A real-time monitoring device for steam admission flow of low-pressure cylinder of power plant heat supply steam turbine

Abstract

This invention discloses a real-time monitoring device for the steam inlet flow rate of the low-pressure cylinder of a power plant heating turbine, specifically relating to the field of flow monitoring. It includes a differential pressure signal generation module, an equivalent differential pressure calculation module, a dynamic flow calculation module, and a feedback maintenance module. The device achieves continuous self-cleaning of the pressure tap orifice and uninterrupted channel switching under fault conditions through the differential pressure signal generation module, avoiding signal distortion or interruption. The equivalent differential pressure calculation module compensates for the nonlinear influence of steam dryness variations on differential pressure measurement, adapting to dynamic operating conditions with frequent flow fluctuations, significantly improving dynamic measurement accuracy, and ensuring the real-time guidance value of monitoring results for unit safety early warning. The differential pressure signal generation module enables manual cleaning and self-maintenance of the pressure tap orifice, avoiding monitoring interruptions and ensuring continuous online real-time monitoring functionality.

Description

Technical Field

[0001] This invention relates to the field of flow monitoring technology, and more specifically, to a real-time monitoring device for the steam inlet flow of the low-pressure cylinder of a power plant heating turbine. Background Technology

[0002] With the in-depth development of cogeneration power plant heating units, high-precision real-time monitoring of the steam inlet flow rate of the turbine low-pressure cylinder is particularly important. Traditional technology mainly relies on indirect calculation methods: by measuring parameters such as main steam, reheat steam, and extraction steam at each stage, the steam inlet flow rate of the low-pressure cylinder is indirectly calculated. However, actual measurements show that under heating conditions, the turbine extracts a large amount of steam after the intermediate-pressure cylinder for external heating, resulting in a significant reduction in the steam flow rate entering the low-pressure cylinder, which is in a state of frequent and violent fluctuations. Due to the reliance on multiple remote measuring points and the long calculation chain, there are inherent defects such as large cumulative errors and slow response, making it difficult to capture rapid fluctuations. Furthermore, when the flow rate is too low, it cannot provide timely and accurate early warning signals to prevent overheating of the last stage blades of the low-pressure cylinder.

[0003] To overcome the shortcomings of traditional sensing systems, existing technologies employ a direct measurement approach: a dedicated scaling tube throttling element is directly installed in the steam inlet pipe of the low-pressure cylinder as a flow sensor. By measuring the static pressure difference before and after the throat of the throttling element, and directly calculating the steam mass flow rate based on classical fluid dynamics formulas, the real-time performance of monitoring is significantly improved, the response time is shortened, and the large-scale transmission of errors in multi-parameter indirect calculations is avoided.

[0004] However, in practical use, existing sensing systems still have some shortcomings. For example, the steam entering the low-pressure cylinder is often saturated or wet steam, which can easily clog the pressure tapping hole at the throat of the dilator tube due to entrained droplets. Traditional sensors cannot detect their own performance degradation and can only detect it when the signal is severely distorted or interrupted, resulting in the inability to guarantee the reliability of key data acquisition. Faced with frequent and rapid fluctuations in steam flow and dryness, the fixed classical flow model is not adaptable enough, and the sensor system itself has no analytical capability to achieve online correction of the model, resulting in a decrease in dynamic measurement accuracy. In order to maintain the measurement, manual cleaning or pneumatic purging is required periodically. The entire process requires interruption of monitoring because the existing system does not have the intelligent execution capability of self-maintenance or non-destructive switching, which seriously affects the continuous online function that real-time monitoring should have. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, the present invention provides a real-time monitoring device for the steam inlet flow of the low-pressure cylinder of a power plant heating turbine, which solves the problems mentioned in the background art through the following scheme.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A real-time monitoring device for the steam inlet flow rate of a low-pressure cylinder in a power plant heating turbine includes a processor and a memory. The memory stores a computer program, and when the computer program is executed by the processor, it further includes: Differential pressure signal generation module: used to acquire measurement signals through a sensor array installed in the steam inlet pipe of the low-pressure cylinder, the measurement signals including at least: A first signal characteristic characterizing fluid dynamic pressure generated by a preset throttling device, the throttling device including a main throttling element and a multi-throat array sensing unit; The second signal characteristic reflecting the steam state includes at least the pipeline static pressure signal, dielectric constant signal, and temperature distribution signal; Equivalent pressure difference calculation module: It is used to perform fusion calculation based on the first signal feature and the second signal feature through a preset wet steam two-phase flow correction model, and output the third signal feature after real-time compensation of steam dryness. Dynamic flow calculation module: Based on the third signal feature, it performs weighted calculation through a parallel benchmark calculation model and an adaptive learning model to output real-time steam mass flow rate. The benchmark calculation model is established based on fluid dynamics formulas, and the adaptive learning model is trained online using the third signal feature in the preceding time period as input features. Feedback maintenance module: used to continuously monitor and analyze the operating status metadata of each module to trigger the switching of the sensing unit in the differential pressure signal generation module, or to trigger the model relearning in the dynamic flow calculation module. The operating status metadata includes at least the signal attenuation rate of the multi-throat array sensing unit, the reading confidence of the second signal feature, and the output deviation value between the benchmark calculation model and the adaptive learning model.

