction module using a convolutional neural network architecture to extract local pattern features of the time series; the correlation feature extraction refers to the operation of using convolutional kernels to slide across the time series to perform weighted summation to extract feature correlation patterns; short-term local correlation features are used to represent the response correlation patterns between the adjustment changes of the fan control parameters and the response changes of the slurry oxidation state parameters within a short time window (such as 3-5 minutes); the first preset time period refers to the length of the time window used to extract short-term local correlation features, for example, 3 minutes or 5 minutes, etc. The LSTM long-term dependency capture module represents a time-series modeling module using a Long Short-Term Memory (LSTM) network architecture. It extracts temporal features from short-term local correlations to capture long-term dependencies in time series. Temporal feature extraction represents the process of feature mining short-term local correlations to reveal their temporal evolution. Long-term cumulative trend features refer to the long-term response characteristics of slurry oxidation state parameters to wind turbine control parameters over a relatively long time period (e.g., 30 minutes to 2 hours). The second preset time period represents the length of the time window used to capture long-term cumulative trend features, such as 30 minutes, 1 hour, or 2 hours. The fully connected output layer represents the connection between all neurons in the neural network and the next layer. A fully connected network layer is used to perform feature mapping on long-term cumulative trend features to obtain the final prediction output. Feature mapping processing refers to the data processing operation that maps long-term cumulative trend features to predicted slurry state parameters through the nonlinear transformation of the fully connected network. Predicted slurry state parameters refer to the prediction results of the DTW-CNN-LSTM prediction model on the future oxidation state of slurry. Predicted regional dissolved oxygen values ​​represent the dissolved oxygen values ​​of each region at future time predicted by the model. Predicted regional redox potential values ​​represent the redox potential values ​​of each region at future time predicted by the model. Predicted outlet pressure values ​​represent the future pressure values ​​at the outlet of the oxidation air duct predicted by the model.

[0043] In the above embodiments, the operating scenario of the wet desulfurization system of the 300MW thermal power generating unit will continue to be used as an example for explanation. The blower control system inputs the blower control parameters, slurry oxidation state parameters, and operating condition parameters into the DTW-CNN-LSTM prediction model. Among them, the blower control parameters include the current motor speed of 18000rpm and the current guide vane opening of 65%; the slurry oxidation state parameters include the total dissolved oxygen value of 1.75mg / L, the total oxidation-reduction potential value of -157.5mV, the first dissolved oxygen deviation value (root mean square value of 0.17mg / L at each monitoring point), the first oxidation-reduction potential deviation value (root mean square value of 8.4mV at each monitoring point), and the average pH value of 5.15; the operating condition parameters include the sulfur dioxide inlet concentration of 450mg / Nm³ and the sulfur dioxide outlet concentration of 28mg / Nm³. The above parameters constitute a total of 8-dimensional feature inputs, including 2-dimensional fan control parameters, 5-dimensional slurry oxidation state parameters, and 1-dimensional operating condition parameters (the sulfur dioxide inlet concentration and sulfur dioxide outlet concentration are used as overall indicators of desulfurization efficiency).

[0044] In the above embodiment, the DTW-CNN-LSTM prediction model converts the turbine control parameters, slurry oxidation state parameters, and operating condition parameters into a three-dimensional feature matrix through the input layer. The above parameter data are collected over the past 30 minutes (one time step per minute, for a total of 30 time steps). The time window length of 30 minutes is set as follows: Through power spectral density analysis of historical operating data, the typical disturbance period of the wet desulfurization system is determined to be 25-30 minutes. This disturbance mainly originates from changes in generator load and fluctuations in flue gas volume. Setting the time window length to 30 minutes can fully cover the typical disturbance period of the wet desulfurization system, ensuring that the model can capture the complete disturbance characteristics and their impact on the slurry oxidation state. The input layer organizes the above parameters according to three dimensions: time step, feature category, and feature dimension, forming a three-dimensional feature matrix with dimensions of 30×3×8. The first dimension represents 30 time steps, the second dimension represents 3 types of parameters (fan control parameters, slurry oxidation state parameters, and operating condition parameters), and the third dimension represents 8 feature dimensions.

[0045] In the above embodiment, the DTW-CNN-LSTM prediction model performs the following target airflow prediction operation on the three-dimensional feature matrix: First, the DTW-CNN-LSTM prediction model performs time-series alignment processing on the three-dimensional feature matrix through the DTW time-series alignment module. Let the fan control parameter sequence (such as the motor speed change sequence) be X, and the slurry oxidation state parameter sequence (such as the total dissolved oxygen value change sequence) be Y. Since the slurry oxidation state parameter requires 1-2 minutes to respond after the fan control parameter changes, there is a time lag between the two sequences. The DTW module constructs a distance matrix and finds the optimal warping path through a dynamic time warping algorithm, adjusting X and Y into a time-synchronized sequence, thus eliminating the influence of response lag time. For example, if the motor speed adjustment at the 10th minute causes a change in the total dissolved oxygen value at the 12th minute, the DTW module will align these two time points to obtain a time-synchronized sequence. and The second step involves the DTW-CNN-LSTM prediction model extracting associated features from the time-synchronized sequence using a CNN local feature extraction module. This module employs a two-layer 1D convolutional structure: the first layer uses 64 kernels of size 3, and the second layer uses 128 kernels of size 3, both with ReLU activation. The convolutional kernels slide along the time dimension, extracting short-term correlation response patterns between changes in the fan control parameters and changes in the slurry oxidation state parameters. The first preset time period is set to 3 minutes, meaning the convolutional kernels cover the current time point and a 1-minute data window before and after it, capturing the short-term local correlation features between the fan control parameters and the slurry oxidation state parameters within 3 minutes. For example, when the motor speed suddenly increases from 18000 rpm to 18500 rpm, the CNN local feature extraction module extracts a short-term response pattern within 3 minutes after this speed increase: a 0.15 mg / L increase in total dissolved oxygen and an 8 mV increase in total redox potential. After convolution and pooling, a 768-dimensional short-term local correlation feature vector is obtained.

[0046] In the above embodiment, in the third step, the DTW-CNN-LSTM prediction model extracts temporal features from short-term local correlation features through the LSTM long-term dependency capture module. The LSTM long-term dependency capture module receives the 768-dimensional short-term local correlation feature vector output by the CNN local feature extraction module as input. The LSTM long-term dependency capture module adopts a two-layer LSTM structure, with the first layer containing 128 hidden units and the second layer containing 64 hidden units. The LSTM long-term dependency capture module performs temporal modeling of short-term local correlation features through gating mechanisms such as forget gate, input gate, cell state, and output gate, capturing the long-term response characteristics of slurry oxidation state parameters to fan control parameters. The second preset time period is set to 30 minutes to 2 hours. For example, when the fan runs at high speed continuously for 1 hour, the LSTM long-term dependency capture module captures the continuous cumulative increase effect (from -180mV to -145mV) of the total redox potential value caused by this continuous high-speed operation state based on the temporal evolution law of short-term local correlation features, obtaining a 64-dimensional long-term cumulative trend feature vector. The fourth step involves the DTW-CNN-LSTM prediction model performing feature mapping on the long-term cumulative trend features through a fully connected output layer. The fully connected layer receives the 64-dimensional long-term cumulative trend feature vector output by the LSTM long-term dependency capture module. After nonlinear transformation by the fully connected network, the long-term cumulative trend features are mapped to predicted slurry state parameters. In this embodiment, the predicted slurry state parameters output by the fully connected output layer include: predicted regional dissolved oxygen values, where the predicted dissolved oxygen value for the lower region is 1.80 mg / L and the predicted dissolved oxygen value for the upper region is 1.90 mg / L; predicted regional redox potential values, where the predicted redox potential value for the lower region is -158 mV and the predicted redox potential value for the upper region is -148 mV; and a predicted outlet pressure value of 12.5 kPa. The blower control system acquires the predicted slurry state parameters output by the DTW-CNN-LSTM prediction model for subsequent determination of the predicted airflow based on the predicted slurry state parameters and operating condition parameters.

[0047] In an optional embodiment, the DO-ORP deviation feedback model is used to perform airflow correction on the slurry oxidation state parameters to obtain the corrected airflow. Specifically, this includes: using the DO-ORP deviation feedback model to perform a third deviation analysis on the total dissolved oxygen value and the total oxidation-reduction potential value to determine a second dissolved oxygen deviation value between the total dissolved oxygen value and a preset dissolved oxygen threshold, and determining a second oxidation-reduction potential deviation value between the total oxidation-reduction potential value and a preset oxidation-reduction potential threshold; determining a basic corrected airflow based on the second dissolved oxygen deviation value and the second oxidation-reduction potential deviation value; determining the spatial distribution non-uniformity of slurry oxidation based on the first dissolved oxygen deviation value and the first oxidation-reduction potential deviation value; comparing the spatial distribution non-uniformity with a preset non-uniformity threshold to determine a compensation corrected airflow when the spatial distribution non-uniformity is greater than the preset non-uniformity threshold; and superimposing the basic corrected airflow and the compensation corrected airflow to obtain the corrected airflow. If the corrected airflow is positive, the output airflow of the magnetic levitation oxidation fan needs to be increased; if the corrected airflow is negative, the output airflow of the magnetic levitation oxidation fan needs to be reduced.

[0048] The third deviation analysis involves comparing the total dissolved oxygen (DO) and total oxidation-reduction potential (ORP) values ​​with the target parameters for slurry oxidation to obtain the deviation data processing operation. The target parameters for slurry oxidation include a preset dissolved oxygen threshold and a preset ORP threshold. The preset dissolved oxygen threshold is a target control value for slurry dissolved oxygen set in advance according to the desulfurization process requirements, which can be set to 2-4 mg / L, etc., to ensure that calcium sulfite is fully oxidized to calcium sulfate. The second dissolved oxygen deviation value represents the difference between the total dissolved oxygen value and the preset dissolved oxygen threshold; a positive value indicates that the total dissolved oxygen value is higher than the preset dissolved oxygen threshold, and a negative value indicates that the total dissolved oxygen value is lower than the preset dissolved oxygen threshold. The preset ORP threshold is a target control value for slurry ORP set in advance according to the desulfurization process requirements, which can be set to -120mV to -80mV, etc., to characterize the oxidizing power of the slurry. The second ORP deviation value is the difference between the total ORP value and the preset ORP threshold. The basic correction air volume is used to represent the overall oxidation state of the slurry. The airflow adjustment amount, determined by the deviation between the slurry oxidation state and the target parameters, is used to adjust the slurry oxidation state to the range defined by the target parameters. Spatial distribution non-uniformity indicates the degree of dispersion of oxidation state parameters at different locations within the slurry pool; a larger value indicates a more uneven distribution of oxidation state within the pool. The preset non-uniformity threshold is a pre-set critical value used to determine whether the oxidation state distribution within the pool is uniform. For example, a normalized non-uniformity greater than 0.3 is considered non-uniform. Distribution comparison refers to the operation of comparing the spatial distribution non-uniformity with the preset non-uniformity threshold to determine the uniformity of the oxidation state distribution within the pool. The compensation correction airflow represents the additional airflow adjustment amount added to improve oxidation uniformity when the oxidation state distribution within the pool is non-uniform. The correction airflow is the total airflow adjustment amount obtained by superimposing the basic correction airflow and the compensation correction airflow. A positive value indicates that the output airflow of the magnetic levitation oxidation fan needs to be increased to improve the oxidation degree, while a negative value indicates that the output airflow of the magnetic levitation oxidation fan needs to be reduced to avoid over-oxidation and energy waste.

[0049] In the above embodiments, the working process of the DO-ORP deviation feedback model is explained in detail using the operating scenario of the 300MW thermal power unit wet desulfurization system as an example. First, the DO-ORP deviation feedback model is used to perform a third deviation analysis on the total dissolved oxygen value and the total redox potential value. Based on the values ​​determined in the aforementioned embodiments... According to the requirements of the desulfurization process, the dissolved oxygen threshold is preset. (A preset dissolved oxygen threshold ensures sufficient oxidation of calcium sulfite). Determine the second dissolved oxygen deviation value. (A negative value indicates that the total dissolved oxygen level is below the preset dissolved oxygen threshold, requiring increased air supply.) Similarly, based on the total redox potential value determined in the aforementioned embodiments... Redox potential threshold (A preset redox potential threshold characterizes the slurry as having good oxidizing properties). Determine the second redox potential deviation value. (A negative value indicates that the total redox potential is lower than the preset redox potential threshold, indicating insufficient oxidizing power). The base correction air volume is determined based on the second dissolved oxygen deviation value and the second redox potential deviation value. A weighted linear combination formula is used for calculation. ,in, This is the dissolved oxygen correction factor, with a value of [value missing]. This indicates that the dissolved oxygen level deviates from the preset dissolved oxygen threshold. Adjustments are needed Air volume; This is the redox potential correction factor, with a value of [value missing]. This indicates that the redox potential deviates from the preset redox potential threshold. Adjustments are needed The air volume. Substitute the values ​​into the calculation. Since both ΔDO and ΔORP are negative (indicating insufficient oxidation), the base correction air volume is taken as positive, i.e. (Increased airflow is required).

