Online unblocking control method and system for pressure sampler

Abstract

The invention belongs to the technical field of power systems, and particularly relates to a pressure sampler on-line unblocking control method and system, and the method comprises the steps: collecting the pressure difference fluctuation data of the inner wall of a pipeline in real time based on a quantum pressure difference sensor, associating historical flow field parameters through a pre-trained quantum neural network model, and dynamically predicting the multi-moment blocking risk level; a dynamic cleaning path is generated in combination with a quantum annealing algorithm, a genetic algorithm parameter library is called to match scene gene data, and cleaning priorities of a fourth pipe body and a seventh pipe body are determined; the control module generates an air cylinder driving instruction according to the cleaning path and the priority, and drives the spiral brush head to execute high-frequency inching unblocking operation, dynamic sensing and accurate cleaning of the blocking risk are achieved, and the data reliability and operation and maintenance safety of thermal power generation pressure measuring points are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of power systems, and particularly relates to an on-line clogging removal control method and system for a pressure sampler. Background Art

[0002] During the operation of a pulverized coal fired boiler in thermal power generation, the pressure measurement data at key parts such as the furnace, air preheater, and electrostatic precipitator are the core basis for boiler combustion efficiency and safety regulation, and its accuracy directly affects the unit load response and pollutant emission control. Traditional anti-clogging devices rely on backwashing or mechanical ash cleaning structures to dredge pipelines, but the backwashing airflow will interfere with the internal pressure field of the pressure taking pipeline, resulting in instantaneous signal distortion and inability to synchronize real working condition data; frequent disassembly and maintenance are required to eliminate blockages, and operations in high-temperature and positive-pressure environments pose risks of scalding and poisoning to personnel, and shutdown maintenance further exacerbates the measurement data fault and system control lag. Especially in the scenario of a fluidized bed boiler with high dust, the flowing material continuously impacts the sampling pipeline. Due to the lack of working condition perception ability, the conventional fixed-cycle clogging removal strategy is difficult to inhibit the dust deposition rate, and the existing pipeline layout design has inherent defects in suppressing airflow disturbance and dust movement, resulting in an increasing blockage probability with the operation time. The above problems form a vicious cycle of "distortion, maintenance, and secondary distortion". Due to the lack of real-time perception, dynamic decision-making, and maintenance-free closed-loop control mechanisms, the existing technologies are difficult to meet the reliability requirements of pressure measurement in complex industrial environments. Summary of the Invention

[0003] Aiming at the deficiencies of the existing technologies, the present invention provides an on-line clogging removal control method and system for a pressure sampler, which solves the problems that the traditional anti-clogging device causes pressure signal distortion due to backwashing interference, high risk of manual maintenance in high-temperature and high-risk environments, insufficient dynamic working condition adaptability of the fixed clogging removal strategy, and pipeline structure design defects, resulting in the decline of the reliability of key pressure measurement points data in thermal power generation and the increase of operation and maintenance risks.

[0004] To solve the above technical problems, the specific technical solutions of the present invention are as follows: In the first aspect, the present invention provides an on-line clogging removal control method for a pressure sampler, including: Step S101, collecting real-time differential pressure data on the inner wall of the sampler pipeline based on a quantum differential pressure sensor, and generating a differential pressure fluctuation sequence including time stamps; Step S102, inputting the differential pressure fluctuation sequence into a pre-trained quantum neural network model, and combining the dust deposition characteristic data in the historical flow field parameter database to output the pipeline clogging risk levels at multiple future moments; Step S103, according to the clogging risk levels, generating a dynamic cleaning path within the preset cleaning path search space through a quantum annealing algorithm, and calling the scenario gene data matching the current working condition in the genetic algorithm parameter library to determine the cleaning action priorities of the fourth pipe body and the seventh pipe body; Step S104: Based on the dynamic cleaning path and the cleaning action priority, send a control instruction to the cylinder drive module to drive the spiral brush head to perform a blockage clearing operation at a set jogging frequency and rotation speed. Step S105: Within a preset time window of the cylinder push rod action, synchronously freeze the transmission of pressure sampling data through the quantum entanglement channel between the quantum sensor and the controller to isolate the interference of the backflush air flow on the pressure data. Step S106: Collect the real-time differential pressure data after the blockage clearing operation through the quantum differential pressure sensor, and calculate the deviation value from the baseline differential pressure data before cleaning. If the deviation value exceeds the preset threshold, send a secondary local cleaning instruction to the control module to trigger the spiral brush head to perform supplementary cleaning on the local area of the target pipe body. Step S107: Based on the prediction result of the wear rate of the spiral brush head by the quantum neural network model, combine the cumulative number of blockage clearings to construct a remaining life prediction curve, generate a maintenance cycle warning signal, and push it to the power plant management system through the wireless communication module.

[0005] Further, for the online blockage clearing control method of the pressure sampler of the present invention, the prediction of the pipeline blockage risk level by the quantum neural network includes: Input the real-time differential pressure change and historical flow field parameters into a pre-trained quantum neural network model. The pre-trained quantum neural network model is a quantum convolutional neural network model trained based on historical differential pressure data. Extract the differential pressure fluctuation characteristics through the convolutional layer of the pre-trained quantum neural network model, and correlate the dust deposition characteristics through the attention mechanism. Based on the output of the fully connected layer of the quantum neural network model, obtain the blockage probability distribution maps at multiple future moments, and the probability distribution maps represent the dust accumulation probabilities of different pipe segments. According to the preset threshold interval where the probability peak is located in the blockage probability distribution map, map it to high-risk, medium-risk, or low-risk level labels.

[0006] The dynamic cleaning path is generated by optimizing the risk level output by the quantum neural network through the quantum annealing algorithm.

[0007] Further, for the online blockage clearing control method of the pressure sampler of the present invention, the generation of the dynamic cleaning path using the quantum annealing algorithm includes: Divide the cleaning path search space of the fourth pipe body and the seventh pipe body according to the risk level labels, where the high-risk level corresponds to the priority cleaning range of the seventh pipe body. Construct a quantum annealing energy function within the search space, and the energy function takes the linear combination of the cleaning energy consumption weight and the dust removal efficiency weight as the optimization target. Traverse the solution space of the energy function through the tunneling effect of qubits, and output the alternating cleaning sequence of the fourth pipe body and the seventh pipe body and the corresponding cylinder action time window.