[0007] Preferably, the differential pressure signal generation module, using a preset throttling device to generate a first signal feature, specifically includes: By continuously monitoring the differential pressure signal generated by the main throttling element, self-draining purging is continuously formed; The real-time quality indicators of the differential pressure signal are monitored in parallel, and the reachability of each of the multi-throat array sensing units is evaluated. The real-time quality indicators include at least the signal-to-noise ratio and attenuation rate of the signal. When the real-time quality index is lower than a preset threshold, the system automatically switches to at least one of the multi-throat array sensing units as a backup signal source to continuously obtain an effective differential pressure value.

[0008] Preferably, the differential pressure signal generation module uses a composite venturi tube structure to generate the first signal characteristic via a throttling device, comprising: A main Venturi tube, which serves as the main throttling element, has an annular steam groove on the inner wall of its throat. It is connected to the high-pressure tapping chamber upstream of the throat through a built-in self-draining channel, and is used to perform the self-draining purging. Additionally, at least two miniature Venturi tubes with different throat diameters, connected in parallel downstream of the main Venturi tube and serving as the multi-throat array sensing unit, are used to form the backup signal source.

[0009] Preferably, the equivalent pressure difference calculation module receives the dielectric constant signal from the second signal feature and defines its signal and dielectric constant of the medium. Relationship: in, , These represent the reference resonant frequency and the corresponding equivalent dielectric constant of the sensor in the cavity state, respectively. , These represent the measured resonant frequency shift and the change in quality factor, respectively. , These are respectively represented as calibration coefficients determined by the sensor's geometry; The dielectric constant Input the pre-stored steam dryness mapping model and output the real-time steam dryness estimate. Specifically, it is expressed as: in, These represent the current static pressure in the pipeline. The dielectric constants of pure saturated water vapor and pure saturated water at [temperature and temperature]. It is represented as the morphological distribution factor.

[0010] Preferably, the equivalent pressure difference calculation module dynamically calculates and outputs a pseudo-single-phase flow equivalent pressure difference after dryness compensation. This pseudo-single-phase flow equivalent pressure difference is expressed as the actual differential pressure generated when the current wet steam fluid flows through the throttling device, equivalently converted to the theoretical differential pressure value generated when single-phase superheated steam of the same mass flow rate flows through the same device under the same conditions. Specifically, it is expressed as follows: in, This represents the actual differential pressure of wet steam directly obtained through the throttling device. Indicated as the static pressure of the pipeline Related stress correction factors, This is expressed as a real-time estimate of steam dryness. , It is represented as the model exponential parameter.

[0011] Preferably, the dynamic flow calculation module obtains the real-time steam mass flow rate, specifically including: The third signal feature and the pipeline static pressure signal are input into the benchmark calculation model to calculate the first flow rate value. Simultaneously, the third signal feature sequence, pipeline static pressure sequence, and historical flow trend from the preceding time period are input into the adaptive learning model to calculate the second flow value; The adaptive learning model is a lightweight neural network model. The first flow rate value and the second flow rate value are weighted according to a real-time weighting coefficient to calculate the real-time steam mass flow rate.

[0012] Preferably, the dynamic traffic calculation module determines the real-time weighting coefficient, specifically including: Based on the fluctuation variance of the third signal feature within a preset time window, the first weight of the benchmark calculation model is determined by a preset monotonically decreasing function, and the second weight of the adaptive learning model is the difference between 1 and the first weight. When the variance of the fluctuation increases, the first weight decreases and the second weight increases accordingly.

[0013] Preferably, the feedback maintenance module is used to perform: Performance degradation prediction is based on the signal attenuation rate and the cumulative operating time and historical operating conditions of the corresponding sensing unit to predict the change in its effective aperture area. Calibration requirement assessment: Based on the confidence level of the second signal characteristic reading, determine the necessity of corresponding sensor calibration; Model relearning judgment: Based on the output deviation value between the benchmark model and the adaptive learning model, determine the necessity of model relearning.

[0014] The technical effects and advantages of this invention are as follows: 1. This invention achieves continuous self-cleaning of the pressure tapping orifice and uninterrupted channel switching under fault conditions through the differential pressure signal generation module, which solves the problem of pressure tapping orifice blockage, avoids signal distortion or interruption, and ensures the long-term continuity of monitoring data and the reliability of system operation. 2. This invention compensates for the nonlinear effect of steam dryness variation on differential pressure measurement through the equivalent differential pressure calculation module, adapts to dynamic operating conditions with frequent flow fluctuations, significantly improves dynamic measurement accuracy, and ensures the real-time guidance value of monitoring results for unit safety early warning. 3. This invention achieves self-maintenance of the pressure tapping hole without manual cleaning through the differential pressure signal generation module, avoiding interruption of the monitoring process and ensuring the continuous online function of real-time monitoring. Attached Figure Description

[0015] Figure 1 This is a block diagram of the computer program executed by the processor in a real-time monitoring device for the steam inlet flow of the low-pressure cylinder of a power plant heating turbine, according to an embodiment of this application.