[0050] In the above embodiments, the spatial distribution non-uniformity of slurry oxidation is determined based on the first dissolved oxygen deviation value and the first redox potential deviation value. The standard deviation of dissolved oxygen spatial distribution is determined using the standard deviation method. ,in, Let be the first dissolved oxygen deviation value at the i-th oxidation state monitoring point, and n be the total number of oxidation state monitoring points. This represents the sum of the squares of the deviations over all monitoring points. Substitute the values ​​into the calculation. Standard deviation of the spatial distribution of redox potential ,in, Let be the redox potential deviation value of the i-th oxidation state monitoring point, and n be the total number of oxidation state monitoring points. Substitute into the numerical calculation. The comprehensive spatial distribution unevenness is determined using a normalized weighting method. ,in, and These are weighting coefficients, taken as 0.6 and 0.4 respectively; , This is the baseline deviation value. Substitute it into the calculation. The spatial distribution non-uniformity is compared with a preset non-uniformity threshold. Due to the current spatial non-uniformity > To determine if the oxidation state distribution within the slurry tank is uneven, a compensation / correction airflow rate needs to be determined. The compensation / correction airflow rate is determined using the following formula. ,in, The compensation coefficient is set to 5. The baseline air volume is the current predicted air volume. Substitute into the calculation (rounded to the nearest integer) Since it is necessary to improve oxidation uniformity, the compensation correction air volume is taken as a positive value, i.e. The corrected air volume is obtained by superimposing the base corrected air volume and the compensated corrected air volume. (rounded to the nearest integer) Correct the airflow to a positive value. This indicates that it is necessary to increase the output air volume of the magnetic levitation oxidation fan in order to improve the oxidation degree of the slurry and improve the oxidation uniformity.

[0051] In an optional embodiment, a target control airflow is generated based on the predicted airflow and the corrected airflow, and the target control airflow is used to perform operation control operations on the magnetic levitation oxidation fan of the wet desulfurization system. Specifically, this includes: performing airflow compensation processing on the predicted airflow and the corrected airflow to generate the target control airflow; obtaining the current outlet airflow and current outlet pressure of the magnetic levitation oxidation fan; determining a fan control strategy based on the target control airflow, the current outlet airflow, and the current outlet pressure, the fan control strategy including a guide vane adjustment strategy and a speed adjustment strategy; comparing the target control airflow with the current outlet airflow and comparing the current motor speed with a preset speed threshold, so that when the target control airflow is less than the current outlet airflow and the current motor speed is less than the preset speed threshold, the control strategy is implemented. When the target control air volume is less than the current outlet air volume and the current motor speed is greater than or equal to the preset speed threshold, the guide vane adjustment strategy is executed; the current guide vane opening is compared with the preset limit opening, so that when the current guide vane opening reaches the preset limit opening, the speed adjustment strategy is executed; or when the target control air volume is greater than or equal to the current outlet air volume, the speed adjustment strategy is executed; the current outlet air volume and the target control air volume are subjected to a fourth deviation analysis to obtain the deviation control air volume; the deviation control air volume is compared with the preset deviation threshold, so that when the deviation control air volume is greater than the preset deviation threshold, the current guide vane opening or the current motor speed is iteratively adjusted until the deviation control air volume is less than or equal to the preset deviation threshold.

[0052] Among them, air volume compensation processing refers to the data processing operation of performing mathematical operations on the predicted air volume and the corrected air volume to generate the final control air volume, which can be done by addition or weighted summation; the target control air volume represents the final air volume setpoint that the magnetic levitation oxidation fan should output after comprehensively considering the predicted air volume and the corrected air volume, and the unit is standard cubic meters per minute. The current outlet air volume represents the actual output air volume of the magnetic levitation oxidation blower, measured in real time by a flow meter installed on the blower's outlet pipe; the current outlet pressure refers to the gas pressure at the current outlet of the magnetic levitation oxidation blower, measured in kilopascals (kPa), measured in real time by a pressure transmitter installed on the blower's outlet pipe; the blower control strategy represents the blower adjustment scheme determined based on the target control air volume and the current operating status, including adjusting the current guide vane opening, adjusting the current motor speed, or a combination of both; the guide vane adjustment strategy represents... The airflow control method that adjusts the airflow by changing the opening angle of the inlet guide vanes of the magnetic levitation oxidation fan is suitable for situations where the airflow needs to be reduced and the current motor speed is low. The speed adjustment strategy refers to the control method that adjusts the airflow by changing the current motor speed of the magnetic levitation oxidation fan. This is suitable for situations where the airflow needs to be increased, decreased, or the current motor speed is high, or the current guide vane opening has reached a preset limit. Airflow comparison refers to comparing the target control airflow with the current outlet airflow to determine the direction of airflow adjustment. Speed ​​comparison refers to comparing the current motor speed with a preset... The operation involves comparing speed threshold values ​​to select an adjustment strategy. A preset speed threshold represents the critical motor speed value used to determine whether to prioritize guide vane adjustment. For example, guide vane adjustment is prioritized when the current motor speed is below 60% of the rated speed, and speed adjustment is used when the current motor speed is above or equal to this critical value. The opening comparison compares the current guide vane opening with a preset limit opening to determine if a switch in adjustment strategy is needed. The preset limit opening represents the minimum and maximum opening limits allowed for the guide vane device, such as a minimum opening of 35% and a maximum opening of 100%, used to protect the fan from unstable operating conditions. The fourth deviation analysis compares the current outlet airflow with the target control airflow to obtain the deviation data processing operation. The deviation control airflow represents the difference between the current outlet airflow and the target control airflow, reflecting the degree of deviation between the actual fan output and the target control airflow. The deviation comparison compares the deviation control airflow with a preset deviation threshold to determine if further adjustment is needed. The preset deviation threshold is the maximum allowable airflow control deviation, for example, ±2. When the deviation exceeds the preset deviation threshold, adjustment is required; iterative adjustment is used to represent a cyclic control process that repeatedly performs adjustment operations until the deviation control air volume is less than or equal to the preset deviation threshold.

[0053] In the above embodiments, taking the operation scenario of the 300MW thermal power unit wet desulfurization system as an example, the generation of the target control air volume and the execution process of the turbine control strategy are explained in detail. First, the predicted air volume and the corrected air volume are processed for air volume compensation to generate the target control air volume. An addition operation method is used. This results in a target control air volume of 197. To obtain the current operating status of the magnetic levitation oxidation fan: the current outlet air volume is measured by a vortex flow meter installed on the fan's outlet duct. The current outlet pressure is measured by a pressure transmitter installed on the blower outlet pipe. The fan control strategy is determined based on the target control air volume, current outlet air volume, and current outlet pressure. The fan control strategy includes two types: guide vane adjustment strategy and speed adjustment strategy. To query the current operating parameters of the magnetic levitation oxidation fan: current motor speed... Current leading leaf opening Preset speed threshold (Approximately 82% of the rated speed of 5500 r / min), preset limit opening includes minimum opening limit. and maximum opening limit .

[0054] In the above embodiments, scenario one is assumed to be the execution of a speed adjustment strategy (airflow increase scenario). The target control airflow is compared with the current outlet airflow: target control airflow... Current outlet air volume Since the target control air volume is greater than the current outlet air volume, a speed adjustment strategy is executed when the target control air volume is greater than or equal to the current outlet air volume. During the speed adjustment process, a fourth deviation analysis is performed between the current outlet air volume and the target control air volume. = The deviation control air volume was obtained as 17. The deviation control airflow is compared with the preset deviation threshold. Due to deviation, the air volume was controlled at 17. Greater than the preset deviation threshold 2 The current motor speed needs to be iteratively adjusted. The first iteration: Based on the fan performance curve, the target speed is set to 4450 r / min, and a speed adjustment command is sent to the frequency converter. After the speed adjustment is completed, the outlet air volume is measured to be 188. Deviation control air volume The iteration continued. Second iteration: The target rotation speed was adjusted to 4650 r / min. After the rotation speed adjustment was completed, the outlet air volume was measured to be 195. Deviation control air volume The deviation control airflow is now less than or equal to the preset deviation threshold, so iterative adjustments stop. At this point, the magnetic levitation oxidation fan is operating stably at a speed of 4650 r / min and an outlet airflow of 195. At an operating point with an outlet pressure of 52 kPa, precise tracking and control of the target air volume were achieved.

[0055] In the above embodiments, it is assumed that scenario two involves implementing a guide vane adjustment strategy. At another operational moment, the target is to control the airflow. Current outlet air volume Current motor speed Current leading leaf opening Compare the target control air volume with the current outlet air volume: Target control air volume 165 Less than the current outlet air volume of 180 The current motor speed is compared with the preset speed threshold: the current motor speed of 4000 r / min is less than the preset speed threshold of 4500 r / min. Since the target controlled air volume is less than the current outlet air volume and the current motor speed is less than the preset speed threshold, the triggering conditions for the guide vane adjustment strategy are met, and the guide vane adjustment strategy is executed. During the guide vane adjustment process, a fourth deviation analysis is performed between the current outlet air volume and the target controlled air volume. The deviation control air volume was obtained as 15. The deviation control airflow is compared with the preset deviation threshold. Since the deviation control airflow is 15... Greater than the preset deviation threshold 2 The current guide vane opening needs to be iteratively adjusted. First iteration: Based on the guide vane characteristic curve, the target guide vane opening is set to 72%. After adjustment, the measured outlet airflow is 170. Deviation control air volume The iteration continued. Second iteration: The target guide vane opening was adjusted to 68%. After adjustment, the outlet airflow was measured to be 164. Deviation control air volume The deviation control airflow is now less than or equal to the preset deviation threshold; iterative adjustment has stopped.

[0056] In the above embodiments, it is assumed that scenario three involves switching to speed regulation after the current guide vane opening reaches a preset limit. At another point in operation, the target control airflow... Current outlet air volume Current motor speed Current leading leaf opening Compare the target control air volume with the current outlet air volume: Target control air volume 145 Less than the current outlet air volume of 155 The current motor speed is compared with the preset speed threshold: the current motor speed of 3800 r / min is less than the preset speed threshold of 4500 r / min. Initially, the conditions for the guide vane adjustment strategy are met, and guide vane adjustment begins. After several iterations, the current guide vane opening is adjusted to 35% (reaching the minimum opening limit). However, the outlet air volume remains at 150. The deviation control air volume is 5 >2 The current guide vane opening is compared with the preset limit opening. Since the current guide vane opening of 35% has reached the preset limit opening (minimum opening limit), the speed adjustment strategy is switched. The current motor speed is iteratively adjusted, reducing the motor speed from 3800 r / min to 3500 r / min, and the outlet air volume is reduced to 146. Deviation control air volume Stop iterative adjustments.

[0057] In the above embodiments, scenario four is assumed to be a scenario of implementing a speed adjustment strategy (a scenario where airflow is reduced but speed is higher). At another point in operation, the target is to control the airflow. Current outlet air volume Current motor speed Current leading leaf opening Compare the target control air volume with the current outlet air volume: Target control air volume 170 Less than the current outlet air volume of 185 The current motor speed is compared with the preset speed threshold: the current motor speed of 4800 r / min is greater than the preset speed threshold of 4500 r / min. Since the target controlled air volume is less than the current outlet air volume and the current motor speed is greater than or equal to the preset speed threshold, a speed adjustment strategy is determined to be executed. During the speed adjustment process, a fourth deviation analysis is performed between the current outlet air volume and the target controlled air volume. The deviation control air volume was obtained as 15. The deviation control airflow is compared with the preset deviation threshold. Since the deviation control airflow is 15... Greater than the preset deviation threshold 2 The current motor speed needs to be iteratively adjusted. The first iteration: Based on the fan performance curve, the target speed is set to 4550 r / min, and a speed adjustment command is sent to the frequency converter. After the speed adjustment is completed, the outlet air volume is measured to be 175. Deviation control air volume The iteration continued. Second iteration: The target rotation speed was adjusted to 4400 r / min. After the rotation speed adjustment was completed, the outlet air volume was measured to be 169. Deviation control air volume The deviation control airflow is now less than or equal to the preset deviation threshold, so iterative adjustments stop. At this point, the magnetic levitation oxidation fan is operating stably at a speed of 4400 r / min and an outlet airflow of 169. The air volume reduction target was achieved by adjusting the speed rather than the guide vane at the operating point, thus avoiding the airflow disturbance and efficiency loss that might occur when using guide vane adjustment at high speed.