[0008] Furthermore, for the online clogging removal control method of the pressure sampler of the present invention, the scenario gene data in the matching genetic algorithm parameter library includes: Analyze the dust concentration and air flow velocity parameters in the current working condition environment characteristics, and call the scenario gene data matching the parameters from the genetic algorithm parameter library. The scenario gene data includes the cylinder thrust coefficient and the spiral brush head rotation speed threshold under the high-sulfur coal working condition; Spatiotemporally couple the cylinder thrust parameter in the scenario gene data with the cleaning sequence output by the quantum annealing algorithm to generate a clogging removal control instruction including the cylinder jogging frequency and the brush head rotation speed.

[0009] Furthermore, for the online clogging removal control method of the pressure sampler of the present invention, the control of the cylinder to drive the spiral brush head to perform high-frequency jogging clogging removal operations includes: According to the dust accumulation thickness data on the inner wall of the diversion groove feedback by the quantum sensor, dynamically adjust the depth parameter of the spiral diversion groove through the proportional-integral algorithm, so that the groove depth is negatively correlated with the accumulation thickness; Based on the adjusted diversion groove depth, match the corresponding upper limit value of the cylinder jogging frequency and the reference value of the brush head rotation speed from the genetic algorithm parameter library; When the vibration signal of the fluidized bed boiler in the start-stop stage is detected, switch to the double-cylinder alternating action mode, and superimpose a high-frequency pulse signal on the clogging removal control instruction to cope with the instantaneous high-dust impact.

[0010] Furthermore, for the online clogging removal control method of the pressure sampler of the present invention, the verification of the effectiveness of the clogging removal operation includes: Collect the real-time pressure difference data after the clogging removal operation through the quantum pressure difference sensor, and call the baseline pressure difference data before cleaning from the historical database; Input the real-time pressure difference data and the baseline pressure difference data into the differential calculation module to generate a pressure difference deviation value and mark the corresponding pipe body position; If the pressure difference deviation value exceeds the preset secondary cleaning threshold, send a local cleaning instruction including the pipe body position coordinates to the control module to trigger the spiral brush head to perform supplementary cleaning on the target area.

[0011] Furthermore, for the online clogging removal control method of the pressure sampler of the present invention, the generation of the maintenance cycle warning signal includes: Input the cumulative clogging removal times of the spiral brush head into the pre-trained quantum neural network wear prediction model to output the remaining life prediction curve; The remaining life prediction curve is superimposed and compared with a preset safe operation threshold curve to determine a spare parts replacement trigger time node; The spare parts replacement trigger time node is encapsulated as a work order instruction through the LoRa wireless communication protocol and pushed to the maintenance task queue of the power plant management system.

[0012] Furthermore, the pressure sampler online blockage clearing control method of the present invention further includes: Before the cylinder performs the blockage clearing operation, a quantum entanglement channel between the sensor and the controller is established through the quantum key distribution protocol; When the start signal of the cylinder push rod action is triggered, the transmission link of the pressure sampling data is interrupted through the channel, and the pressure data in the current time window is cached; When the cylinder reset completion signal is received, the real-time transmission of the pressure sampling data is resumed and the data cache state is released.

[0013] Furthermore, the pressure sampler online clearing control method of the present invention further includes: under high temperature conditions, encapsulating the quantum sensor and the cylinder drive module in a high temperature resistant isolation cavity of a multilayer ceramic composite material, wherein a thermocouple is built into the cavity to monitor the cavity temperature in real time; The temperature data collected by the thermocouple is input into a temperature compensation algorithm to dynamically correct the thermal drift error component in the pressure difference data output by the quantum sensor.

[0014] In a second aspect, the present invention provides an online blockage clearing control system for a pressure sampler, which is applied to the online blockage clearing control method for a pressure sampler as described above, comprising: A quantum pressure difference sensor module is integrated in the inner wall of the fourth tube body and the seventh tube body of the sampler tube body, collects pipeline pressure difference fluctuation data in real time through quantum tunneling effect, and outputs a pressure difference sequence containing a timestamp; a data processing module is connected to the quantum pressure difference sensor module, receives the pressure difference sequence and inputs the pre-trained quantum neural network model, combines the dust deposition characteristics in the historical flow field parameter database, outputs the blockage risk level and dynamic cleaning path of the fourth tube body and the seventh tube body, and calls the scene gene data matching the current working condition in the genetic algorithm parameter library; The control module receives the dynamic cleaning path and scenario gene data, generates a clogging removal control instruction including the jogging frequency of the cylinder, the rotation speed of the spiral brush head and the cleaning sequence, and establishes a quantum entanglement channel with the quantum differential pressure sensor module through the quantum key distribution protocol to freeze the transmission of pressure sampling data within the cylinder action window; The execution module includes a double-cylinder drive device and a spiral brush head with adjustable depth, responds to the clogging removal control instruction, dynamically adjusts the rotation speed of the brush head according to the dust accumulation thickness on the inner wall of the diversion groove, and switches the alternating action mode during the start-up and shutdown stages of the fluidized bed boiler; The verification module is connected to the quantum differential pressure sensor module, calls the baseline differential pressure data before cleaning from the historical database, calculates the differential pressure deviation value after cleaning, and if the deviation value exceeds the threshold, sends a secondary cleaning instruction including the target pipe body coordinates to the control module; The maintenance warning module generates a spare part replacement work order based on the remaining life prediction curve of the spiral brush head output by the quantum neural network model and superimposes the cumulative number of clogging removals, and pushes it to the maintenance queue of the power plant management system through the LoRa wireless communication protocol; The high-temperature resistant protection module encapsulates the quantum differential pressure sensor module and the execution module in a multi-layer ceramic composite cavity, internally installs a thermocouple to collect the cavity temperature data, and corrects the thermal drift error in the differential pressure data through a temperature compensation algorithm. The multi-layer ceramic composite cavity includes an alternately stacked alumina-based bottom layer and a silicon nitride heat insulation layer.

[0015] Advantages of the present invention; The advantages of the present invention are embodied in: real-time collection of pipeline differential pressure fluctuation data through a quantum differential pressure sensor, combined with the dynamic prediction of the clogging risk level at multiple moments by the quantum neural network model, effectively eliminating the indexing deviation caused by traditional static clustering rules; using the quantum annealing algorithm to generate an optimized sequence in the dynamic cleaning path search space, combined with the scenario gene data call mechanism of the genetic algorithm parameter library, to achieve adaptive fusion of cross-modal features and collaborative analysis of multi-dimensional parameters; synchronously freezing the pressure sampling data based on the quantum entanglement channel to construct an anti-blowing interference isolation mechanism to ensure the real-time and integrity of data collection; forming a closed-loop control logic through the superposition of the differential pressure deviation feedback of the verification module and the remaining life prediction curve of the maintenance warning module to enhance the stability of the hierarchical semantic association network; the high-temperature resistant protection module and the temperature compensation algorithm work together to suppress the influence of environmental thermal interference on the quantum sensing accuracy, combined with the disassembly-free maintenance and remote self-check mechanism, significantly improving the reliability and operation and maintenance efficiency of the pressure measurement point system for thermal power generation. Description of the drawings

[0016] In order to more clearly illustrate the technical solutions of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, for those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on the drawings.