[0016] Figure 2 This is a block diagram of the overall structure of a real-time monitoring device for the steam inlet flow of a low-pressure cylinder of a power plant heating turbine, according to an embodiment of this application.

[0017] Explanation of reference numerals in the attached drawings: 100, Overall block diagram of a real-time monitoring device for the steam inlet flow rate of a low-pressure cylinder of a power plant heating turbine; 101, Processor; 102, Memory; 103, Input device; 104, Acquisition device; 105, Output device; 106, Bus. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] 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 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 and includes any or all possible combinations of one or more of the listed items.

[0020] Hereinafter, the terms "first," "second," and "third" 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," "second," and "third" 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.

[0021] As attached Figure 1 The device shown is a real-time monitoring device for the steam inlet flow of the low-pressure cylinder of a power plant heating turbine. Its core includes at least one or more processors 101 and an associated memory 102. The memory 102 stores a computer program. When the computer program is executed by the processor 101, it also includes a differential pressure signal generation module, an equivalent differential pressure calculation module, a dynamic flow calculation module, and a feedback maintenance module.

[0022] The processor 101 is configured to execute the computer program configured in the real-time monitoring device for the steam inlet flow of the low-pressure cylinder of a power plant heating turbine provided in the following embodiments of this application. The processor 101 may be an embedded industrial control unit dedicated to the device, or a control processing module in a power plant distributed control system or programmable logic controller, or a device containing a central processing unit, a digital signal processor, or other forms of processing units with real-time data processing and complex algorithm execution capabilities. The processor 101 is responsible for processing the raw data from various sensors, executing core algorithms such as flow calculation, model learning and feedback maintenance, and controlling the various components of the entire monitoring device to work together to complete the desired monitoring function.

[0023] The memory 102 may include one or more computer programs, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include random access memory and / or cache memory, etc., for temporarily storing real-time processing data and running programs. The non-volatile memory may include read-only memory, solid-state drive, industrial-grade flash memory, etc., for long-term firmware storage. One or more computer program instructions are stored on the computer-readable storage medium, and the processor 101 can run the computer program instructions to realize real-time monitoring of the steam inlet flow of the low-pressure cylinder of the power plant heating turbine. The computer-readable storage medium may also store various data necessary for the operation of various applications and systems, including but not limited to the wet steam correction model parameters, neural network model weights, historical flow data, equipment health status logs, and various configuration thresholds used and / or generated by the real-time monitoring.

[0024] In one embodiment, a real-time monitoring device for the steam inlet flow rate of the low-pressure cylinder of a power plant heating turbine may further include an input device 103, an output device 105, and a data acquisition device 104, with each component interconnected via an internal bus 106. It should be noted that the components and structure of the monitoring device illustrated are merely exemplary and not limiting; the device may also have other components and structures as needed.

[0025] It should be noted that the input device 103 may be an interface for receiving user configuration commands or maintenance commands, and may include one or more of the following: an industrial keyboard, a touch screen, buttons, and a remote communication interface; the output device 105 is used to output monitoring results, early warning information, and system status to external systems or personnel, and may include one or more of the following: a display, status indicator lights, an industrial alarm, and a data communication interface for connecting to a power plant DCS or a plant-level monitoring information system; the acquisition device 104 is used to acquire physical signals in the low-pressure cylinder inlet pipe, and converts the acquired raw signals into electrical signals and sends them to the processor 101 or memory 102 for processing; for example, the acquisition device 104 may include, but is not limited to: a differential pressure transmitter array for acquiring differential pressure signals, a microwave resonant sensor for acquiring steam dryness-related signals, a thermocouple array for acquiring temperature distribution, and a pressure transmitter for acquiring pipeline static pressure.

[0026] In one embodiment, the components in the example device used to implement the embodiments of this application can be integrated or distributed; for example, the processor 101, memory 102, input device 103 and output device 105 can be integrated into an industrial chassis or control cabinet to form an independent monitoring station; while the acquisition device 104 can be separated and installed at the steam inlet pipeline of the steam turbine, and the two are connected by a bus 106.

[0027] In one embodiment, the example device for implementing the embodiments of this application can be implemented in various forms, including but not limited to: a dedicated function station integrated into the existing DCS or PLC system of a power plant, an integrated intelligent monitoring instrument cabinet installed on-site, or a hardware and software integrated solution deployed on an edge computing gateway.

[0028] The differential pressure signal generation module is used to acquire measurement signals through a sensor array installed in the steam inlet pipe of the low-pressure cylinder. The measurement signals include at least: a first signal feature characterized by fluid dynamic pressure generated by a preset throttling device, the throttling device including a main throttling element and a multi-throat array sensing unit; and a second signal feature reflecting the steam state, which includes at least a pipeline static pressure signal, a dielectric constant signal, and a temperature distribution signal.