[0058] In an optional embodiment, when the target controlled air volume is less than the current outlet air volume and the current motor speed is less than a preset speed threshold, a guide vane adjustment strategy is executed, specifically including: determining the air volume difference between the target controlled air volume and the current outlet air volume; determining the target guide vane opening based on the air volume difference and a preset guide vane characteristic curve, where the guide vane characteristic curve characterizes the correspondence between the guide vane opening and air volume changes; determining the opening adjustment amount of the guide vane device of the magnetic levitation oxidation fan based on the target guide vane opening and the current guide vane opening; and decomposing the opening adjustment amount into multiple [various parameters] when the opening adjustment amount is greater than a preset maximum single adjustment angle. The adjustment steps involve multiple adjustment steps, each with an adjustment range not exceeding a preset maximum single adjustment angle. The control device adjusts the current guide vane opening according to multiple adjustment steps to generate a pre-rotation component in the same direction as the impeller rotation of the magnetic levitation oxidation fan. During the guide vane opening adjustment process, the vibration and surge states of the magnetic levitation oxidation fan are monitored in real time. If the vibration value of the magnetic levitation oxidation fan exceeds a preset vibration threshold or the surge margin of the magnetic levitation oxidation fan is less than a preset safety margin, the adjustment of the current guide vane opening is stopped.

[0059] The air volume difference refers to the difference between the target control air volume and the current outlet air volume, expressed in standard cubic meters per minute (m³ / min). The system reflects the amount of airflow change that needs to be achieved through adjustment; the preset guide vane characteristic curve represents the characteristic curve describing the relationship between the guide vane opening and the airflow change, which can be obtained through fan performance testing or CFD simulation. For example, when the guide vane opening decreases from 100% to 80%, the airflow decreases by about 15%; the target guide vane opening is used to represent the guide vane opening angle value that needs to be adjusted to achieve the target controlled airflow, in degrees (°) or percentages (%); the guide vane device refers to the adjustable guide vane device installed at the air inlet of the magnetic levitation oxidation fan, consisting of multiple (e.g., 8-12) synchronously rotating guide vanes, a transmission mechanism, and an electric motor. The actuator consists of the following components: The current guide vane opening indicates the degree of opening of the guide vanes relative to the fully open position; 100% indicates the vanes are fully open, and 0% indicates the vanes are fully closed. The opening adjustment amount represents the required angle change in the guide vane opening, equal to the difference between the target guide vane opening and the current guide vane opening. The preset maximum single adjustment angle refers to the maximum allowable angle change in a single adjustment, limited to ensure smooth adjustment and avoid fan surge, for example, 5 degrees or 10 degrees. The adjustment steps represent the decomposition of a large opening adjustment amount into multiple smaller adjustment actions, with the amplitude of each adjustment action in the multiple adjustment steps not exceeding the preset value. The maximum adjustment angle in a single operation; the pre-swirl component represents the tangential velocity component of the airflow before it enters the magnetic levitation oxidation fan, which is in the same direction as the impeller rotation. Pre-swirl reduces the relative velocity between the airflow and the impeller, thus reducing the impeller's work and power consumption; vibration state refers to the vibration of the magnetic levitation oxidation fan during operation, quantified by vibration values; vibration value refers to the amplitude of the vibration of the magnetic levitation oxidation fan during operation, measured in millimeters per second (mm / s), and measured by vibration sensors installed on the fan bearing housing or casing; the preset vibration threshold represents the maximum safe limit of the allowable fan vibration value, for example... For example, a vibration rate of 4.5 mm / s, exceeding the preset vibration threshold, may damage the fan. Surge state refers to the safe state of the magnetic levitation oxidation fan relative to the surge boundary during operation, which is quantified by the surge margin. The surge margin is used to represent the safe distance between the fan's current operating point and the surge boundary. The larger the surge margin, the more stable the fan operation; the smaller the surge margin, the closer the fan is to the surge zone. The preset safety margin refers to the minimum surge margin required to ensure the safe operation of the fan, such as 12%, 14%, 15%, etc. (not limited here). When the surge margin is less than the preset safety margin, adjustment needs to be stopped or anti-surge measures need to be taken.

[0060] In the above embodiments, taking the operating conditions of scenario two as an example, the execution process of the guide vane adjustment strategy is explained in detail. The current operating state is: target control air volume. Current outlet air volume Current motor speed (Less than the preset speed threshold of 4500 r / min), current guide vane opening When the target control airflow is less than the current outlet airflow and the current motor speed is less than a preset speed threshold, the guide vane adjustment strategy is executed. The first step is to determine the airflow difference between the target control airflow and the current outlet airflow. A negative airflow difference indicates that the airflow needs to be reduced. The second step is to determine the target guide vane opening based on the airflow difference and the preset guide vane characteristic curve. The preset guide vane characteristic curve is obtained through fan performance testing and characterizes the relationship between the guide vane opening and airflow changes. According to the preset guide vane characteristic curve, under the current motor speed of 4000 r / min, when the guide vane opening decreases from 85% to 68%, the airflow decreases by approximately 15%. Therefore, determining the target leading edge opening is crucial. The third step is to determine the opening adjustment amount of the guide vane device of the magnetic levitation oxidation fan based on the target guide vane opening and the current guide vane opening. The opening adjustment amount is -17%, indicating that the current leading vane opening needs to be reduced by 17 percentage points (corresponding to an angle change of approximately 15.3 degrees). The fourth step is to determine if the opening adjustment amount is greater than the preset maximum single adjustment angle. The preset maximum single adjustment angle is 5 degrees (corresponding to an opening change of approximately 5.5%). Since the opening adjustment amount of 17% (approximately 15.3 degrees) is greater than the preset maximum single adjustment angle of 5 degrees, the opening adjustment amount needs to be broken down into multiple adjustment steps.

[0061] In the above embodiment, the 17% opening adjustment can be decomposed into four adjustment steps: Step 1 is to adjust from 85% to 80%, with an adjustment range of 5% (approximately 4.5 degrees); Step 2 is to adjust from 80% to 75%, with an adjustment range of 5% (approximately 4.5 degrees); Step 3 is to adjust from 75% to 70%, with an adjustment range of 5% (approximately 4.5 degrees); Step 4 is to adjust from 70% to 68%, with an adjustment range of 2% (approximately 1.8 degrees). The adjustment range of each adjustment step in the multiple adjustment steps does not exceed the preset maximum single adjustment angle of 5 degrees. In the fifth step, the guide vane device is controlled to adjust the current guide vane opening according to multiple adjustment steps. After each adjustment step is completed, a 3-second wait is allowed for the airflow to stabilize before proceeding to the next adjustment step. The guide vane device consists of 10 adjustable guide vanes evenly distributed around the fan inlet. Each vane is connected to an electric actuator via a transmission mechanism and can rotate synchronously. When the current guide vane opening decreases, the guide vanes rotate towards the closing direction, causing the airflow entering the magnetic levitation oxidation fan to generate a pre-swirl component in the same direction as the impeller rotation. This pre-swirl component reduces the relative velocity between the airflow and the impeller, thus reducing the work done by the impeller. According to Euler's equation for centrifugal fans, the work done by the impeller on the airflow is proportional to the difference in tangential velocity between the inlet and outlet airflows. When the inlet airflow has a positive pre-swirl in the same direction as the impeller rotation, the inlet tangential velocity increases, the difference in tangential velocity between the inlet and outlet decreases, and the work done by the impeller on the airflow decreases accordingly, thereby reducing the fan's power consumption. Step 1 adjustment: Adjust the current guide vane opening from 85% to 80%, wait 3 seconds; Step 2 adjustment: Adjust the current guide vane opening from 80% to 75%, wait 3 seconds; Step 3 adjustment: Adjust the current guide vane opening from 75% to 70%, wait 3 seconds; Step 4 adjustment: Adjust the current guide vane opening from 70% to 68%, adjustment complete.

[0062] In the above embodiment, in the sixth step, during the adjustment of the guide vane opening, the vibration and surge states of the magnetic levitation oxidation fan are monitored in real time. Vibration values ​​are measured using vibration sensors installed on the fan bearing housing to characterize the vibration state, with a preset vibration threshold of 4.5 mm / s. The surge margin is obtained by determining the distance between the current operating point and the surge boundary to characterize the surge state. Specifically, the surge boundary curve is obtained based on the performance curve provided by the magnetic levitation oxidation fan manufacturer. This curve characterizes the relationship between the critical pressure and critical flow rate at different speeds when the fan experiences surge. The current outlet airflow is measured in real time using a flow meter and pressure transmitter installed on the fan outlet pipe. and current export pressure Determine the current operating point; find the relationship between the current outlet pressure and the surge boundary curve. Critical surge flow rate for the same pressure value The surge margin is calculated using the flow margin method, and the formula is as follows: For example, the current outlet air volume The surge critical flow rate was determined under the current outlet pressure of 48 kPa. ,but The preset safety margin is 15%, a value derived from industry standards or manufacturer recommendations for magnetic levitation oxidation fans. When the surge margin is below 10%, the fan is at risk of entering the surge zone; below 5%, surge may occur. To ensure the fan operates safely and stably under various conditions and to provide sufficient safety margin to cope with disturbances such as sudden load changes, the preset safety margin is set at 15%.

[0063] In the above embodiment, during the above adjustment process: after performing step 1, the vibration value is 2.1 mm / s < 4.5 mm / s, the surge margin is 28% > 15%, both the vibration and surge states are safe, and adjustment continues; after performing step 2, the vibration value is 2.3 mm / s < 4.5 mm / s, the surge margin is 25% > 15%, both the vibration and surge states are safe, and adjustment continues; after performing step 3, the vibration value is 2.5 mm / s < 4.5 mm / s, the surge margin is 22% > 15%, both the vibration and surge states are safe, and adjustment continues; after performing step 4, the vibration value is 2.4 mm / s < 4.5 mm / s, the surge margin is 20% > 15%, both the vibration and surge states are safe, and adjustment is complete. All adjustment steps have been performed, and both the vibration and surge states are within safe ranges, indicating that the lead vane adjustment strategy has been successfully executed. In a safety protection scenario, during another adjustment process, when the current guide vane opening was adjusted from 60% to 55%, a sudden increase in vibration was detected to 4.8 mm / s, exceeding the preset vibration threshold of 4.5 mm / s. The blower control system immediately stopped adjusting the current guide vane opening, maintaining the current opening at 55%, and generated a vibration over-limit alarm signal. Based on the vibration over-limit alarm signal, the blower control system switched to a speed regulation strategy, reducing the current motor speed to achieve the target control airflow. In yet another adjustment process, when the current guide vane opening was adjusted from 45% to 40%, a surge margin was detected to drop to 12%, less than the preset safety margin of 15%. The blower control system immediately stopped adjusting the current guide vane opening, returning the opening to 45% to restore the safety margin, and generated a surge warning signal.

[0064] In an optional embodiment, when the current guide vane opening reaches a preset limit or the target control air volume is greater than or equal to the current outlet air volume, a speed adjustment strategy is executed. Specifically, this includes: determining whether the current guide vane opening has reached a preset limit, where the preset limit includes a first opening limit and a second opening limit; when the current guide vane opening reaches the first opening limit and the target control air volume is less than the current outlet air volume, a speed adjustment mode is activated; or when the current guide vane opening reaches the second opening limit and the target control air volume is greater than or equal to the current outlet air volume, a speed adjustment mode is activated; or when the target control air volume is greater than the current outlet air volume, a speed adjustment mode is activated; in the speed adjustment mode, the target control air volume, the current motor speed, and a preset fan performance curve are used to determine the target... Rotational speed is preset, and the fan performance curve represents the mapping relationship between air volume and outlet pressure under different rotational speed conditions. A rotational speed adjustment command is generated and sent to the magnetic levitation oxidation fan to adjust the current motor speed to the target speed. During the adjustment of the current motor speed, the rate of change of the current motor speed is constrained to ensure that the rate of change does not exceed a preset change threshold. After determining that the current motor speed has been adjusted to the target speed, the current outlet pressure is compared with the ultimate pressure requirement of the absorber slurry pool. If the current outlet pressure is determined to be less than the ultimate pressure requirement, the current motor speed is increased and the current guide vane opening is decreased until the current outlet air volume reaches the target control air volume and the current outlet pressure is greater than or equal to the ultimate pressure requirement.