[0017] Figure 1Flowchart of an online blockage clearing control method for a pressure sampler provided by an embodiment of the present invention. Detailed implementation manners

[0018] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only a part rather than all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention. The technical solutions provided by various embodiments of the present invention are described in detail below with reference to the drawings. To better understand the objectives of the present invention, the present invention is further described in detail below.

[0019] Please refer to Figure 1 , on the first hand, the present invention provides an online blockage clearing control method for a pressure sampler, and the specific implementation process includes: Step S101: Based on a quantum differential pressure sensor, collect real-time differential pressure data on the inner wall of the sampler pipeline, and generate a differential pressure fluctuation sequence containing timestamps; Step S102: Input the differential pressure fluctuation sequence into a pre-trained quantum neural network model, and combine the dust deposition characteristic data in the historical flow field parameter database to output the pipeline blockage risk levels at multiple future moments; Step S103: According to the blockage risk levels, generate a dynamic cleaning path within a preset cleaning path search space through a quantum annealing algorithm, and call the scenario gene data matching the current working condition in the genetic algorithm parameter library to determine the cleaning action priorities of the fourth pipe body and the seventh pipe body; Step S104: Based on the dynamic cleaning path and the cleaning action priorities, send control instructions to the cylinder drive module to drive the spiral brush head to perform blockage clearing operations at a set jogging frequency and rotation speed; Step S105: Within a preset time window of the cylinder push rod action, synchronously freeze the transmission of pressure sampling data through the quantum entanglement channel between the quantum sensor and the controller to isolate the interference of the backflush air flow on the pressure data; Step S106: Collect real-time differential pressure data after the blockage clearing operation through a quantum differential pressure sensor, and calculate the deviation value from the baseline differential pressure data before cleaning; If the deviation value exceeds a preset threshold, send a secondary local cleaning instruction to the control module to trigger the spiral brush head to perform supplementary cleaning on a local area of the target pipe body; Step S107: Based on the prediction result of the wear rate of the spiral brush head by the quantum neural network model, combine the cumulative blockage clearing times to construct a remaining life prediction curve, generate a maintenance period warning signal and push it to the power plant management system through a wireless communication module.

[0020] In step S101, the quantum differential pressure sensor monitors the differential pressure change on the inner wall of the sampler pipeline in real time through the quantum tunneling effect. The sensor is installed at specific detection points on the fourth pipe body and the seventh pipe body, and captures the differential pressure fluctuation signal at a sampling frequency of milliseconds. The data acquisition module converts the differential pressure signal into a digital sequence and attaches a timestamp to form a fluctuation data set with time series characteristics. The packaging structure of the sensor uses a multi-layer ceramic composite material, and a temperature compensation unit is integrated inside to suppress the interference of ambient temperature changes on the quantum tunneling current, thereby improving the acquisition accuracy of differential pressure data.

[0021] In step S102, the pre-trained quantum neural network model extracts features from the input differential pressure fluctuation sequence through the convolutional layer, and identifies periodic fluctuations and abnormal spike signals. The model also correlates the dust deposition characteristic data in the historical flow field parameter database, including the dust particle size distribution and the historical records of air flow velocity in different pipe sections, and calculates the correlation weight between the differential pressure fluctuation and the dust deposition through the attention mechanism. The fully connected layer generates the clogging probability distribution maps at multiple future moments based on the above features, where the horizontal axis of the probability distribution map represents the pipe body position, and the vertical axis is the dust accumulation probability. According to the preset threshold interval division rule, the interval where the probability peak is located is mapped to high, medium, and low risk labels, providing a basis for subsequent path planning.

[0022] In step S103, the quantum annealing algorithm divides the search space of the cleaning path based on the risk level label, and the high risk level corresponds to the priority cleaning area of the seventh pipe body. The energy function constructed by the algorithm takes the weighted sum of the cleaning energy consumption and the dust removal efficiency as the optimization goal, and traverses the solution space through the tunneling effect of quantum bits to generate an alternating cleaning sequence for the fourth pipe body and the seventh pipe body. The genetic algorithm parameter library stores the scenario gene data under different working conditions, such as the cylinder thrust coefficient and the spiral brush head rotation speed threshold corresponding to the high-sulfur coal working condition. The control module analyzes the dust concentration and air flow velocity parameters of the current working condition, calls the matching scenario gene data, and couples the cleaning sequence and the working condition parameters in space and time to generate a clogging removal control instruction including the jogging frequency and the rotation speed.

[0023] In step S104, the control instruction drives the double-cylinder device to drive the spiral brush head to move along the preset path. The depth of the diversion groove of the spiral brush head is dynamically adjusted according to the dust accumulation thickness feedback by the quantum sensor, and the proportional-integral algorithm controls the reverse correlation relationship between the depth of the diversion groove and the accumulation thickness through the negative feedback mechanism. When the vibration signal in the start-up and shutdown stages of the fluidized bed boiler is detected, the control module switches the alternating action mode of the double cylinders and superimposes a high-frequency pulse signal on the instruction to enhance the cleaning intensity under the instantaneous high-dust impact.

[0024] In step S105, before the start of the plugging removal operation, the quantum key distribution protocol establishes a quantum entanglement channel between the sensor and the controller. When the cylinder push rod moves, this channel synchronously interrupts the transmission link of the pressure sampling data and caches the data within the current time window in the non-volatile memory. After the cylinder reset signal arrives, the data link resumes real-time transmission, and at the same time, the cached state is released to avoid pressure sampling distortion caused by the backflush air flow.

[0025] In step S106, the verification module calls the baseline differential pressure data in the historical database and performs differential calculation with the cleaned real-time differential pressure data to generate the differential pressure deviation values at the positions of each pipe body. If the deviation value of a certain pipe body exceeds the preset threshold, the control module triggers a local cleaning instruction according to the coordinate information, and the spiral brush head performs a supplementary cleaning operation on the target area until the deviation value returns to the allowable range.

[0026] In step S107, the quantum neural network wear prediction model outputs a remaining life prediction curve based on the cumulative plugging removal times of the spiral brush head and the working condition parameters. The maintenance warning module performs superposition analysis on this curve and the preset safe operation threshold curve to determine the time node for spare part replacement. The work order instruction is pushed to the maintenance queue of the power plant management system through the low-power wide-area network protocol to realize remote monitoring of the equipment status and automatic scheduling of the maintenance plan.