[0029] It should be noted that after power-on, the main venturi tube is used as the initial signal source by default, and the differential pressure transmitters of each backup miniature venturi tube are automatically zeroed online. The acquisition of all channels is triggered by the same hardware clock to ensure data timestamp synchronization. If the differential pressure signal of the backup signal source does not reach the effective range within 2 seconds after the switching command is issued, the system will attempt to switch to the next backup signal source while keeping the main channel valve open as a backup.

[0030] In one possible implementation, generating the first signal feature includes: continuously monitoring the differential pressure signal generated by the main throttling element and continuously forming a self-draining purge; monitoring the real-time quality indicators of the differential pressure signal in parallel and evaluating the reachability of each of the multi-throat array sensing units, wherein the real-time quality indicators include at least the signal-to-noise ratio and attenuation rate of the signal; and automatically switching to at least one of the multi-throat array sensing units as a backup signal source when the real-time quality indicators are lower than a preset threshold, so as to continuously obtain an effective differential pressure value.

[0031] It should be noted that the throttling device is a composite Venturi tube structure, comprising: a main Venturi tube serving as the main throttling element, with an annular steam groove formed on the inner wall of its throat, which is connected to a high-pressure tapping chamber upstream of the throat via a built-in self-draining channel, so as to utilize the kinetic energy of the steam flowing through the throat to form a continuous self-draining purge at the tapping orifice, preventing droplets in the wet steam from adhering or condensate from accumulating; and at least two miniature Venturi tubes with different throat diameters connected in parallel downstream of the main Venturi tube, serving as the multi-throat array sensing unit, each equipped with an independent A differential pressure transmitter and isolation valve are used to form the backup signal source; the preset threshold is determined based on a large number of tests and the performance specifications of the differential pressure transmitter; in this embodiment, the signal-to-noise ratio threshold is set to 70% of the average value under normal operating conditions, and the attenuation rate threshold is set to 0.5% of the full scale per hour; the automatic signal source switching process includes: closing the isolation valve of the currently operating throttling element; opening the isolation valve of the backup miniature venturi tube and preheating its differential pressure transmitter; switching the data acquisition channel to the new transmitter, and sending a calibration coefficient update command to the main processor.

[0032] In this embodiment, the inlet end of the self-draining channel opens into the high-pressure tapping chamber, and the outlet end is connected to the annular steam tank in the form of multiple nozzles evenly distributed circumferentially or a continuous slit. This ensures that when the Venturi tube is working normally, the stable pressure difference between the high-pressure tapping chamber and the throat is sufficient to drive a continuous, minute amount of steam flow through the channel and eject it from the slit of the annular steam tank, thereby forming an effective positive air curtain barrier at the orifice of the high-pressure tapping hole. The self-draining channel is designed as a capillary structure with an inner diameter between 0.5 mm and 2 mm and a length-to-diameter ratio greater than 20. This ensures that under the typical pressure difference (5-50 kPa) between the throat and upstream of the Venturi tube, a minute amount of purging steam flow with a velocity in the range of 5-20 m / s can be generated. The effectiveness of the air curtain can be verified by observing that there are no obvious droplets or solid deposits adhering to the orifice of the high-pressure tapping hole and the surrounding wall during shutdown inspection.

[0033] It should be noted that the throat diameter of the miniature Venturi tube is configured according to the percentage of the design flow rate of the main Venturi tube. In this embodiment, three miniature Venturi tubes are set, and their throat diameters correspond to the 60%, 80%, and 100% design flow rate points of the main Venturi tube, respectively. Each miniature Venturi tube is calibrated on a standard device before leaving the factory to determine the relationship between the flow coefficient and the Reynolds number, and the result is stored in the processor. When switching, the flow coefficient of the corresponding unit is automatically called according to the Reynolds number of the current operating condition to calculate the flow rate, ensuring measurement continuity.

[0034] Furthermore, the pipeline static pressure acquisition in the second signal feature is directly obtained through an absolute pressure transmitter installed on the pipeline, providing the necessary working fluid state parameters for flow calculation; the dielectric constant signal is realized through an embedded microwave resonant sensor, which is non-invasively installed on the outer wall of the pipeline to emit microwaves of a specific frequency to the flowing steam and detect changes in the resonant frequency; the temperature distribution signal is realized through a fast-response thermocouple array, which consists of multiple armored thermocouples uniformly arranged circumferentially along the pipeline cross-section, capable of capturing the temperature field distribution of the steam cross-section, providing data support for judging flow stability and assisting in dryness compensation.

[0035] In this embodiment, the embedded microwave resonant sensor can be a vector network analysis unit based on phase-locked loop technology, with an operating frequency range of 8-12 GHz and a dryness measurement resolution better than 0.01; the fast-response thermocouple array uses K-type armored thermocouples with a time constant of less than 200 ms; the main venturi tube and each miniature venturi tube are connected to the main pipe of the steam inlet pipe through parallel branch pipes, sharing the same upstream inlet cross section and downstream outlet cross section; at the same time, the signal output terminals of each differential pressure transmitter, absolute pressure transmitter, microwave resonant sensor and thermocouple array all upload the first signal feature and the second signal feature in digital form to the processor 101 through bus 106.