[0065] The preset limit opening includes a first opening limit and a second opening limit. The first opening limit refers to the minimum opening degree to which the guide vane is allowed to close, which can be set to 30%-40%, etc. Below the first opening limit, it may cause severe distortion of the intake airflow field or fan surge. The second opening limit refers to the maximum opening degree to which the guide vane is fully open, such as 98%, 99%, 100%, etc. The speed adjustment mode indicates the working mode of adjusting the airflow by changing the current motor speed of the magnetic levitation oxidation fan, and is the main adjustment method of the fan. The preset fan performance curve refers to a family of characteristic curves describing the mapping relationship between airflow and outlet pressure under different speed conditions of the magnetic levitation oxidation fan. These can be provided by the fan manufacturer or obtained through performance testing. The target speed indicates the motor speed value that needs to be adjusted to achieve the target controlled airflow, in revolutions per minute (rpm). The speed adjustment command indicates the speed setting command sent by the blower control system to the frequency converter of the magnetic levitation oxidation fan, including the target speed value and adjustment... The parameters include speed and rate of change; the rate of change refers to the speed at which the current motor speed changes over time, measured in revolutions per minute per second (rpm / s). An excessively rapid rate of change may cause fan vibration or electrical shock. Constraint control refers to the control operation that limits the rate of change of the current motor speed to ensure a smooth adjustment process. The preset change threshold represents the upper limit of the speed change rate limited to ensure a smooth adjustment process, for example, 500 rpm / s. Pressure comparison refers to the operation of comparing the current outlet pressure with the ultimate pressure requirement of the absorber slurry pool to determine whether the pressure requirement is met. The pressure comparison result represents the comparison conclusion obtained from the pressure comparison operation, used to determine whether the current outlet pressure meets the ultimate pressure requirement. The ultimate pressure requirement represents the minimum gas supply pressure required for normal oxidation in the absorber slurry pool, determined based on the slurry pool depth and pipeline resistance. Specifically, the ultimate pressure requirement equals the sum of the static pressure of the slurry pool liquid column, the pipeline friction loss, the local resistance loss, the oxidation air distributor resistance loss, and the safety margin pressure. The formula for calculating the static pressure of the slurry pool liquid column is as follows: ρ is the density of the slurry (for example, take...). (No specific limit is specified here), g is the acceleration due to gravity (e.g., 9.8 m / s², no specific limit is specified here), H is the effective liquid level depth of the slurry pool; the pipe friction loss is calculated according to the Darcy-Weisbach formula, the formula is: λ is the friction factor (e.g., 0.02, not limited here), L is the pipe length, and D is the pipe inner diameter. For air density (e.g., take air density) (Not limited here), v is the air velocity inside the pipe; local resistance loss includes the resistance generated by local components such as elbows and valves, and the calculation formula is: , The local resistance coefficients of each component are given; the resistance loss of the oxidation air distributor is determined based on the distributor structure, generally taken as 5-10 kPa; the safety margin pressure is generally taken as 10%-15% of the total pressure. Taking a medium-sized wet desulfurization system as an example, the calculation is as follows: effective liquid level in the slurry tank H = 5 meters, pipe length L = 80 meters, pipe inner diameter D = 0.3 meters, air velocity in the pipe v = 15 m / s, and there are 4 90-degree elbows (single...). ) and 2 butterfly valves (single) Static pressure of the liquid column in the slurry tank. (The slurry column needs to be overcome to allow the oxidizing air to enter the slurry tank); however, since the oxidizing air distributor is located at the bottom of the slurry tank, the actual static pressure to be overcome is determined by the distributor's submersion depth. Taking the submersion depth as 60% of the effective liquid level, i.e., 3 meters, then... Pipeline friction loss Local resistance loss The resistance loss of the oxidation air distributor is taken as 8 kPa; the sum of various pressure losses = 32 + 0.7 + 0.3 + 8 = 41 kPa; the safety margin pressure is taken as 15%, that is, 41 × 0.15 ≈ 6 kPa; the ultimate pressure requirement = 41 + 6 = 47 kPa, which is rounded to the range of 48-55 kPa. According to the above determination method, for a small wet desulfurization system with an effective liquid level of 3 meters in the slurry tank and a length of 50 meters in the oxidation air duct, the ultimate pressure requirement is approximately 35-40 kPa; for a medium-sized wet desulfurization system with an effective liquid level of 5 meters in the slurry tank and a length of 80 meters in the oxidation air duct, the ultimate pressure requirement is approximately 48-55 kPa; and for a large wet desulfurization system with an effective liquid level of 8 meters in the slurry tank and a length of 120 meters in the oxidation air duct, the ultimate pressure requirement is approximately 65-75 kPa. "Increase adjustment" refers to an adjustment operation that increases the value, such as increasing the current motor speed; "decrease adjustment" refers to an adjustment operation that decreases the value, such as decreasing the current guide vane opening; "coordinated adjustment" is used to indicate a control method that simultaneously adjusts multiple control parameters to achieve the optimal control effect.

[0066] In the above embodiments, three triggering scenarios and specific execution processes of the speed regulation strategy are described in detail. The preset limit opening includes a first opening limit value. Second opening limit This is used to define the adjustable range of the guide vane. The ultimate pressure requirement is determined based on the depth of the absorber slurry pool and the pipeline resistance. Specific examples are as follows: for a small wet desulfurization system with an effective slurry pool level of 3 meters and an oxidation air pipeline length of 50 meters, the ultimate pressure requirement is approximately 35-40 kPa; for a medium-sized wet desulfurization system with an effective slurry pool level of 5 meters and an oxidation air pipeline length of 80 meters, the ultimate pressure requirement is approximately 48-55 kPa; and for a large wet desulfurization system with an effective slurry pool level of 8 meters and an oxidation air pipeline length of 120 meters, the ultimate pressure requirement is approximately 65-75 kPa. This embodiment uses a medium-sized wet desulfurization system as an example. .

[0067] In the above embodiments, assuming scenario one is that the current leading vane opening reaches the first opening limit and the target control air volume is less than the current outlet air volume, the current operating state is: target control air volume. Current outlet air volume Current motor speed Current leading leaf opening (First opening limit reached). First step: Determine if the current leader vane opening has reached the preset limit. Current leader vane opening. , equal to the first opening limit value The first step determines that the current guide vane opening has reached the preset limit. The second step, because the current guide vane opening has reached the first opening limit and the target control airflow is less than the current outlet airflow (…),… < ), activate the speed adjustment mode. Third, in speed adjustment mode, control the airflow to the target volume of 140. The target speed is determined using the current motor speed of 3800 r / min and the preset fan performance curve. The preset fan performance curve, provided by the fan manufacturer, represents the mapping relationship between airflow and outlet pressure under different speed conditions. Based on this curve, with the current guide vane opening at 35%, to increase the airflow from 155... Dropped to 140 The rotational speed needs to be reduced from 3800 r / min to 3450 r / min. Determine the target rotational speed. The fourth step involves generating a speed adjustment command and sending it to the inverter of the magnetic levitation oxidation fan to adjust the current motor speed to the target speed. The speed adjustment command includes the target speed value of 3450 r / min and the adjustment rate parameter. The fifth step involves constraining the rate of change of the current motor speed during the adjustment process. The preset change threshold is 500 r / min / s, meaning the speed change per second should not exceed 500 revolutions per second, ensuring the rate of change does not exceed the preset threshold. Adjusting from 3800 r / min to 3450 r / min involves a speed change of 350 r / min, determined according to the preset change threshold of 500 r / min / s, with a minimum adjustment time of 0.7 seconds. During actual adjustment, the inverter smoothly adjusts the speed according to the constrained rate of change, avoiding fan vibration or electrical shock caused by sudden speed changes. The sixth step involves comparing the current outlet pressure with the ultimate pressure requirement of the absorber slurry tank after determining that the current motor speed has been adjusted to the target speed of 3450 r / min, obtaining the pressure comparison result. The current outlet pressure is then measured. Extreme pressure demand .because The pressure comparison results show that the current outlet pressure meets the ultimate pressure requirement, and the speed adjustment is complete.

[0068] In the above embodiments, assuming scenario two is that the current leading vane opening reaches the second opening limit and the target control air volume is greater than the current outlet air volume, the current operating state is: target control air volume. Current outlet air volume Current motor speed Current leading leaf opening (The second opening limit has been reached). Determine that the current leading vane opening has reached the second opening limit. Furthermore, if the target control air volume is greater than the current outlet air volume, the speed regulation mode is activated. According to the fan performance curve, under the current condition of 100% front guide vane opening, to increase the air volume from 200... Increased to 220 The rotational speed needs to be increased from 4800 r / min to 5100 r / min. Determine the target rotational speed. A speed adjustment command is generated and sent to adjust the current motor speed to the target speed. The frequency converter adjusts the current motor speed from 4800 r / min to 5100 r / min at a rate of change not exceeding a preset threshold of 500 r / min / s. After confirming that the current motor speed has been adjusted to the target speed, the current outlet pressure is compared with the ultimate pressure requirement to obtain the current outlet pressure. The pressure comparison results show that the ultimate pressure requirement is met, and the adjustment was successful.

[0069] In the above embodiments, assuming that scenario three is that the target control air volume is greater than the current outlet air volume, the current operating state is: target control air volume. Current outlet air volume Current motor speed Current leading leaf opening Because the target control air volume is 195... Greater than the current outlet air volume of 175 Activate the speed regulation mode. According to the fan performance curve, under the current condition of 80% front guide vane opening, to increase the air volume from 175... Increased to 195 The rotational speed needs to be increased from 4100 r / min to 4500 r / min. Determine the target rotational speed. A speed adjustment command is generated and sent to adjust the current motor speed to the target speed. The frequency converter adjusts the current motor speed from 4100 r / min to 4500 r / min at a rate of change not exceeding a preset threshold of 500 r / min / s. After confirming that the current motor speed has been adjusted to the target speed, the current outlet pressure is compared with the ultimate pressure requirement to obtain the current outlet pressure. The pressure comparison results show that the ultimate pressure requirement is met, and the adjustment was successful.

[0070] In the above embodiments, assuming scenario four is the coordinated adjustment when the current outlet pressure does not meet the ultimate pressure requirement based on the pressure comparison result, the current operating state is: target control air volume. Current outlet air volume Current motor speed 3600 r / min, current leading vane opening Since the target control air volume is greater than the current outlet air volume, the speed regulation mode is activated. The target speed is determined based on the fan performance curve. After the motor speed adjustment is completed, the current outlet air volume is measured to be 158. (Approaching the target control air volume of 160) However, current export pressure The current outlet pressure was compared with the ultimate pressure requirement, and the result was: current outlet pressure 46 kPa < ultimate pressure requirement 50 kPa. Based on this comparison, the current outlet pressure is less than the ultimate pressure requirement, failing to meet the ultimate pressure requirement of the absorber slurry tank. Therefore, the current motor speed needs to be increased, and the current guide vane opening needs to be decreased until the current outlet airflow reaches the target control airflow and the current outlet pressure is greater than or equal to the ultimate pressure requirement. The specific operation is as follows: First coordinated adjustment: Increase the current motor speed from 3900 r / min to 4100 r / min (an increase of 200 r / min), and simultaneously decrease the current guide vane opening from 95% to 88% (a decrease of 7%). According to the fan characteristics, increasing the speed will increase both airflow and pressure, while decreasing the guide vane opening will decrease the airflow but further increase the pressure. The synergistic effect of these two adjustments can increase the outlet pressure while maintaining a relatively constant airflow. Measurement after adjustment: Current outlet airflow 161 kPa. (Approaching the target control air volume of 160) The current outlet pressure is 49 kPa. Comparing the current outlet pressure with the ultimate pressure requirement: 49 kPa < 50 kPa, still less than the ultimate pressure requirement, further coordinated adjustment is needed. Second coordinated adjustment: Increase the current motor speed from 4100 r / min to 4200 r / min (an increase of 100 r / min), while simultaneously decreasing the current guide vane opening from 88% to 85% (a decrease of 3%). Measurement after adjustment: Current outlet airflow 160... (Target control airflow achieved), current outlet pressure 51 kPa. Comparing the current outlet pressure with the ultimate pressure requirement: 51 kPa > 50 kPa, the current outlet pressure exceeds the ultimate pressure requirement, thus meeting the requirement. At this point, since the current outlet airflow has reached the target control airflow and the current outlet pressure exceeds the ultimate pressure requirement, coordinated regulation stops. The magnetic levitation oxidation fan operates stably at a speed of 4200 r / min, a guide vane opening of 85%, and an outlet airflow of 160... The operating point with an outlet pressure of 51 kPa achieved both the target air volume control and the ultimate pressure requirement, demonstrating successful coordinated regulation.