[0027] The above steps form a complete control link from data acquisition, risk prediction, path optimization to execution verification through the coordination of quantum sensing, dynamic path planning and closed-loop feedback mechanism. The logical relationship between each step is as follows: real-time data drives model prediction, the prediction result guides path generation, the path and working condition parameters are combined to generate execution instructions, interference is synchronously isolated during the execution process, the cleaning effect triggers supplementary operations through deviation calculation, and finally maintenance warning is realized based on the equipment wear status. This technical solution solves the problems of insufficient dynamic adaptability and low data reliability in traditional plugging removal methods through hierarchical logical association and multi-algorithm fusion.

[0028] Specifically, for the online plugging removal control method of the pressure sampler described in the present invention, the risk level of pipeline blockage is predicted by a quantum neural network, including: Input the real-time differential pressure change and historical flow field parameters into the pre-trained quantum neural network model, extract the differential pressure fluctuation characteristics through the convolutional layer of the pre-trained quantum neural network model, and associate the dust deposition characteristics through the attention mechanism; Based on the fully connected layer of the quantum neural network model, output the plugging probability distribution maps at multiple future moments, and the probability distribution maps represent the dust accumulation probabilities of different pipe segments; According to the preset threshold interval where the probability peak is located in the plugging probability distribution map, map it to high-risk, medium-risk or low-risk level labels.

[0029] In the on-line blockage clearing control method of the pressure sampler described in the present invention, the steps of predicting the pipeline blockage risk level through a quantum neural network are as follows: The input layer of the quantum neural network model receives real-time differential pressure change data and historical flow field parameters. Among them, the differential pressure data is normalized to eliminate the dimension difference. The historical flow field parameters include the dust particle size distribution, air flow velocity and pipe section geometric parameters under different working conditions, which are stored in a structured database. The data preprocessing module aligns the real-time differential pressure data and historical parameters according to a time window to form a multi-dimensional feature vector, and inputs it into the convolutional layer of the model. The convolutional layer uses multi-scale convolutional kernels to extract local features from the differential pressure fluctuation sequence, identifies periodic fluctuation patterns and mutation features, and at the same time captures correlation information with different time spans through a sliding window mechanism.

[0030] The attention mechanism module performs cross-modal association on the extracted differential pressure features and dust deposition features, and calculates the dynamic weight coefficient between the dust deposition rate and the differential pressure fluctuation. After being normalized by the softmax function, the weight coefficient is weighted and fused with the differential pressure feature matrix to generate a comprehensive feature matrix with spatio-temporal correlation. The model processes multiple feature sub-spaces in parallel through a multi-head attention mechanism, captures the coupling relationship between dust deposition and differential pressure change in different dimensions, and improves the robustness of feature expression.

[0031] The fully connected layer maps the fused feature matrix into blockage probability distribution maps for multiple future time moments. The horizontal axis of the distribution map represents the segment number of the sampler pipe body, and the vertical axis is the dust accumulation probability value of each pipe segment at a preset time point. The output layer uses a sigmoid activation function to constrain the probability value to the 0-1 interval, generating a visual probability heat map. The probability distribution map divides the grid area according to the pipe body structure parameters, and each grid corresponds to the cumulative risk value of a specific pipe segment in the time series, forming a multi-dimensional risk prediction result.

[0032] The risk level mapping module classifies the probability peaks according to a preset threshold interval. The threshold interval is set based on the statistical results of historical data. The high risk level corresponds to the interval where the probability value is greater than 0.7, the medium risk level corresponds to the interval of 0.4-0.7, and the low risk level corresponds to the interval less than 0.4. During the mapping process, the system preferentially selects the pipe segment position where the peak value is located in the probability distribution map as the risk marking object, and incorporates the probability gradient change of adjacent pipe segments into the evaluation range to avoid misjudgment of local extreme values. The classification result is output in the form of three-dimensional coordinates, including the pipe segment number, risk level and prediction time node, providing dynamic input parameters for subsequent path planning.

[0033] The above steps combine real-time sensing data with historical parameters through multi-level feature fusion and spatio-temporal correlation analysis of a quantum neural network to establish a dynamic risk prediction model. The probability distribution map output by the model reflects the dust accumulation difference between pipe segments, and combined with the threshold mapping rule, a hierarchical risk assessment result is formed, providing data support for the dynamic adjustment of the cleaning path. The logical relationship between the steps is as follows: data input drives feature extraction, cross-modal correlation strengthens feature expression, probability distribution generates quantitative risk, and threshold mapping realizes decision output, forming a closed-loop analysis link from data to decision.

[0034] Specifically, for the online plugging removal control method of the pressure sampler described in the present invention, the use of the quantum annealing algorithm to generate a dynamic cleaning path includes: Dividing the cleaning path search space of the fourth pipe body and the seventh pipe body according to the risk level label, where the high-risk level corresponds to the priority cleaning range of the seventh pipe body; Constructing a quantum annealing energy function within the search space, where the energy function takes the linear combination of the cleaning energy consumption weight and the dust removal efficiency weight as the optimization target; Traversing the solution space of the energy function through the tunneling effect of quantum bits, and outputting the alternating cleaning sequence of the fourth pipe body and the seventh pipe body and the corresponding cylinder action time window.

[0035] In the online plugging removal control method of the pressure sampler described in the present invention, the steps of using the quantum annealing algorithm to generate a dynamic cleaning path are as follows: The division module based on the risk level label discretizes the three-dimensional structure model of the fourth pipe body and the seventh pipe body into a grid search space, where the high-risk level corresponds to a specific sub-region of the seventh pipe body. The boundary of the sub-region is dynamically adjusted according to historical blockage records and real-time dust concentration data, and preferentially covers the pipe segments with a dust deposition rate higher than the set threshold. The path nodes in the search space correspond to the diversion groove structure inside the pipe body, and each node is associated with the initial parameter values of the cleaning energy consumption and the dust removal efficiency, forming the basic data for topological path planning.

[0036] The quantum annealing energy function construction module maps the cleaning path optimization problem into a combinatorial optimization model. The cleaning energy consumption weight coefficient in the energy function is dynamically calculated according to the cylinder thrust parameter and the movement distance of the spiral brush head, and the dust removal efficiency weight coefficient is determined based on the dust accumulation density and particle hardness values detected in real time. The weight coefficients are dynamically adjusted through the regression analysis of historical working condition data, so that the energy function automatically adapts to the balance relationship between cleaning efficiency and energy consumption under different working conditions. The function optimization target is transformed into minimizing the total energy value of the cleaning path, where the total energy value is the linear weighted sum of the energy consumption and efficiency of each path node, and the weight ratio is updated in real time according to the wear degree of the pipe body material.