[0036] In this embodiment, the device of the present invention was compared with the existing technology that uses a traditional single-diffuser tube throttling and fixed formula calculation on a test platform simulating typical wet steam conditions of a low-pressure cylinder in a power plant. The data shows that after about four weeks of continuous operation, the differential pressure signal attenuated by more than 15% due to droplet adhesion in the pressure tap of the existing technology, requiring shutdown for manual cleaning. However, the main Venturi tube of the present invention, with its self-draining and purging structure, maintained a signal attenuation rate of less than 3% for three months of continuous operation under the same conditions. When intelligently switching to the backup micro Venturi tube based on the signal-to-noise ratio and attenuation rate threshold, the entire process was completed within a few hundred milliseconds without data interruption, achieving a data acquisition continuity rate of over 99.9%, fundamentally solving the monitoring interruption problem caused by blockage and maintenance.

[0037] The equivalent pressure difference calculation module is used to perform fusion calculation based on the first signal feature and the second signal feature through a preset wet steam two-phase flow correction model, and output a third signal feature after real-time compensation of steam dryness.

[0038] Furthermore, the first and second signal features of the input are preprocessed to ensure data quality. This preprocessing includes at least: validity verification, checking whether each input signal is within a preset reasonable range and masking obviously abnormal jump values; temperature field uniformity judgment, analyzing the temperature readings of the thermocouple array and calculating its variance; if the variance is less than the threshold, the average value is taken as the current steam temperature; if the variance is too large, the flow state is marked as unstable and the adaptive weight is increased in subsequent calculations.

[0039] Furthermore, the dielectric constant signal, which is acquired in real time by the embedded microwave resonant sensor, is received from the second signal feature, and its relationship with the dielectric constant of the medium is defined. The relationship is usually determined by the sensor's own resonant characteristics, which can be obtained through calibration: in, , These represent the reference resonant frequency and the corresponding equivalent dielectric constant of the sensor in the cavity state, respectively. , These represent the measured resonant frequency shift and the change in quality factor, respectively. , These are respectively represented as calibration coefficients determined by the sensor's geometry, which are pre-determined experimentally; the dielectric constant is... Input the pre-stored steam dryness mapping model, output the current real-time steam mass dryness value, and use it as the estimated real-time steam dryness value. Specifically, it is expressed as: in, γ represents the dielectric constants of pure saturated steam and pure saturated water under the current pipeline static pressure Ps and temperature, respectively. γ represents the morphology distribution factor, which describes the influence of the distribution morphology of water droplets in steam on the overall dielectric properties. It was obtained through experimental calibration of a typical flow field of steam at the inlet of the low-pressure cylinder of a power plant.

[0040] It should be noted that the steam dryness mapping model is stored in memory 102 in the form of a two-dimensional lookup table, with its row index being the dielectric constant. The discrete values, with the column index being the discrete values of the pipeline static pressure Ps, and the data in the table being the corresponding steam dryness Dq determined through calibration experiments; its model construction is carried out by preparing a series of wet steam samples with known estimated values of steam dryness within the pressure and temperature ranges simulating the inlet conditions of the low-pressure cylinder in the laboratory; the known estimated values of steam dryness are accurately determined in advance by a reference measuring device combining the separator method and the heat balance method; the core principle of the reference measuring device is to obtain the liquid-phase mass through a high-precision separator and calculate the total enthalpy through heat balance, thereby inversely deriving the known estimated value of steam dryness; using the embedded microwave resonance sensor to measure the dielectric constant signal of the corresponding sample, and solving to obtain , and recording the current pipeline static pressure Ps; performing surface fitting on multiple groups of experimental data points of known estimated values of steam dryness, dielectric constants, and pipeline static pressures to generate a two-dimensional look-up table with and Ps as indices and outputting Dq.

[0041] In this embodiment, the morphological distribution factor is synchronously optimized and determined in the same set of experimental calibration data for constructing the steam dryness mapping model. Substituting multiple groups of experimental data points of known estimated values of steam dryness, dielectric constants, and pipeline static pressures into for non-linear regression fitting to minimize the overall error between the theoretically calculated dielectric constant value and the experimentally measured value according to the formula. The obtained value is the calibration value.