[0071] It should be noted that the embodiments described above are only some embodiments of this application, and not all embodiments. The present application will be described in detail below with reference to specific embodiments.

[0072] This application provides a control process for precise blowing in wet desulfurization, including the following steps:

[0073] S1: Multiple samples are taken from the slurry pool of the absorption tower. The slurry pool of the absorption tower is measured in real time at multiple points by integrating online devices for pH, density, DO, and ORP.

[0074] S2: Based on parameters such as sulfur dioxide inlet concentration, sulfur dioxide outlet concentration, fan control parameters, and slurry oxidation state parameters, the oxidation air volume is predicted using the DTW-CNN-LSTM prediction model, and the air volume is corrected by combining the real-time measured values ​​of slurry oxidation state parameters with the DO-ORP deviation feedback model, thereby achieving the determination of the target control air volume by combining the predicted air volume and the corrected air volume.

[0075] S3: Automatically adjustable regulating valves and pressure measuring points are installed in the branch pipes of the oxidation air distribution in the slurry tank. The air volume of the branch pipes is adjusted according to the feedback from the pressure measuring points to solve the problem of uneven air volume distribution in the branch pipes.

[0076] S4: The oxidation blower adopts a magnetic levitation blower with high-speed motor direct drive and magnetic levitation non-contact support technology, which solves the problems of low efficiency, high noise, high energy consumption, large size and weight, and poor flow regulation performance of traditional Roots blowers and ordinary centrifugal blowers.

[0077] S5: The magnetic levitation oxidation fan's guide vane device regulates airflow by adjusting the current guide vane opening. Simultaneously, feedback from the outlet fan's volumetric flow rate and pressure measurements pre-rotates the fan inlet, improving the airflow field and impeller efficiency, effectively reducing power loss. Adjusting fan control parameters to regulate different outlet flow rates and pressures maximizes overall efficiency while ensuring surge margin and preventing surge. The guide vane adjustment pre-rotates the airflow before it enters the high-speed impeller to improve airflow conditions and enhance impeller efficiency. Compared to using an inlet throttling method, adjusting the current guide vane opening angle effectively reduces power loss.

[0078] This application provides a multi-point sampling device for the slurry pool of an absorption tower. Figure 2 This is a schematic diagram of the deployment of a multi-point sampling and detection device for the slurry pool of the absorption tower in this application embodiment. (See attached diagram.) Figure 2 , Figure 2The left side of the image shows the absorption tower and its bottom slurry pool. Multiple sampling points are located on the sidewall of the slurry pool, arranged in layers at different depths. The multi-point sampling and testing device for the slurry pool includes a manual shut-off valve 21 for the slurry sampling points, an electric shut-off valve 22 for the slurry sampling points, a process water flushing shut-off valve 23 for the sampling pipeline, a process water flushing discharge valve 24 for the sampling pipeline, an inlet shut-off valve 25 for the integrated measurement device, a comprehensive DO and ORP measurement device 26, a DO measuring instrument 27, an ORP measuring instrument 28, an overflow port 29 for the comprehensive measurement device, and a bottom flow port 210 for the comprehensive measurement device. Each sampling point is equipped with a manual shut-off valve 21 and an electric shut-off valve 22. The sampling pipelines from all sampling points converge and connect to the inlet of the slurry DO and ORP integrated measurement device 26 via the inlet shut-off valve 25. A process water flushing shut-off valve 23 is also installed on the sampling pipeline to introduce flushing water for rinsing the pipeline. The flushed wastewater is discharged through the process water flushing discharge valve 24. The core instrument's working principle: The dissolved oxygen value and oxidation-reduction potential value in the calcium sulfite oxidation reaction in the absorber slurry pool are correlated with its oxidation rate. The DO measuring instrument operates on an optical principle (fluorescence quenching method), suitable for high-solids content or corrosive slurries, with stronger anti-fouling properties. The sensor surface is coated with a fluorescent material (such as ruthenium complex), which emits red light after blue light excitation. Collisions between oxygen molecules and the fluorescent material cause fluorescence intensity decay or shortened lifetime (quenching effect). The oxygen concentration is inferred by detecting changes in fluorescence intensity / lifetime. The basic principle of the ORP measuring instrument is that the electrode is immersed in the slurry, and electron exchange occurs between redox substances on the platinum surface, forming a potential difference. The instrument collects the potential signal, and after filtering, amplification, and temperature compensation, outputs the ORP value (response time can be 10-60 seconds). The working principle of the multi-point sampling device is as follows: Multiple branches are led out from different typical locations in the slurry pool of the absorption tower. By controlling the electric shut-off valve 22 of the slurry sampling point, the slurry medium at different locations is automatically switched to flow into the slurry DO and ORP integrated measuring device 26. This slurry DO and ORP integrated measuring device 26 is equipped with a DO measuring instrument 27 and an ORP measuring instrument 28. After each measurement of the data at a slurry pool location is completed, the process water flushing shut-off valve 23 of the sampling pipeline is opened to flush the integrated measuring device (i.e., including the integrated measuring device overflow port 29 and the integrated measuring device underflow port 210) in preparation for the next measurement process.

[0079] This application provides an adjustable front guide vane energy-saving integrated magnetic levitation oxidation fan. Figure 3 This is a schematic diagram of a structure of an adjustable front guide vane energy-saving integrated magnetic levitation oxidation fan in an embodiment of this application. (See attached diagram.) Figure 3 , Figure 3 The left side of the middle section shows the absorption tower and the absorption tower slurry pool at its bottom. The absorption tower slurry pool is divided into multiple slurry oxidation zones along the depth direction. Figure 3 The area within the dotted box in the upper right corner represents the magnetic levitation oxidation fan 32 and its supporting equipment. Air enters through the inlet of the adjustable guide vane 31 of the magnetic levitation fan, is compressed by the magnetic levitation oxidation fan 32, and is then output to each slurry oxidation zone through the oxidation air main pipe. The integrated magnetic levitation oxidation fan includes the adjustable guide vane 31, the magnetic levitation oxidation fan 32, the fan main pipe shut-off valve 33, the oxidation air main pipe pressure transmitter 34, the oxidation air main pipe pre-cooling temperature transmitter 35, the oxidation air main pipe post-cooling temperature transmitter 36, the oxidation air branch pipe electric regulating valve 37, and the oxidation air branch pipe pressure transmitter 38. The working principle of the magnetic levitation oxidation fan: In wet desulfurization systems, the magnetic levitation oxidation fan is mainly used for supplying oxidation air in key stages. Its core working principle is based on the combination of magnetic levitation bearing technology and a high-speed centrifugal impeller, achieving frictionless and efficient gas transport. An outlet silencer is also installed on the oxidation air header to reduce the noise of the airflow at the outlet of the magnetic levitation oxidation fan 32. An automatic pressure relief valve is installed on the bypass of the fan header shut-off valve 33 to automatically relieve pressure and protect the fan in case of shutdown or abnormal operation. A pressure gauge is also installed near the fan header shut-off valve 33 to display the pressure of the oxidation air header locally. An oxidation air header pressure transmitter 34 and an oxidation air header pre-cooling temperature transmitter 35 are installed sequentially on the oxidation air header to monitor the header pressure and temperature before cooling. A cooling water inlet is provided on the oxidation air header. Cooling water is connected to the oxidation air header through a pipeline to cool the high-temperature oxidation air. The temperature after cooling is monitored by the oxidation air header post-cooling temperature transmitter 36. Downstream of the main oxidation air pipe, there are multiple oxidation air branch pipes. Each oxidation air branch pipe corresponds to a slurry oxidation zone. Each oxidation air branch pipe is equipped with an oxidation air branch pipe electric regulating valve 37 and an oxidation air branch pipe pressure transmitter 38. The oxidation air branch pipe electric regulating valve 37 is used to automatically regulate the branch air volume of each slurry oxidation zone, and the oxidation air branch pipe pressure transmitter 38 is used to monitor the pressure of each branch pipe to provide feedback control for the distribution of branch air volume. Working principle of integrated magnetic levitation oxidation fan: The adjustable guide vane energy-saving integrated magnetic levitation oxidation fan and blower control system adjust the output of the oxidation fan based on parameters such as sulfur dioxide inlet concentration, sulfur dioxide outlet concentration, fan control parameters, and slurry oxidation state parameters, using the DTW-CNN-LSTM prediction model and DO-ORP deviation feedback model. Specifically, the air volume is controlled by adjusting the current opening of the adjustable guide vane 31 of the magnetic levitation fan and the current motor speed of the magnetic levitation oxidation fan 32. At the same time, the air volume of each absorption tower slurry pool zone is adjusted by the slurry oxidation-reduction potential value and dissolved oxygen value measured by multi-point sampling of different zones through an integrated measuring device, and the electric regulating valve 37 of the oxidation air branch pipe of each corresponding zone is adjusted to ensure the oxidation efficiency of each zone.

[0080] This application provides a DTW-CNN-LSTM prediction model. Figure 4 This is a schematic diagram of the architecture of the DTW-CNN-LSTM prediction model in the embodiments of this application. Figure 1 See Figure 4 , Figure 4 The five core processing modules and data flow of the DTW-CNN-LSTM prediction model are shown from left to right:

[0081] The input layer is located at Figure 4 On the far left, the input layer receives the 3D feature matrix as model input. Figure 4 The numbers 1 to 12 arranged vertically in the middle represent the 12 input feature dimensions contained in the three-dimensional feature matrix. Figure 4 The waveform curves in the figure represent time-series data formed by the changes in the values ​​of the 12 input feature dimensions over time;

[0082] The DTW timing alignment module is located on the right side of the input layer. Figure 4 The middle label is DTW time alignment. The DTW time alignment module receives the three-dimensional feature matrix output by the input layer. The DTW time alignment module eliminates the response lag time between the fan control parameters and the slurry oxidation state parameters through the dynamic time warping algorithm. The DTW time alignment module outputs the time synchronization sequence.

[0083] The CNN local feature extraction module is located to the right of the DTW temporal alignment module. Figure 4 The text is labeled as Convolutional Neural Network and CNN Local Feature Extraction. The CNN Local Feature Extraction module receives the time synchronization sequence output by the DTW time alignment module. The CNN Local Feature Extraction module performs one-dimensional convolution operation on the time synchronization sequence. The CNN Local Feature Extraction module extracts the local correlation features between the control parameters of the magnetic levitation oxidation fan and the slurry oxidation state parameters within a short time window.

[0084] The LSTM long-term dependency capture module is located to the right of the CNN local feature extraction module. Figure 4 The text is labeled as Long Short-Term Memory Network and Long-Term Dependency Capture. The LSTM Long-Term Dependency Capture module receives the local correlation features output by the CNN Local Feature Extraction module. The LSTM Long-Term Dependency Capture module performs temporal modeling on the time synchronization sequence through the Long Short-Term Memory Network. The LSTM Long-Term Dependency Capture module captures the cumulative influence features of the control parameters of the magnetic levitation oxidation fan on the slurry oxidation state parameters over a long period of time.

[0085] The fully connected prediction layer is located in Figure 4 The far right, Figure 4The text is labeled as a fully connected prediction layer and a prediction output layer. The fully connected prediction layer receives short-term local correlation features output by the CNN local feature extraction module and long-term cumulative trend features output by the LSTM long-term dependency capture module. The fully connected prediction layer performs feature fusion processing on the short-term local correlation features and long-term cumulative trend features. The fully connected prediction layer outputs regional dissolved oxygen prediction values, regional redox potential prediction values, and magnetic levitation oxidation fan outlet pressure prediction values.