[0037] The qubit encoding module maps the cleaning path sequence into the superposition state of qubits, where each qubit represents the cleaning state of a specific pipe segment. The tunneling effect is applied through the transverse magnetic field of the quantum annealer, enabling the system to traverse possible combinations of cleaning sequences in the solution space. During the iterative process, the algorithm preferentially retains the cleaning path solutions with lower energy function values and synchronously updates the alternating cleaning order of the fourth and seventh pipe bodies through the quantum entanglement mechanism. The output module decodes the optimal solution into a cleaning sequence including the time dimension, where the length of the cylinder action time window is calculated based on the pipe segment length and the brush head movement speed, and the window interval time is synchronously calibrated with the boiler operating state signal.

[0038] The above steps transform the dynamic path planning problem into an energy minimization solution process through the quantum annealing algorithm, and achieve the adaptive partitioning of the search space and the dynamic adjustment of weights in combination with real-time working condition parameters. The alternating cleaning sequence output by the algorithm takes into account both the energy consumption efficiency and the dust removal requirements. The time window parameters match the device action characteristics, forming an executable plugging removal control instruction. The logical relationship between each step is as follows: the risk level drives the space partitioning, the energy function defines the optimization goal, the qubit encoding realizes the traversal of the solution space, and the final output sequence and time parameters jointly guide the action of the actuator, constituting a complete dynamic path generation mechanism.

[0039] Specifically, in the online plugging removal control method of the pressure sampler described in the present invention, the scenario gene data in the matching genetic algorithm parameter library includes: Analyze the dust concentration and air flow velocity parameters in the current working condition environment characteristics, and call the scenario gene data matching the parameters from the genetic algorithm parameter library. The scenario gene data includes the cylinder thrust coefficient and the spiral brush head rotation speed threshold under the high-sulfur coal working condition; Couple the cylinder thrust parameter in the scenario gene data with the cleaning sequence output by the quantum annealing algorithm in space and time to generate a plugging removal control instruction including the cylinder jogging frequency and the brush head rotation speed.

[0040] In the online plugging removal control method of the pressure sampler described in the present invention, the steps of matching the scenario gene data in the genetic algorithm parameter library are as follows: The working condition environment characteristics analysis module collects the dust concentration and air flow velocity parameters in real time through multi-source sensors. The data preprocessing unit performs noise reduction and normalization processing on the parameters to generate a standardized feature vector. The genetic algorithm parameter library adopts a hierarchical storage structure, where the scenario gene data is stored classified by sulfur content, dust viscosity, and pipe body material. Each type of gene data includes the cylinder thrust coefficient, the spiral brush head rotation speed threshold, and the allowable vibration amplitude range. The parameter matching engine calculates the similarity between the current working condition and historical scenarios based on the Euclidean distance of the feature vector, and preferentially calls the scenario gene data group with a similarity higher than the preset threshold, such as the cylinder thrust coefficient threshold and the maximum rotation speed limit parameter of the spiral brush head corresponding to the high-sulfur coal working condition.

[0041] The spatio-temporal coupling module performs multi-dimensional alignment of the scene gene data called and the cleaning sequence output by the quantum annealing algorithm. The pipe segment numbers and time nodes in the cleaning sequence are mapped to spatial coordinates and timestamps, and the cylinder thrust parameters are dynamically corrected according to the pipe segment length and the number of elbows in the cleaning path. The corrected thrust parameters and the brush head rotation speed threshold are embedded in the time window of the cleaning sequence to generate a control instruction template including the segmented jogging frequency, acceleration curve, and rotation speed gradient interval. The instruction template fills the parameter gaps between discrete time nodes through the spatio-temporal interpolation algorithm to form a continuous executable clog-removing action sequence.

[0042] The control instruction generation module converts the action sequence after spatio-temporal coupling into a pulse signal recognizable by the cylinder drive module. The frequency of the pulse signal is proportional to the number of cylinder jogs, and the duty cycle is linearly related to the rotation speed of the spiral brush head. During the instruction transmission process, the system coordinates the alternating action timing of the double cylinders through the timestamp synchronization mechanism to avoid path deviation caused by execution delay. After the generated instruction set is verified for logical consistency through the check code, it is sent to the actuator to achieve the adaptive matching of the cleaning path and working condition parameters.

[0043] The above steps, through the scene matching and spatio-temporal coupling mechanism of the genetic algorithm parameter library, dynamically adapt the static gene data to the real-time cleaning requirements. The working condition feature analysis provides data input, the parameter library matching completes the scene adaptation, the spatio-temporal coupling realizes the fusion of parameters and paths, and finally generates executable instructions. The logical relationship between each step is as follows: the environmental features drive data call, the gene parameters dynamically correct the cleaning path, and the instruction generation realizes physical execution, forming a closed-loop control link from scene recognition to action execution.

[0044] Specifically, for the online clog-removing control method of the pressure sampler described in the present invention, the control cylinder drives the spiral brush head to perform high-frequency jogging clog-removing operations, including: According to the dust accumulation thickness data on the inner wall of the diversion groove feedback by the quantum sensor, the depth parameter of the spiral diversion groove is dynamically adjusted through the proportional-integral algorithm, so that the groove depth is negatively correlated with the accumulation thickness; Based on the adjusted diversion groove depth, the corresponding upper limit value of the cylinder jogging frequency and the reference value of the brush head rotation speed are matched from the genetic algorithm parameter library; When the vibration signal of the fluidized bed boiler in the start-stop stage is detected, switch to the alternating action mode of the double cylinders, and superimpose high-frequency pulse signals on the clog-removing control instructions to cope with instantaneous high-dust impact.

[0045] In the online clog-removing control method of the pressure sampler described in the present invention, the steps for the control cylinder to drive the spiral brush head to perform high-frequency jogging clog-removing operations are as follows: Quantum sensors monitor the dust accumulation thickness on the inner wall of the diversion trough in real time through the piezoelectric effect. The thickness data is input into a proportional-integral controller after analog-to-digital conversion. The proportional-integral algorithm calculates the adjustment amount of the diversion trough depth according to the preset mapping relationship between the trough depth and the accumulation thickness, and drives a micro stepping motor to adjust the telescopic mechanism of the spiral diversion trough. The depth parameter and the accumulation thickness form a closed-loop negative feedback control. When an increase in thickness is detected, the control module outputs a signal to shorten the diversion trough depth, enhancing the peeling effect of the airflow on the deposited dust and reducing the probability of secondary deposition at the same time.