[0042] In a possible implementation manner, according to the pre-stored wet steam two-phase flow correction model, the differential pressure value in the first signal feature is fused with the real-time steam dryness estimated value and the pipeline static pressure signal in the second signal feature; dynamically calculating and outputting a dryness-compensated pseudo-single-phase flow equivalent differential pressure as the third signal feature; where the pseudo-single-phase flow equivalent differential pressure is expressed as converting the actual differential pressure generated when the current wet steam fluid flows through the throttling device into the theoretical differential pressure value generated when a single-phase superheated steam with the same mass flow rate flows through the same device under the same conditions, and is specifically expressed as: Where represents the actual differential pressure of the wet steam directly obtained through the throttling device, that is, the core value in the first signal feature, C(Ps) represents the pressure correction factor related to the pipeline static pressure Ps, used to compensate for the influence of pressure changes on the two-phase flow state, Dq represents the real-time steam dryness estimated value, 0 < Dq ≤ 1, mq represents the model exponential parameter, 0.​​It should be noted that the determination of the model exponential parameter mq and the pressure correction factor C(Ps) involves establishing a two-phase flow simulation model of the throttling device under different static pressures Ps and different inlet steam dryness fractions Dq using computational fluid dynamics software, obtaining multiple sets of wet steam differential pressures; on a physical flow calibration bench, wet steam with a known mass flow rate is passed through the throttling device, and the wet steam differential pressure is measured; the simulation and measured data are then substituted into the formula. The theoretical differential pressure calculated using the classical single-phase steam flow formula is used as... The target value is determined by jointly optimizing and fitting the parameters mq and C(Ps) to ensure that the calculated result of the formula best matches the target value across the entire operating range. The classical single-phase steam flow rate formula refers to the single-phase compressible fluid mass flow rate calculation formula specified in ISO 5167 and applicable to the throttling device. The joint optimization fitting employs the least squares method to ensure that the calculated result obtained by substituting simulation and measured data into the formula is optimal. The sum of squared residuals between the target value and the target value is minimized.

[0044] In this embodiment, the existing technology uses a single throttling element, and the pressure tapping hole is easily blocked by droplets in wet steam, resulting in signal distortion or interruption. The system reliability depends entirely on frequent manual maintenance. In contrast, this technology utilizes the kinetic energy of the fluid to achieve continuous self-cleaning of the pressure tapping hole from a physical structure perspective, fundamentally reducing the probability of blockage. It innovatively introduces a dielectric constant that is strongly related to dryness and solves the problem of unknown core variables by establishing a mapping relationship.

[0045] The dynamic flow calculation module is used to perform weighted calculations based on the third signal feature through a parallel benchmark calculation model and an adaptive learning model to output the real-time steam mass flow rate. The benchmark calculation model is established based on fluid dynamics formulas, and the adaptive learning model is a lightweight neural network model that is trained online using the third signal feature in the preceding time period as input features.

[0046] It should be noted that the adaptive learning model includes a fully connected feedforward network comprising an input layer, 1-2 hidden layers, and an output layer; its training objective is set to predict instantaneous mass flow rate values, including: when the operating condition is determined to be stable, i.e., the fluctuation variance of the pseudo-single-phase flow equivalent pressure difference is lower than a threshold, and the confidence of the benchmark calculation model is high, the current time and the preceding N-second time window are used to predict the mass flow rate. The sequence, the pipeline static pressure sequence, and the final flow rate calculated by weighted fusion together constitute a training sample, which is stored in a training sample queue of fixed capacity. Every 24 hours, all samples are used to perform an incremental training on the neural network, with the learning rate set to 0.001. Its output layer is a node, representing the predicted flow rate at the current moment.

[0047] In this embodiment, the high confidence level determination condition is that the following conditions are met simultaneously: the fluctuation variance of the third signal feature in the past 60 seconds is lower than the first set threshold; the absolute value of the difference between the flow rate values ​​output by the benchmark calculation model and the adaptive learning model is lower than the second set threshold. In this embodiment, the first set threshold and the second set threshold can be initially set by observing the distribution range of the fluctuation variance and the model output difference of the unit in the recognized stable load phase and taking its 95th percentile as the initial set value.

[0048] In one possible implementation, the third signal characteristic and the pipeline static pressure signal are input into the benchmark calculation model to calculate the first flow rate value. Specifically, it is expressed as: Where C represents the design constant. Expressed as the flow beam expansion coefficient, Expressed as medium density, This is expressed as the equivalent pressure difference of a pseudo-single-phase flow; simultaneously, the third signal feature sequence, the pipeline static pressure sequence, and the historical flow trend from the preceding time period are input into the adaptive learning model to calculate the second flow value. The first flow rate value and the second flow rate value are weighted according to a real-time weighting coefficient to calculate the real-time steam mass flow rate. Specifically, it is expressed as: in, , These are respectively represented as the first flow rate value. With the second flow value Real-time weighting coefficients.

[0049] It should be noted that the design constant C and the flow expansion coefficient are... The density of the medium is obtained by calibrating the geometric dimensions and hydrodynamic characteristics of the throttling device. This is obtained by checking the current pipeline static pressure Ps and the saturated steam properties at the steam temperature.

[0050] Furthermore, the real-time weighting coefficients are dynamically adjusted based on the fluctuation variance of the third signal feature within a preset time window, including: determining the first weight of the benchmark calculation model based on the fluctuation variance using a preset monotonically decreasing function. The second weight of the adaptive learning model Then it is 1 and the first weight The difference is ( When the variance of the fluctuation increases, the first weight... Decrease, the second weight The corresponding increase.

[0051] In this embodiment, the preset time window is defined as 10 seconds, and the first weight... The calculation formula is: ,in This is expressed as the variance of the equivalent pressure difference within the most recent 10-second time window, with k=0.1 representing an empirical constant; the variance of the equivalent pressure difference fluctuation during rapid changes in heating load. Enlargement, leading to This reduction makes the output more dependent on the neural network model that has learned dynamic characteristics. This reduces the need for traditional pure physics models. Improved dynamic accuracy is achieved by addressing response lag and modeling errors during transient processes.