[0086] Figure 4 The arrows between the modules indicate the direction of data transmission. Data flows sequentially from the input layer through the DTW temporal alignment module, the CNN local feature extraction module, and the LSTM long-term dependency capture module, and finally reaches the fully connected prediction layer. The fully connected prediction layer outputs the prediction result, and the DTW-CNN-LSTM prediction model forms a complete monitoring-control prediction closed loop.

[0087] Figure 5 This is a schematic diagram of the architecture of the DTW-CNN-LSTM prediction model in the embodiments of this application. Figure 2 See Figure 5 , Figure 5 The parameter configuration and data processing of each processing layer of the DTW-CNN-LSTM prediction model are illustrated in flowchart form:

[0088] 1) Input layer:

[0089] Input features: an 8-dimensional time-series feature matrix, with dimensions of 30×8 (time steps × number of features), where:

[0090] Fan control parameters (2D): Current motor speed (r / min), current guide vane opening (%);

[0091] Slurry oxidation state parameters (5 dimensions): total dissolved oxygen (i.e., total DO, unit mg / L), total oxidation-reduction potential (i.e., total ORP, unit mV), first dissolved oxygen deviation (unit mg / L, i.e., the difference between the dissolved oxygen value at each oxidation state monitoring point and the total dissolved oxygen value in the slurry pool of the absorption tower), first oxidation-reduction potential deviation (unit mV, i.e., the difference between the oxidation-reduction potential value at each oxidation state monitoring point and the total oxidation-reduction potential value in the slurry pool of the absorption tower), pH value;

[0092] Operating parameters (1-dimensional): Sulfur dioxide inlet concentration and sulfur dioxide outlet concentration are used as indicators of overall desulfurization efficiency (units). ).

[0093] Time step: 30 minutes (1 minute per step), covering the typical disturbance cycle of wet desulfurization system.

[0094] 2) DTW timing alignment module:

[0095] Eliminate the 1-2 minute time lag between the fan control parameters and the slurry oxidation state parameters, and output a time synchronization sequence.

[0096] Technical principle: Figure 6 This is a schematic diagram illustrating the principle of DTW timing sequence alignment in an embodiment of this application. (See attached diagram.) Figure 6 , Figure 6 The left side shows the original sequence with a 2-minute lag. Figure 6 The right side shows the synchronization sequence after DTW alignment. The arrows indicate the optimal regularization path, ensuring accurate time correlation between operation and response. Specifically, two types of time-series data are defined as: the magnetic levitation oxidation fan operation sequence (e.g., current motor speed). ,in The current motor speed (r / min) at minute i; the slurry oxidation state sequence (e.g., DO). ,in Let be the DO value (mg / L) at minute j. The DTW algorithm eliminates lag by constructing a time-warped path. The core steps are as follows: Step 1, construct the distance matrix: calculate the Manhattan distance between each data point in X and Y, using the following formula: This forms a 30×30 distance matrix D, where Step two, finding the optimal regular path: using dynamic programming to find the path from... arrive Minimum cumulative distance path The following constraints must be met: Path continuity constraint is... and To avoid jumping across time points; the path monotonicity constraint is... and Maintain chronological order; maximum offset constraint is The maximum delay is limited to 2 minutes. The cumulative distance formula is: Where t ranges from 1 to k. In the process of finding the optimal regularized path, the DTW algorithm uses dynamic programming to construct the cumulative distance matrix. Cumulative distance matrix elements in This represents the distance from the starting point (1,1) to the current point. The minimum cumulative distance is calculated recursively using the following formula: ,in, The corresponding vertical movement indicates that the slurry sequence lags behind the operation sequence; The corresponding horizontal shift indicates that the operation sequence lags behind the slurry sequence; A diagonal shift indicates that the two sequences have no lag. The boundary conditions are set as follows: , , To limit the maximum lag time to no more than 2 minutes, only those satisfying the condition are calculated in the cumulative distance matrix. Elements that do not meet the constraints are set to infinity to avoid physical and logical distortion caused by excessive path offset. The path backtracking phase starts from the endpoint of the cumulative distance matrix. Starting from the beginning, backtrack gradually along the direction of minimum cumulative distance to the starting point (1,1) to obtain the optimal regular path. If the path length k is less than the original time step of 30, linear interpolation is used to complete the aligned sequence to ensure the consistency of the input dimensions. Step 3, Data Alignment Output: Adjust X and Y into time-synchronized sequences according to the optimal path P. and To ensure the elimination of lag interference. Distance metric selection: Manhattan distance is used instead of Euclidean distance because data such as DO and rotational speed in industrial scenarios are easily affected by instantaneous noise. Manhattan distance is less sensitive to outliers and is suitable for the complex operating conditions of the wet desulfurization absorption tower slurry pool. Path constraint: The maximum time offset is set to 2 (i.e., a maximum lag alignment of 2 minutes is allowed). According to field experiments, after the rotational speed of the magnetic levitation oxidation fan is adjusted, the response lag of the dissolved oxygen value in the slurry does not exceed 2 minutes. This path constraint can avoid ineffective long-distance regularization.

[0097] To address the multi-sequence alignment requirements of 8-dimensional input features, the DTW time-series alignment module employs a group alignment + parallel computing strategy to improve processing efficiency. In the group alignment stage, the input features are divided into an operation group (including current motor speed and current guide vane opening) and a response group (including total dissolved oxygen, total redox potential, first dissolved oxygen deviation, first redox potential deviation, sulfur dioxide inlet concentration, sulfur dioxide outlet concentration, and pH). Using the operation group as a reference, DTW alignment is performed with each feature sequence in the response group, ensuring that all response features are time-synchronized with the operation features. In the parallel computing stage, the distance matrix and optimal path of multiple feature sequences are calculated in parallel using GPU multi-threading, reducing the alignment processing time from 0.8 seconds / sample in serial computing to 0.1 seconds / sample, meeting the real-time requirements of the blower control system.

[0098] 3) CNN Local Feature Extraction Module (capturing short-term parameter correlations):

[0099] Extract the short-term local correlation characteristics between the blower control parameters and the slurry oxidation state parameters (such as the trend of the first dissolved oxygen deviation value within 3 minutes after the current motor speed changes). Figure 7 This is a schematic diagram of a CNN local feature extraction module in an embodiment of this application. (See attached diagram.) Figure 7This demonstrates the hierarchical structure of a CNN. 1D convolutional kernels slide along the time axis to extract local features, max-pooling layers compress dimensionality and retain key peaks (such as DO mutation points), and the final output is a flattened feature adapted to LSTM. Specifically, the network structure of the CNN local feature extraction module is shown in Table 1 below:

[0100] Table 1 Parameter Table of CNN Local Feature Extraction Module

[0101]

[0102] The computational principle of 1D convolution operation in the CNN local feature extraction module is as follows: For the input temporal feature map... ,in, This represents an L-row, C-column real matrix, where L is the time step and C is the number of feature channels, using a convolutional kernel. Perform convolution operations, where k is the kernel size, and output feature maps. The calculation formula is: Where l is the temporal position index of the output feature map, ranging from 1 to L-k+1; i is the temporal offset index within the convolutional kernel, ranging from 0 to k-1; and c is the input feature channel index, ranging from 1 to C. σ represents the output feature channel index; σ is the ReLU activation function. This represents the value of the input feature map at time position l+i and channel c; This indicates that the convolution kernel has a time offset of i, an input channel of c, and an output channel of c. The weight value at the location; For output channel The bias is calculated by sliding the convolution kernel along the time axis, weighting and summing the features within the local time window to extract locally correlated features. Max pooling is used instead of average pooling because in industrial time series data, the peaks of local features (such as abrupt changes in dissolved oxygen values ​​and rapid fluctuations in redox potential values) have a greater impact on the prediction results. Max pooling can effectively preserve these key peak features, while average pooling smooths the peaks, leading to feature distortion. Input layer: Receives time series data aligned by DTW, with a dimension of 30×8 (time step × number of features). The time step 30 corresponds to 30 minutes of historical data, which can cover the typical disturbance cycle of the wet desulfurization system. The number of features 8 includes fan control parameters (current motor speed, current guide vane opening), slurry oxidation state parameters (total dissolved oxygen value, total redox potential value, first dissolved oxygen deviation value, first redox potential deviation value, pH value), and operating condition parameters (sulfur dioxide inlet concentration and sulfur dioxide outlet concentration are used as overall indicators of desulfurization efficiency).

[0103] The first convolutional layer employs 64 1D convolutional kernels with a kernel size of 3, a stride of 1, and the ReLU activation function. The 1D convolutional kernels are well-suited to the characteristics of time-series data, and the kernel size of 3 captures the local correlation between the current time point and one time point before and after it, matching the short-term response cycle of the oxidation state. The ReLU activation function addresses the vanishing gradient problem, adapting to the nonlinear characteristics of industrial time-series data.

[0104] First pooling layer: Max pooling is used, with a pooling kernel size of 2 and a stride of 2. Max pooling can preserve key peak information in local features (such as the mutation point of DO and the rapid fluctuation of ORP). After pooling, the time step is halved, reducing the amount of subsequent computation.

[0105] The second convolutional layer uses 128 1D convolutional kernels with a kernel size of 3, a stride of 1, and the activation function ReLU. This second layer uses more kernels to increase feature complexity and avoid overfitting or underfitting.

[0106] The second pooling layer uses max pooling with a kernel size of 2 and a stride of 2 to further compress the dimensionality and focus on core local features.

[0107] Feature flattening layer: After two convolutions and pooling, the time step is compressed from 30 to 6, and finally the 6×128 feature matrix is ​​flattened into a 768-dimensional feature vector to meet the input requirements of the subsequent LSTM module.

[0108] 4) LSTM Long-Term Dependency Capture Module:

[0109] Capture long-term trends (such as the cumulative effect of fan control parameters on slurry oxidation state parameters over 2 consecutive hours). Figure 8 This is a schematic diagram of a long-term dependency capture module of LSTM in an embodiment of this application. (See attached diagram.) Figure 8 The hierarchical structure of a two-layer LSTM. Figure 8 This demonstrates the two-layer LSTM hierarchy and data flow that LSTM relies on a long-term capture module. Figure 8 From left to right, the diagram shows the input of the LSTM long-term dependency capture module, the two-layer LSTM processing unit, and the output: Figure 8 On the left is the input of the LSTM long-term dependency capture module. The input receives the 768-dimensional short-term local correlation feature vector output by the CNN local feature extraction module. This feature vector is expanded by time steps to form a temporal input sequence. Figure 8 The central part is the core processing unit of the two-layer LSTM structure, where the first layer LSTM is located in Figure 8 In the upper middle section, the first LSTM layer contains 128 hidden units. The first LSTM layer receives the input sequence and outputs a complete temporal feature sequence. Figure 8The output arrow of the first LSTM layer points to the second LSTM layer, indicating that the first LSTM layer passes the complete sequence features to the second LSTM layer; the second LSTM layer is located in... Figure 8 In the lower middle section, the second LSTM layer contains 64 hidden units. The second LSTM layer receives the complete sequence features output by the first LSTM layer. The second LSTM layer only outputs the hidden state of the last time step as the long-term cumulative trend feature. Figure 8 Each LSTM layer contains multiple LSTM cell units that unfold according to time steps. Figure 8 The horizontal arrows between adjacent LSTM cell units represent the transfer of hidden state h and cell state C between time steps. Figure 8 The vertical arrow in the middle indicates that the input data for the current time step is flowing into the LSTM cell. Figure 8 On the right is the output of the LSTM long-term dependency capture module, which outputs a 64-dimensional long-term cumulative trend feature vector. This feature vector will be concatenated and fused with the 768-dimensional short-term local correlation feature vector output by the CNN local feature extraction module, and then input together into the fully connected output layer. Figure 8 The two-layer LSTM structure is designed as follows: The first LSTM layer uses 128 hidden units to provide sufficient feature representation capabilities. The first LSTM layer is configured as follows: True outputs the complete temporal feature sequence for further processing by the second LSTM layer. The second LSTM layer uses 64 hidden units to compress the feature dimension and focus on the core information of long-term trends. The second LSTM layer only outputs the final hidden state to obtain a comprehensive representation of the entire input sequence. Compared with the single-layer LSTM, the dual-layer LSTM structure has a stronger ability to capture long-term dependencies and can adapt to the long-cycle prediction requirements of 30-60 minutes for wet desulfurization systems.