[0046] The genetic algorithm parameter library stores the operation parameter groups corresponding to different diversion trough depths using a multi-dimensional index structure. The parameter matching module performs a nearest neighbor search along the depth dimension in the parameter library according to the adjusted depth value, and extracts the corresponding upper limit value of the cylinder jogging frequency and the reference value of the brush head rotation speed. During the search process, the system compensates and corrects the reference value of the speed in combination with the current pipe material parameters to avoid excessive wear of the brush head caused by differences in material hardness. The matched parameter group is used as the basic template of the control instruction and is input into the dynamic adjustment unit.

[0047] The vibration signal detection module collects the vibration spectrum of the fluidized bed boiler through an acceleration sensor, and the feature extraction unit identifies the low-frequency resonance peak and high-frequency harmonic components in the start-stop stage. When the amplitude of the vibration signal exceeds the preset threshold, the control module switches the driving mode of the double cylinders, changing from synchronous propulsion to phase-difference alternating action. The high-frequency pulse signal generator superimposes a pulse sequence with an adjustable duty cycle on the clogging removal instruction, and the pulse frequency is synchronized with the main frequency component of the vibration signal to enhance the penetration of the brush head under instantaneous high-dust impact. In the alternating action mode, the timing logic controller of the double cylinders adopts a redundant check mechanism to prevent action conflicts caused by signal delays.

[0048] The above steps achieve the adaptive control of the clogging removal operation through the dynamic parameter adjustment and the working condition response mechanism. The sensor data drives the adjustment of the diversion trough structure, the parameter matching completes the adaptation of the action parameters, and the vibration detection triggers the mode switch, forming a closed-loop link from environmental perception to execution optimization. The logical relationship between the steps is as follows: the thickness feedback adjusts the diversion structure, the depth parameter indexes the operation threshold, the vibration signal triggers the mode switch, and finally, a stable clogging removal effect under dynamic working conditions is achieved through instruction superposition, improving the robustness of the system in complex industrial environments.

[0049] Specifically, for the online clogging removal control method of the pressure sampler described in the present invention, the verification of the effectiveness of the clogging removal operation includes: Collect the real-time differential pressure data after the clogging removal operation through a quantum differential pressure sensor, and call the baseline differential pressure data before cleaning from the historical database; Input the real-time differential pressure data and the baseline differential pressure data into a differential calculation module to generate a differential pressure deviation value and mark the corresponding pipe position; If the differential pressure deviation value exceeds a preset secondary cleaning threshold, a local cleaning instruction including the position coordinates of the pipe body is sent to the control module to trigger the spiral brush head to perform supplementary cleaning on the target area.

[0050] In the online blockage clearing control method of the pressure sampler described in the present invention, the steps for verifying the effectiveness of the blockage clearing operation are as follows: After the blockage clearing operation is completed, the quantum differential pressure sensor starts the high-precision sampling mode to collect the real-time differential pressure data of the target pipe body with a millisecond-level time resolution. The data acquisition module aligns the real-time data with the baseline differential pressure data in the historical database according to the time stamp, and the baseline data is selected from the stable differential pressure records after the last three effective blockage clearings under the same working conditions. During the data alignment process, the system eliminates the baseline drift caused by equipment aging or environmental temperature fluctuations through the sliding window mechanism, and retains the comparable characteristic data segments.

[0051] After the differential calculation module normalizes the real-time data and the baseline data, it performs point-by-point difference operations to generate a differential pressure deviation matrix. Each element in the matrix corresponds to a specific position in the grid coordinate system of the pipe body, and the element value represents the differential pressure change amount in this area. The module is built-in with a low-pass filter to eliminate high-frequency noise interference, and at the same time, a threshold truncation algorithm is used to filter non-significant deviation signals. The deviation matrix is expanded into a continuous distribution map through a spatial interpolation algorithm, and the pipe body coordinate area with a deviation value exceeding the preset threshold is marked to form a visual abnormal hot spot map.

[0052] The local cleaning instruction generation module extracts the geometric boundary parameters of the target area according to the coordinate information of the abnormal hot spot map. The system converts the pipe body coordinate system into the motion coordinate system of the actuator, and combines the motion range and joint angle limit of the mechanical arm of the spiral brush head to generate an executable local cleaning path. The instruction transmission protocol uses redundant check coding to avoid coordinate offset errors during data transmission. When it is detected that the deviation of the same coordinate area exceeds the standard twice continuously, the control module automatically increases the cleaning priority of this area, and adds a brush head rotation speed gradient parameter to the instruction to strengthen the local cleaning intensity.

[0053] The above steps form a closed-loop verification system for the blockage clearing effect through data alignment, difference calculation and space mapping mechanisms. The comparison between the real-time data and the baseline data reveals the dust residue area, the generation of the deviation matrix quantifies the cleaning effect, and the coordinate mapping of the exceeded standard area triggers targeted supplementary cleaning. The logical relationship between each step is as follows: data acquisition provides the verification basis, difference analysis locates the abnormal area, coordinate conversion generates the execution path, and finally dynamic optimization is achieved through priority adjustment, forming a complete feedback link from effect evaluation to re-cleaning execution.

[0054] Specifically, for the online blockage clearing control method of the pressure sampler described in the present invention, the generation of the maintenance cycle warning signal includes: Input the cumulative blockage clearing times of the spiral brush head into a pre-trained quantum neural network wear prediction model to output a remaining life prediction curve; Overlay and compare the remaining life prediction curve with a preset safe operation threshold curve to determine the spare part replacement trigger time node; Package the spare part replacement trigger time node as a work order instruction through the LoRa wireless communication protocol and push it to the maintenance task queue of the power plant management system.

[0055] In the online blockage clearing control method of the pressure sampler described in the present invention, the steps of generating a maintenance cycle warning signal are as follows: The input layer of the quantum neural network wear prediction model receives the cumulative blockage clearing times of the spiral brush head and associated operating condition parameters, including the dust hardness for each blockage clearing, the friction coefficient of the pipe body material, and the drive current data of the actuator. The model extracts the non-linear relationship between the blockage clearing times and the wear rate through the convolutional layer, combines the attention mechanism to associate the abnormal wear characteristics under high dust concentration conditions, and outputs a remaining life prediction curve. The horizontal axis of the curve is the time dimension, the vertical axis is the percentage of the remaining life, and the slope of the curve reflects the dynamic change trend of the brush head wear rate under different operating conditions. The model training data is sourced from the actual replacement cycles in historical maintenance records and laboratory accelerated wear test data, and is adapted to the operating environment differences of different power plants through the transfer learning algorithm.