[0052] In this embodiment, a comparative test was conducted on the steam inlet pipe of the low-pressure cylinder of a cogeneration unit. Compared with the traditional method of throttling with a single scaling tube and fixed formula calculation, the real-time monitoring device of the present invention exhibits comprehensive and significant technical advantages. After 500 hours of continuous operation, the differential pressure signal drifted by more than ±15% due to condensation blockage of the pressure tap, and finally required shutdown and cleaning after 750 hours due to complete signal failure. However, the self-purge venturi tube and multi-throat array used in the present invention maintained the main channel signal drift within ±2.5% throughout the 3000-hour test cycle, and achieved 100% data availability through the non-disruptive switching of the array, fundamentally ensuring the continuity and reliability of monitoring.

[0053] The feedback maintenance module is used to continuously monitor and analyze the operating status metadata of each module to trigger the switching of the sensing unit in the differential pressure signal generation module or to trigger the model relearning in the dynamic flow calculation module. The operating status metadata includes at least the signal attenuation rate of the multi-throat array sensing unit, the reading confidence of the second signal feature, and the output deviation value between the benchmark calculation model and the adaptive learning model.

[0054] It should be noted that the feedback maintenance module has preset performance thresholds corresponding to signal attenuation rate, calibration thresholds corresponding to reading confidence, and model drift thresholds corresponding to output deviation values; by comparing the real-time acquired metadata with the corresponding thresholds, the module generates corresponding switching instructions or relearning trigger signals based on the comparison results.

[0055] In one possible implementation, based on the operating status metadata, the performance degradation trend of the current working unit in the multi-throat array sensing unit is predicted, and before the predicted performance reaches a preset first maintenance threshold, a first maintenance decision signal is generated to trigger the differential pressure signal generation module to perform a disturbance-free switch. The generation of the first maintenance decision signal is based on the output of the performance degradation prediction. The execution of the disturbance-free switch includes: sending a switch command containing a backup signal source identifier and its real-time calibration coefficient to the differential pressure signal generation module; the differential pressure signal generation module completes the signal source switch between two adjacent data sampling periods, and the dynamic flow calculation module synchronously uses the new calibration coefficient to perform flow calculation.

[0056] It should be noted that the switching instruction includes a switching timing synchronization signal. After receiving the instruction, the differential pressure signal generation module first locks the current sampling period data, and then completes the physical channel switching and signal stabilization to the backup sensing unit before the start of the next sampling period. The dynamic flow calculation module then uses the new signal source and its calibration coefficients to perform calculations from the start of the new sampling period, thereby ensuring the continuity of the output flow.

[0057] Furthermore, in the feedback maintenance module, the following steps need to be performed: performance degradation prediction, establishing a time series model based on the signal attenuation rate and the cumulative runtime and historical operating conditions of the corresponding sensing unit to predict future changes in its effective aperture area; calibration requirement judgment, determining the necessity of corresponding sensor calibration based on whether the confidence level of the second signal feature reading is continuously lower than the confidence threshold; and model relearning judgment, determining the necessity of model relearning based on whether the output deviation between the benchmark calculation model and the adaptive learning model exceeds the dynamic deviation threshold.

[0058] In this embodiment, the time series model can employ a prediction algorithm based on linear regression or exponential smoothing, using the sequence of signal attenuation rate changes over time to extrapolate the time point at which it reaches the performance threshold; the reading confidence level can be calculated based on the signal-to-noise ratio and spectral width of the microwave resonant sensor; and the dynamic deviation threshold can be adaptively adjusted according to the current flow rate operating range.

[0059] It should be noted that the model relearning is a background task triggering mechanism. When it is determined that relearning is needed, the feedback maintenance module sends an instruction to the dynamic traffic calculation module. While maintaining the normal operation of the foreground, the adaptive learning model retraining process will be started in the background using the high confidence dataset in the historical database. After training is completed, the validated new model parameters will smoothly replace the original parameters, realizing hot updating of the model.

[0060] In this embodiment, since the pressure taps become significantly clogged every 1-3 months on average, leading to signal distortion or even complete failure, the annual availability of the measurement system is usually less than 80%, and maintenance requires downtime. In contrast, the present invention uses a Venturi tube with a self-purge structure and a redundant design of a multi-throat array, combined with a non-disruptive switching mechanism based on performance degradation prediction, which can minimize the risk of signal interruption caused by pressure tap problems.