[0110] Figure 9 This is a schematic diagram of the internal structure of an LSTM cell unit in an embodiment of this application. (See attached diagram.) Figure 9 The LSTM long-term dependency capture module implements long-term dependency capture through a gating mechanism. Let the input at time t (i.e., the current time) be... (Local features output by CNN, 768 dimensions), the hidden state at the previous time step (t-1) was: (64-dimensional), the cell state at the previous time step was (64-dimensional), the calculation formulas for each component are as follows: Forget gate (filtering historical information): determines how much of the cell state from the previous time step is retained. The formula is ,in, Let be the forget gate output vector (64-dimensional) at time t, with values ​​ranging from 0 to 1; The forget gate weight matrix has a dimension of 64×832 (832=768+64, corresponding to the input vector). (the splicing dimension) The hidden state of the previous moment With the current input splicing; σ is the forget gate bias vector (64-dimensional); σ is the Sigmoid activation function, which compresses the output to the range of 0-1, where 0 represents complete forgetting and 1 represents complete retention; This indicates that the previous state will be hidden. With the current input Perform vector concatenation. In wet desulfurization, historical oxidation states (such as total dissolved oxygen value one hour ago) influence the current trend. The forget gate can dynamically adjust the weight of historical information to avoid interference from irrelevant historical data. Input gate (filtering current information): determines how much of the current input x is updated. t The information is transmitted to the cell state, and the formula is: , ,in, The input gate output vector (64-dimensional) at time t has a value range of 0-1, representing the updated weights; The input gate weight matrix has a dimension of 64×832; The input gate bias vector (64-dimensional); The current input candidate cell state vector (64-dimensional) has a value range from -1 to 1; The candidate state weight matrix has a dimension of 64×832; The candidate state bias vector (64-dimensional); The hyperbolic tangent activation function compresses the output to the range of -1 to 1 to generate candidate cell states. The impact of current fan control parameters (such as a sudden adjustment in the current lead vane opening) on ​​the oxidation state needs to be prioritized; the input gate can strengthen the weight of key current information. Cell state update (long-term memory): Long-term cell states are updated by combining the forget gate and the input gate, using the following formula: ,in, The cell state vector at time t (64-dimensional) serves as the long-term memory unit of the LSTM. This represents the Hadamard product (element-wise multiplication), which is the multiplication of corresponding elements of two vectors of the same dimension. This indicates selective forgetting of the cell's state at the previous moment; This indicates that the candidate cell state is selectively updated. It can accumulate several hours of operation-oxidation correlation information, adapting to the slow dynamic characteristics of wet desulfurization oxidation reactions (long DO and ORP change cycles). Output gate (outputs the current state): determines the cell state. How much information should be output to the hidden state? The formula is , ,in, The output gate output vector (64-dimensional) at time t has a value range of 0-1 and controls the output ratio of cell state. The output gate weight matrix has a dimension of 64×832; The output gate bias vector (64-dimensional); The hidden state vector (64-dimensional) at time t is used as the output of the LSTM and passed to the next LSTM layer or fully connected layer. After compressing the cell state to a range of -1 to 1, through Selective output is performed. The output gate filters out redundant information from the cell state, ensuring that the features passed to the fully connected layer are more focused on predicting relevant long-term trends.

[0111] The LSTM network, which relies heavily on the capture module, employs a two-layer LSTM structure. The first layer contains 128 hidden units, and the second layer contains 64 hidden units. The first layer outputs complete sequence features (…). The second layer outputs the final hidden state (64 dimensions). The two-layer structure can enhance the ability to capture long-term dependencies and is suitable for long-term prediction needs of 30-60 minutes.

[0112] 5) Fully connected output layer:

[0113] The fully connected output layer takes as input the 64-dimensional hidden state output by the LSTM; the output consists of three types of predictions: region-specific predictions. (i.e., regional dissolved oxygen prediction value 1), regional (i.e., regional dissolved oxygen prediction value 2); regional (i.e., predicted redox potential values ​​1 for different regions), different regions (i.e., predicted redox potential values ​​for different regions 2); fan outlet (i.e., the predicted value of the wind turbine outlet pressure); the activation function is Linear (without range compression, suitable for regression scenarios).

[0114] Because the model training data is compressed to the [0,1] range using the Min-Max normalization method, the predicted values ​​output by the fully connected output layer are normalized values, which need to be restored to the actual physical quantities using an inverse normalization algorithm. The inverse normalization calculation formula is as follows: ,in, The predicted value after inverse normalization. The normalized predicted value is the output of the fully connected output layer. and These represent the maximum and minimum values ​​of each predicted target in the training set. Through inverse normalization, the output results are ensured to conform to the physical range of industrial parameters. For example, the output range for dissolved oxygen is 0-10 mg / L, the output range for redox potential is -200 to +200 mV, and the output range for outlet pressure is 10-20 kPa.

[0115] Based on the above explanation of the structure and working principle of each processing module in the DTW-CNN-LSTM prediction model, the DTW-CNN-LSTM prediction model is the core hub connecting the wet desulfurization oxidation state monitoring and the coordinated control of the magnetic levitation oxidation blower. It aims to solve the problems of traditional models being unable to simultaneously address time lag elimination, multi-parameter correlation extraction, and long-term trend prediction. This DTW-CNN-LSTM prediction model eliminates the 1-2 minute response lag between blower control parameters and slurry oxidation state parameters through DTW time-series alignment, extracts short-term local correlation features through CNN, and captures long-term cumulative trend features through LSTM. Ultimately, it achieves accurate prediction of the dissolved oxygen value, redox potential value, and magnetic levitation oxidation blower outlet pressure in different regions of the slurry within the next 30-60 minutes, providing core decision-making basis for adjusting the current guide vane opening and motor speed of the magnetic levitation oxidation blower.

[0116] This application also provides a method for training and validating a DTW-CNN-LSTM prediction model, including the following steps:

[0117] 1) Training data sources and preprocessing:

[0118] Data source: Historical operating data of a 300MW thermal power plant's wet desulfurization system for one year (or 0.5, 0.9, or 1.5 years, not limited here) was selected. For example, operating parameters: sulfur dioxide inlet concentration (200-800... ), sulfur dioxide outlet concentration (10-35) Slurry oxidation state parameters: total dissolved oxygen (0-2 mg / L), total oxidation-reduction potential (-200~-100 mV), first dissolved oxygen deviation (±0.5 mg / L), first oxidation-reduction potential deviation (±20 mV), pH (4.5-5.5); Fan control parameters: current motor speed (1800-5500 r / min), current guide vane opening (35%-95%), outlet pressure (14.5-15.5 kPa); Data sampling frequency: 1 time / minute, totaling approximately 525,600 data points.

[0119] Data preprocessing:

[0120] Outlier removal: Data from the start-up and shutdown phases of the magnetic levitation oxidation fan (current motor speed <1800r / min) and sensor malfunctions (total dissolved oxygen value >2mg / L or <0) were deleted, retaining approximately 480,000 valid data entries;

[0121] Normalization: The Min-Max normalization method is used to preprocess the data. The normalization formula is as follows: Where X is the original data value, These are the normalized data values. This is the minimum value of the parameter in the training set. The maximum value of this parameter in the training set is used to compress the normalized data to the range of [0,1], so as to avoid the difference in magnitude of different parameters from affecting the model training.

[0122] Deviation value calculation: The first dissolved oxygen deviation value is determined based on the difference between the dissolved oxygen value at each oxidation state monitoring point and the total dissolved oxygen value; the first oxidation-reduction potential deviation value is determined based on the difference between the oxidation-reduction potential value at each oxidation state monitoring point and the total oxidation-reduction potential value.

[0123] Dataset partitioning: The dataset was divided into a training set (336,000 records), a validation set (96,000 records), and a test set (48,000 records) in a 7:2:1 ratio, with the partitioning occurring consecutively over time to avoid data leakage.

[0124] 2) Training parameters and convergence criteria:

[0125] Optimizer: The Adam optimizer is used with a learning rate of 0.001. The Adam optimizer uses adaptive moment estimation to update model weights, which avoids the slow convergence problem caused by differences in parameter magnitudes during training on industrial data. The core parameters of the Adam optimizer are set as follows: initial learning rate... First-order moment estimation of attenuation rate Second-order moment estimation of attenuation rate Numerical stability parameters Simultaneously, a learning rate decay strategy is employed, reducing the learning rate to 0.9 times its original value every 20 training epochs to improve training stability in the later stages of model convergence. The weight update formula for the Adam optimizer is as follows: First, the first-order moment estimate is calculated. ,in Let be the first moment estimate for the t-th round. This is the estimate of the first moment in the (t-1)th round. Let be the gradient of round t. Estimate the attenuation rate for the first moment; then calculate the second moment estimate. ,in Let be the estimated value of the second moment in round t. This is the estimate of the second moment in the (t-1)th round. The square of the gradient. Estimate the attenuation rate for the second moment; then perform bias correction. and ,in This is the first-order moment estimate after bias correction. This is the second moment estimate after bias correction. for t to the power of for The power of t; last update the weights ,in Let be the weights after the t-th round update. Let be the weights for the (t-1)th round, α be the learning rate, and ε be a numerical stability parameter to prevent the denominator from being zero. Adam's adaptive learning rate characteristic avoids the slow convergence problem caused by differences in parameter magnitudes during industrial data training.

[0126] Loss function: Mean squared error (MSE) is used, and the formula is as follows: Where Loss is the loss function value, N is the number of samples in the batch, and i is the sample index. Let i be the true value of the i-th sample. Let be the predicted value for the i-th sample. Let be the square of the prediction error for the i-th sample. MSE penalizes prediction errors more heavily for larger deviations, which is suitable for the need in industrial scenarios to avoid excessive DO / ORP prediction deviations that could lead to fan control errors.

[0127] Training rounds: 80 rounds, using an early stopping strategy. Training is stopped when the validation set loss does not decrease for 5 consecutive rounds to avoid overfitting.

[0128] Batch size: 64. This batch size allows the model to run stably on an industrial computer (16GB memory) while ensuring the stability of parameter updates. To avoid gradient explosion during deep LSTM network training, a gradient norm clipping strategy is used to limit the gradients. The maximum gradient norm is set to 5.0, and the gradient clipping formula is as follows: ,in, This is the original gradient vector. This is the clipped gradient vector. Let g be the L2 norm of the gradient g. The maximum gradient norm threshold, Indicates taking 1 and The smaller value in the range is used as the scaling factor. When the L2 norm of the gradient exceeds the threshold, the gradient is scaled proportionally to the threshold range to ensure the numerical stability of the model training process.

[0129] 3) Model performance verification:

[0130] The model's prediction accuracy was validated using a test set, employing mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (CME). The evaluation is based on four indicators: mean absolute error, 30-minute prediction accuracy, and 30-minute prediction accuracy. The formulas for each indicator are as follows: Where N is the total number of samples in the test set, and i is the sample index. Let be the predicted value for the i-th sample. Let i be the true value of the i-th sample. The root mean square error (RMSE) is the absolute value of the prediction error for the i-th sample. The MAE reflects the average deviation between the predicted and actual values. The meanings of the parameters are the same as above. The RMSE is the square of the prediction error for the i-th sample. It can amplify the impact of larger errors to highlight extreme biases; the coefficient of determination... ,in The arithmetic mean of the true values ​​of all samples in the test set. Let be the square of the difference between the true value and the mean of the true values ​​for the i-th sample. This reflects the model's ability to explain variance. A value closer to 1 indicates a better model fit. The 30-minute prediction accuracy is the percentage of samples where the error between the predicted and actual values ​​is less than the industrial allowable error, specifically 0.2 mg / L for dissolved oxygen, 10 mV for redox potential, and 0.1 kPa for outlet pressure. The DTW-CNN-LSTM prediction model is compared with traditional models (single LSTM, single CNN), and the results are shown in Table 2 below.

[0131] Table 2 Performance Comparison of DTW-CNN-LSTM Prediction Models

[0132]

[0133] Note: Smaller MAE and RMSE values ​​indicate higher prediction accuracy. A value closer to 1 indicates a better model fit. Accuracy is the percentage of samples where the error between the predicted and actual values ​​is less than the industrial allowable error, specifically 0.2 mg / L for dissolved oxygen and 10 mV for redox potential. Data is derived from statistical results of 48,000 data points in the test set. The DTW-CNN-LSTM prediction model significantly outperforms traditional models in all of the above prediction metrics, demonstrating its synergistic advantages in eliminating time lag, extracting multi-parameter correlations, and capturing long-term trends, fully meeting the accuracy requirements for wet desulfurization magnetic levitation oxidation fan control.