[0056] The safe operation threshold curve construction module generates a segmented threshold curve based on the theoretical life parameters provided by the equipment manufacturer and historical failure statistics results. The threshold curve adopts a linear decay mode in the initial stage and switches to an exponential decay mode when the cumulative blockage clearing times exceed the critical value, matching the fatigue characteristics of the brush head material. The overlay comparison module uses the dynamic time warping algorithm to align the time axes of the prediction curve and the threshold curve, calculates the minimum distance value between the two curves, and determines the crossover point where the predicted value is first lower than the threshold as the spare part replacement trigger time node. The time error range of the trigger node is calibrated through the sliding window mechanism to eliminate misjudgments caused by short-term operating condition fluctuations.

[0057] The work order instruction packaging module binds the trigger time node with the equipment code and maintenance suggestion measures to generate a structured data packet. The data packet is packaged in a lightweight JSON format and sent through the low-power wide-area network transmission channel of the LoRa wireless communication protocol. A timestamp synchronization calibration mechanism is added during the transmission process to keep the clock of the power plant management system and the on-site equipment synchronized at the millisecond level. After parsing the content of the data packet, the maintenance task queue receiving module calls the application program interface of the power plant work order system, inserts the maintenance task into the queue according to the preset priority, and at the same time triggers the spare part inventory query and personnel scheduling processes to form a closed-loop maintenance link from prediction to execution.

[0058] The above steps achieve accurate prediction of equipment maintenance cycles through multi-source data fusion and dynamic threshold comparison. The model input integrates operating parameters and historical data. The prediction curve reflects the actual wear state. Threshold comparison determines the replacement node. The communication protocol completes instruction transmission. Finally, it is integrated into the power plant maintenance system. The logical relationship between each step is as follows: The data-driven model predicts, the threshold constraints the decision boundary, the protocol realizes information flow, forming a complete early warning system from equipment status monitoring to maintenance resource scheduling, and improving the preventive maintenance ability of the thermal power generation system.

[0059] Specifically, the online blockage clearing control method for the pressure sampler described in the present invention further includes: Before the cylinder performs the blockage clearing operation, a quantum entanglement channel between the sensor and the controller is established through the quantum key distribution protocol; When the start signal of the cylinder push rod action is triggered, the transmission link of the pressure sampling data is interrupted through the channel, and the pressure data within the current time window is cached; After receiving the cylinder reset completion signal, the real-time transmission of the pressure sampling data is restored and the data caching state is released.

[0060] In the online blockage clearing control method for the pressure sampler described in the present invention, the data transmission control steps before and during the cylinder blockage clearing operation are as follows: The quantum key distribution protocol is activated in the blockage clearing operation preparation stage. A quantum random number sequence is generated through a single photon source. A quantum entanglement channel based on polarization coding is established between the sensor and the controller. During the channel establishment process, both parties exchange basis vector selection information and verify the channel security through error rate detection, generating an uncopyable encryption key pair. This key pair is used for the encryption verification of the subsequent data transmission link to prevent the tampering of control instructions by external interference signals during the blockage clearing process.

[0061] The start signal of the cylinder push rod action is triggered by a position sensor. The rising edge of the signal drives an optical switch to cut off the physical transmission link of the pressure sampling data. The data caching module captures the pressure data within the current time window at the moment of link interruption and stores it in the circular buffer of the non-volatile memory. The cached data is appended with a timestamp and a device status mark, recording the system clock value, ambient temperature, and pipe body vibration intensity parameters at the starting moment of the cylinder action, forming a complete working condition snapshot data packet for subsequent data integrity verification.

[0062] The cylinder reset completion signal is generated after the Hall sensor detects the return position of the push rod, and the falling edge of the signal triggers the transmission link recovery instruction. The controller sends a synchronous handshake signal through the quantum entanglement channel to verify the consistency between the cached data hash value at the sensor end and the control end. After successful verification, the optical switch of the real-time transmission link is re-closed, and the pressure sampling data is released frame by frame in a sliding window mode and seamlessly connected to the subsequent real-time data stream through the timestamp alignment mechanism. After the cached data is unlocked, it is transferred to the historical database for archiving, and at the same time, the temporary storage content of the circular buffer is cleared to release memory resources.

[0063] The above steps construct a data isolation and protection mechanism during the plug removal operation through quantum encryption and hardware-level link control. Channel establishment ensures the security of control instructions, link interruption prevents data contamination, cache storage retains the working condition snapshot, and reset recovery realizes data coherence. The logical relationship between the steps is as follows: the key protocol is the security foundation, the action signal drives the link switch, and the reset signal triggers the synchronous recovery, forming a closed-loop control process from operation preparation to execution protection and then to status recovery, effectively isolating the interference of mechanical actions on the pressure sampling data.

[0064] Specifically, the online plug removal control method for the pressure sampler described in the present invention further includes: in high-temperature working conditions, the quantum sensor and the cylinder drive module are encapsulated in a high-temperature isolation cavity made of multi-layer ceramic composite materials, and a thermocouple is built in the cavity to monitor the cavity temperature in real time; The temperature data collected by the thermocouple is input into the temperature compensation algorithm to dynamically correct the thermal drift error component in the differential pressure data output by the quantum sensor.

[0065] In the online plug removal control method for the pressure sampler described in the present invention, the equipment encapsulation and data compensation steps under high-temperature working conditions are as follows: The multi-layer ceramic composite material cavity adopts an alternating laminated structure of alumina and silicon nitride, with microporous heat dissipation channels provided on the outer layer and a vacuum insulation chamber integrated on the inner layer. During cavity assembly, the quantum sensor and the cylinder drive module are fixed through a high-temperature ceramic substrate, and the surface of the substrate is coated with a silicon carbide anti-oxidation coating to prevent changes in contact impedance caused by high-temperature oxidation. The cavity sealing interface adopts a metal-ceramic gradient transition structure to match the thermal expansion coefficient differences of different materials and avoid seal failure caused by sudden temperature changes. Three groups of K-type thermocouples are arranged inside the cavity, which are respectively installed in the sensor signal line interface, the cylinder solenoid valve housing, and the heat dissipation channel inlet area to form a temperature monitoring network.