[0061] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other. In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A real-time monitoring device for the steam inlet flow rate of the low-pressure cylinder of a power plant heating turbine, comprising a processor and a memory, wherein the memory stores a computer program, characterized in that, When the computer program is executed by the processor, it further includes: Differential pressure signal generation module: used to acquire measurement signals through a sensor array installed in the steam inlet pipe of the low-pressure cylinder, the measurement signals including at least: A first signal characteristic characterizing fluid dynamic pressure is generated by a preset throttling device, the throttling device including a main throttling element and a multi-throat array sensing unit, comprising: continuously monitoring the differential pressure signal generated by the main throttling element and continuously forming a self-draining purge; monitoring the real-time quality index of the differential pressure signal in parallel and evaluating the reachability of each of the multi-throat array sensing units, the real-time quality index including at least the signal-to-noise ratio and attenuation rate; when the real-time quality index is lower than a preset threshold, automatically switching to at least one of the multi-throat array sensing units as a backup signal source to continuously obtain an effective differential pressure value; The second signal characteristic reflecting the steam state includes at least the pipeline static pressure signal, dielectric constant signal, and temperature distribution signal; The equivalent pressure difference calculation module is used to perform fusion calculations based on the first and second signal characteristics using a preset wet steam two-phase flow correction model, and outputs a third signal characteristic after real-time compensation for steam dryness. Specifically, it is a pseudo-single-phase flow equivalent pressure difference, which is expressed as the actual differential pressure generated by the current wet steam fluid flowing through the throttling device, equivalently converted to the theoretical differential pressure value generated by single-phase superheated steam of the same mass flow rate flowing through the same device under the same conditions. Specifically, it is expressed as follows: in, This is expressed as the actual differential pressure of wet steam obtained directly through the throttling device. Indicated as the static pressure of the pipeline Related stress correction factors, This is expressed as a real-time estimate of steam dryness. , Represented as model exponential parameters; The dynamic flow calculation module is used to perform weighted calculations based on the third signal feature using a parallel benchmark calculation model and an adaptive learning model to output the real-time steam mass flow rate. This includes: inputting the third signal feature and pipeline static pressure signal into the benchmark calculation model to calculate a first flow rate value; simultaneously, inputting the third signal feature sequence, pipeline static pressure sequence, and historical flow trend from the preceding time period into the adaptive learning model to calculate a second flow rate value; wherein the adaptive learning model is a lightweight neural network model; and weighting the first flow rate value and the second flow rate value according to real-time weighting coefficients to output the real-time steam mass flow rate. The benchmark calculation model is established based on fluid dynamics formulas, and the adaptive learning model is trained online using the third signal features in the preceding time period as input features. The adaptive learning model is a lightweight neural network model that uses the third signal feature sequence, pipeline static pressure sequence, and historical flow trend from the preceding time period as input features for online training, and outputs a second flow value. ; Feedback and maintenance module: This module continuously monitors and analyzes the metadata of the operating status of each module to trigger the switching of sensor units in the differential pressure signal generation module or to trigger model relearning in the dynamic flow calculation module. It also performs the following: performance degradation prediction, based on the signal attenuation rate and the cumulative operating time and historical conditions of the corresponding sensor unit, to predict changes in its effective flow area; calibration requirement judgment, based on the confidence level of the second signal characteristic reading, to determine the necessity of corresponding sensor calibration; and model relearning judgment, based on the output deviation between the benchmark calculation model and the adaptive learning model, to determine the necessity of model relearning. The operational status metadata includes at least the signal attenuation rate of the multi-throat array sensing unit, the confidence level of the second signal feature reading, and the output deviation between the benchmark calculation model and the adaptive learning model.

2. The real-time monitoring device for the steam inlet flow rate of the low-pressure cylinder of a power plant heating turbine according to claim 1, characterized in that: The differential pressure signal generation module generates a throttling device for the first signal characteristic, which is a composite venturi tube structure, including: A main Venturi tube, which serves as the main throttling element, has an annular steam groove on the inner wall of its throat. It is connected to the high-pressure tapping chamber upstream of the throat through a built-in self-draining channel, and is used to perform the self-draining purging. Additionally, at least two miniature Venturi tubes with different throat diameters, connected in parallel downstream of the main Venturi tube and serving as the multi-throat array sensing unit, are used to form the backup signal source.

3. The real-time monitoring device for the steam inlet flow rate of the low-pressure cylinder of a power plant heating turbine according to claim 1, characterized in that: The equivalent pressure difference calculation module receives the dielectric constant signal from the second signal feature and defines its signal and dielectric constant. Relationship: ， in, , These represent the reference resonant frequency and the corresponding equivalent dielectric constant of the sensor in the cavity state, respectively. , These represent the measured resonant frequency shift and the change in quality factor, respectively. , These are respectively represented as calibration coefficients determined by the sensor's geometry; The dielectric constant Input the pre-stored steam dryness mapping model and output the real-time steam dryness estimate. Specifically, it is expressed as: ， in, These represent the current static pressure in the pipeline. The dielectric constants of pure saturated water vapor and pure saturated water at [temperature and temperature]. It is represented as the morphological distribution factor.

4. The real-time monitoring device for the steam inlet flow rate of the low-pressure cylinder of a power plant heating turbine according to claim 1, characterized in that: The dynamic traffic calculation module determines the real-time weight coefficient, specifically including: Based on the fluctuation variance of the third signal feature within a preset time window, the first weight of the benchmark calculation model is determined by a preset monotonically decreasing function, and the second weight of the adaptive learning model is the difference between 1 and the first weight. When the variance of the fluctuation increases, the first weight decreases and the second weight increases accordingly.

Images