[0134] The blower control system in the embodiments of this invention is described below from the perspective of hardware processing. (See attached document.) Figure 10 , Figure 10 This is a schematic diagram of the physical device structure of a blower control system in an embodiment of this application.

[0135] It should be noted that, Figure 10 The structure of the blower control system shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0136] like Figure 10 As shown, the blower control system includes a Central Processing Unit (CPU) 1001, which can perform various appropriate actions and processes according to a program stored in Read-Only Memory (ROM) 1002 or a program loaded from storage section 1008 into Random Access Memory (RAM) 1003, such as executing the methods described in the above embodiments. The RAM 1003 also stores various programs and data required for system operation. The CPU 1001, ROM 1002, and RAM 1003 are interconnected via a bus 1004. An Input / Output (I / O) interface 1005 is also connected to the bus 1004.

[0137] The following components are connected to I / O interface 1005: input section 1006 including audio input devices, push-button switches, etc.; output section 1007 including a liquid crystal display (LCD) and audio output devices, indicator lights, etc.; storage section 1008 including a hard disk, etc.; and communication section 1009 including a network interface card such as a LAN (Local Area Network) card, modem, etc. Communication section 1009 performs communication processing via a network such as the Internet. Drive 1010 is also connected to I / O interface 1005 as needed. Removable media 1011, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 1010 as needed so that computer programs read from them can be installed into storage section 1008 as needed.

[0138] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing computer programs for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 1009, and / or installed from removable medium 1011. When the computer program is executed by central processing unit (CPU) 1001, it performs the various functions defined in the present invention.

[0139] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0140] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Each block in a flowchart or block diagram may represent a module, program segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those shown in the drawings.

[0141] Specifically, the blower control system of this embodiment includes a processor and a memory. The memory stores a computer program. When the computer program is executed by the processor, it implements the blower control method of the wet desulfurization system provided in the above embodiment.

[0142] In another aspect, the present invention also provides a computer-readable storage medium, which may be included in the blower control system described in the above embodiments; or it may exist independently and not assembled into the blower control system. The storage medium carries one or more computer programs that, when executed by a processor of the blower control system, cause the blower control system to implement the blower control method for the wet desulfurization system provided in the above embodiments.

[0143] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0144] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A method for controlling the blower air in a wet desulfurization system, characterized in that, include: Multiple oxidation state monitoring points are set in the slurry pool of the absorber tower of the wet desulfurization system to collect the slurry oxidation state parameters of each oxidation state monitoring point in real time, and to obtain the operating condition parameters of the wet desulfurization system and the fan control parameters of the magnetic levitation oxidation fan of the wet desulfurization system. During the operation of the wet desulfurization system, the fan control parameters, slurry oxidation state parameters, and operating condition parameters are input into the DTW-CNN-LSTM prediction model to obtain the predicted slurry state parameters output by the DTW-CNN-LSTM prediction model after performing a target airflow prediction operation on the fan control parameters, slurry oxidation state parameters, and operating condition parameters. Specifically, this includes inputting the fan control parameters, slurry oxidation state parameters, and operating condition parameters into the DTW-CNN-LSTM prediction model so that the DTW-CNN-LSTM prediction model performs the following target airflow prediction operation: The DTW-CNN-LSTM prediction model converts the wind turbine control parameters, the slurry oxidation state parameters, and the operating condition parameters into a three-dimensional feature matrix through the input layer; The DTW-CNN-LSTM prediction model performs time-series alignment processing on the three-dimensional feature matrix through the DTW time-series alignment module to obtain a time-synchronized sequence. The DTW-CNN-LSTM prediction model extracts correlation features from the time synchronization sequence through the CNN local feature extraction module to obtain short-term local correlation features. The DTW-CNN-LSTM prediction model extracts temporal features from the short-term local correlation features through the LSTM long-term dependency capture module to obtain long-term cumulative trend features; The DTW-CNN-LSTM prediction model performs feature mapping on the long-term cumulative trend features through a fully connected output layer to obtain the predicted slurry state parameters. The predicted air volume is determined based on the predicted slurry state parameters and the operating condition parameters. The DO-ORP deviation feedback model is used to perform airflow correction operation on the slurry oxidation state parameters to obtain the corrected airflow. Specifically, the DO-ORP deviation feedback model is used to perform a third deviation analysis on the total dissolved oxygen value and the total oxidation-reduction potential value in the slurry oxidation state parameters to determine the second dissolved oxygen deviation value between the total dissolved oxygen value and the preset dissolved oxygen threshold, and to determine the second oxidation-reduction potential deviation value between the total oxidation-reduction potential value and the preset oxidation-reduction potential threshold. The basic corrected air volume is determined based on the second dissolved oxygen deviation value and the second redox potential deviation value; The spatial distribution non-uniformity of slurry oxidation is determined based on the first dissolved oxygen deviation value and the first redox potential deviation value in the slurry oxidation state parameters. The spatial distribution non-uniformity is compared with a preset non-uniformity threshold to determine the compensation correction air volume when the spatial distribution non-uniformity is greater than the preset non-uniformity threshold. The basic corrected air volume is superimposed with the compensation corrected air volume to obtain the corrected air volume. If the corrected air volume is positive, the output air volume of the magnetic levitation oxidation fan needs to be increased. If the corrected air volume is negative, the output air volume of the magnetic levitation oxidation fan needs to be reduced. A target control air volume is generated based on the predicted air volume and the corrected air volume, and the target control air volume is used to perform operation control operations on the magnetic levitation oxidation fan.

2. The method according to claim 1, characterized in that, The method involves setting up multiple oxidation state monitoring points in the slurry pool of the absorber tower in the wet desulfurization system to collect the slurry oxidation state parameters at each monitoring point in real time, and to obtain the operating condition parameters of the wet desulfurization system and the fan control parameters of the magnetic levitation oxidation fan of the wet desulfurization system. Specifically, this includes: Multiple integrated measuring devices are arranged at different depths and radial orientations in the slurry pool of the absorption tower to obtain multiple oxidation state monitoring points. The dissolved oxygen value, oxidation-reduction potential value and pH value of each oxidation state monitoring point are collected in real time using the multiple integrated measuring devices. The total dissolved oxygen value of the absorber slurry pool is determined based on the dissolved oxygen value at each oxidation state monitoring point, and the total oxidation-reduction potential value of the absorber slurry pool is determined based on the oxidation-reduction potential value at each oxidation state monitoring point. A first deviation analysis is performed between the dissolved oxygen value at each oxidation state monitoring point and the total dissolved oxygen value to obtain a first dissolved oxygen deviation value for each oxidation state monitoring point. A second deviation analysis is performed between the oxidation-reduction potential value at each oxidation state monitoring point and the total oxidation-reduction potential value to obtain a first oxidation-reduction potential deviation value for each oxidation state monitoring point. The total dissolved oxygen value, the total oxidation-reduction potential value, the first dissolved oxygen deviation value, the first oxidation-reduction potential deviation value, and the pH value are used as the oxidation state parameters of the slurry; The sulfur dioxide inlet concentration is collected from the flue gas inlet of the absorption tower, and the sulfur dioxide outlet concentration is collected from the flue gas outlet of the absorption tower. The sulfur dioxide inlet concentration and the sulfur dioxide outlet concentration are used as the operating condition parameters. The current motor speed and current guide vane opening of the magnetic levitation oxidation fan are obtained, and the current motor speed and current guide vane opening are used as the fan control parameters.

3. The method according to claim 2, characterized in that, The step of generating a target control airflow based on the predicted airflow and the corrected airflow, and using the target control airflow to perform operation control operations on the magnetic levitation oxidation fan of the wet desulfurization system, specifically includes: The predicted air volume and the corrected air volume are subjected to air volume compensation processing to generate the target control air volume; Obtain the current outlet air volume and current outlet pressure of the magnetic levitation oxidation fan; The fan control strategy is determined based on the target control air volume, the current outlet air volume, and the current outlet pressure. The fan control strategy includes a guide vane adjustment strategy and a speed adjustment strategy. The target control air volume is compared with the current outlet air volume, and the current motor speed is compared with a preset speed threshold. When the target control air volume is less than the current outlet air volume and the current motor speed is less than the preset speed threshold, the guide vane adjustment strategy is executed; when the target control air volume is less than the current outlet air volume and the current motor speed is greater than or equal to the preset speed threshold, the speed adjustment strategy is executed. The current guide vane opening is compared with the preset limit opening. When the current guide vane opening reaches the preset limit opening, the speed adjustment strategy is executed; or when the target control air volume is greater than or equal to the current outlet air volume, the speed adjustment strategy is executed. A fourth deviation analysis is performed between the current outlet air volume and the target control air volume to obtain the deviation control air volume; The deviation control air volume is compared with a preset deviation threshold. When the deviation control air volume is greater than the preset deviation threshold, the current guide vane opening or the current motor speed is iteratively adjusted until the deviation control air volume is less than or equal to the preset deviation threshold.

4. The method according to claim 3, characterized in that, The step of executing the guide vane adjustment strategy when the target control air volume is less than the current outlet air volume and the current motor speed is less than a preset speed threshold specifically includes: Determine the air volume difference between the target control air volume and the current outlet air volume; The target guide vane opening is determined based on the air volume difference and the preset guide vane characteristic curve, wherein the guide vane characteristic curve characterizes the correspondence between the guide vane opening and the air volume change. The opening adjustment amount of the guide vane device of the magnetic levitation oxidation fan is determined based on the target guide vane opening and the current guide vane opening. When the opening adjustment amount is greater than the preset maximum single adjustment angle, the opening adjustment amount is decomposed into multiple adjustment steps, and the adjustment range of each adjustment step in the multiple adjustment steps is not greater than the preset maximum single adjustment angle. The guide vane device is controlled to adjust the current guide vane opening according to the multiple adjustment steps, so that the airflow entering the magnetic levitation oxidation fan generates a pre-swirl component in the same direction as the impeller rotation of the magnetic levitation oxidation fan. During the adjustment of the guide vane opening, the vibration and surge states of the magnetic levitation oxidation fan are monitored in real time. When the vibration value of the magnetic levitation oxidation fan is found to be greater than a preset vibration threshold or the surge margin of the magnetic levitation oxidation fan is less than a preset safety margin, the adjustment of the current guide vane opening is stopped.

5. The method according to claim 4, characterized in that, The step of executing the speed regulation strategy when the current guide vane opening reaches the preset limit opening or the target control air volume is greater than or equal to the current outlet air volume specifically includes: Determine whether the current opening of the leading vane has reached the preset limit opening. The preset limit opening includes a first opening limit and a second opening limit. The first opening limit is the minimum opening value allowed by the leading vane device, and the second opening limit is the maximum opening value allowed by the leading vane device. When the current guide vane opening reaches the first opening limit and the target control air volume is less than the current outlet air volume, the speed adjustment mode is activated; or when the current guide vane opening reaches the second opening limit and the target control air volume is greater than or equal to the current outlet air volume, the speed adjustment mode is activated; or when the target control air volume is greater than the current outlet air volume, the speed adjustment mode is activated. In the speed adjustment mode, the target speed is determined based on the target control air volume, the current motor speed, and the preset fan performance curve; A speed adjustment command is generated and sent to the magnetic levitation oxidation fan to adjust the current motor speed to the target speed; During the current motor speed adjustment process, the rate of change of the current motor speed is constrained to ensure that the rate of change does not exceed a preset change threshold. After determining that the current motor speed is adjusted to the target speed, the current outlet pressure is compared with the ultimate pressure requirement of the absorber slurry tank. If the current outlet pressure is determined to be less than the ultimate pressure requirement, the current motor speed is increased and the current guide vane opening is decreased until the current outlet air volume reaches the target control air volume and the current outlet pressure is greater than or equal to the ultimate pressure requirement.

6. A blower control system, characterized in that, The blower control system includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code including computer instructions, and the one or more processors call the computer instructions to cause the blower control system to perform the method as described in any one of claims 1-5.

7. A computer-readable storage medium comprising instructions, characterized in that, When the instruction is executed on the blower control system, the blower control system performs the method as described in any one of claims 1-5.

8. A computer program product, characterized in that, When the computer program product is run on the blower control system, it causes the blower control system to perform the method as described in any one of claims 1-5.

Images