[0066] The temperature data collected by the thermocouple is transmitted to the signal conditioning module via a shielded cable. The module is built with an anti-aliasing filter to eliminate high-frequency electromagnetic interference and compensates for the non-linear response characteristics of the thermocouple through a polynomial interpolation algorithm. The data preprocessing unit fuses multiple temperature signals into a cavity temperature distribution field and extracts the temperature gradient eigenvalue in the sensor installation area. The temperature field eigenvalue and the original differential pressure data output by the quantum sensor are synchronously input into the compensation algorithm to establish a temperature-differential pressure drift correlation matrix.

[0067] The temperature compensation algorithm constructs a thermal drift error model based on historical calibration data. The model fits a cubic polynomial correction function through the sensor output deviation curves under different temperature conditions. During the dynamic compensation process, the algorithm calculates the difference between the current temperature field eigenvalue and the reference temperature in real time, calls the corresponding polynomial coefficient matrix, and corrects the DC component and the high-frequency noise component in the differential pressure data respectively. For the scenario of rapid temperature fluctuations, the compensation module uses a sliding window mechanism to analyze the short-term temperature change trend and adaptively adjusts the response speed of the correction function to suppress the compensation lag effect caused by transient thermal shocks.

[0068] The above steps, through the collaborative design of the high-temperature resistant packaging structure and dynamic temperature compensation, solve the problem of interference of the high-temperature environment on the quantum sensing accuracy. The cavity structure provides physical protection, the thermocouple network monitors the temperature distribution, and the compensation algorithm realizes data correction, forming a complete technical link from environmental isolation to data calibration. The logical relationship between the steps is as follows: material selection and structure design lay the foundation for high-temperature resistance, the temperature monitoring network provides input data, the algorithm model realizes dynamic error elimination, ultimately ensuring the accuracy of the sensor output data and improving the operation reliability of the system under extreme working conditions.

[0069] In a second aspect, the present invention provides an on-line clogging removal control system for a pressure sampler, which is applied to the on-line clogging removal control method for a pressure sampler as described above, and includes: A quantum differential pressure sensor module, integrated on the inner walls of the fourth and seventh pipe bodies of the sampler pipe body, collects the differential pressure fluctuation data of the pipeline in real time through the quantum tunneling effect and outputs a differential pressure sequence including time stamps. The quantum differential pressure sensor operates in the temperature range of -40°C to 600°C, and the baseline differential pressure data is the calibration data in the initial commissioning stage of the equipment; a data processing module, connected to the quantum differential pressure sensor module, receives the differential pressure sequence and inputs it into a pre-trained quantum neural network model, combines the dust deposition characteristics in the historical flow field parameter database, and outputs the clogging risk levels and dynamic cleaning paths of the fourth and seventh pipe bodies, and at the same time calls the scenario gene data in the genetic algorithm parameter library that matches the current working conditions; The control module receives the dynamic cleaning path and the scenario gene data, generates a clogging removal control instruction including the cylinder jogging frequency, the rotating speed of the spiral brush head, and the cleaning sequence, and establishes a quantum entanglement channel with the quantum differential pressure sensor module through the quantum key distribution protocol to freeze the pressure sampling data transmission within the cylinder action window; The execution module includes a double-cylinder drive device and a spiral brush head with adjustable depth, responds to the clogging removal control instruction, dynamically adjusts the brush head rotation speed according to the dust accumulation thickness on the inner wall of the diversion groove, and switches to an alternating action mode during the start-up and shutdown stages of the fluidized bed boiler; The verification module is connected to the quantum differential pressure sensor module, calls the baseline differential pressure data before cleaning from the historical database, calculates the differential pressure deviation value after cleaning, and if the deviation value exceeds the threshold, sends a secondary cleaning instruction including the target pipe body coordinates to the control module; The maintenance warning module generates a spare part replacement work order by superimposing the cumulative clogging removal times based on the remaining life prediction curve of the spiral brush head output by the quantum neural network model, and pushes it to the maintenance queue of the power plant management system through the LoRa wireless communication protocol; The high-temperature protection module encapsulates the quantum differential pressure sensor module and the execution module in a multi-layer ceramic composite cavity, internally installs a thermocouple to collect the cavity temperature data, and corrects the thermal drift error in the differential pressure data through a temperature compensation algorithm.

[0070] For the on-line clogging removal control system of the pressure sampler provided by the present invention, the technical solutions and logical relationships of each module are as follows: A bidirectional communication link is established between the quantum differential pressure sensor module and the data processing module through an optical fiber channel. After the differential pressure sequence collected by the sensor is marked with a time stamp, it is transmitted to the input buffer of the data processing module in the form of a data frame. When the data processing module parses the data frame, it synchronously calls the dust particle size distribution characteristics in the historical flow field parameter database, matches the spatio-temporal dimensions of the current working condition and the historical scenario through a data alignment algorithm, and forms a multi-dimensional feature vector to input into the quantum neural network model. The clogging risk level label and the dynamic cleaning path instruction output by the model are transmitted to the instruction queue of the control module through a high-speed serial bus.

[0071] After receiving the dynamic cleaning path, the control module indexes the scenario gene data corresponding to the current dust concentration from the genetic algorithm parameter library, and performs a spatio-temporal coupling operation on the path coordinate sequence and the cylinder thrust parameter. During the coupling process, the pipe section number in the path coordinate system is mapped to the robotic arm movement trajectory of the execution module, and the cylinder thrust parameter is converted into the duty cycle parameter of the driving pulse. The quantum key distribution protocol completes the identity authentication before the instruction is issued, establishes a quantum entanglement channel between the sensor and the controller, and the channel state monitoring unit real-time detects the photon loss rate and dynamically adjusts the encryption key update frequency to maintain communication security.

[0072] After the dual-cylinder drive device of the execution module receives the blockage clearing control instruction, the push rod motion controller analyzes the time-frequency characteristics of the pulse signal and drives the spiral brush head to perform cleaning actions along the inner wall of the pipe. The diversion groove depth adjustment unit dynamically adjusts the micro-step amount of the stepping motor through the PID control algorithm according to the dust accumulation thickness data fed back by the quantum sensor, so that the depth of the diversion groove forms an inverse linkage relationship with the accumulation thickness. When the vibration sensor detects the characteristic frequency spectrum during the start-up and shutdown stages of the fluidized bed boiler, the execution module switches to the dual-cylinder alternating pulse mode, the push rod stroke is shortened, and the brush head rotation speed is increased to cope with the instantaneous high dust load impact.

[0073] After the blockage clearing operation is completed, the verification module starts the differential pressure data comparison process. When calling the baseline data under the same working conditions from the historical database, the sliding window matching algorithm is used to eliminate the time drift error. The deviation matrix generated by the differential calculation is converted into a three-dimensional thermal map through the spatial interpolation algorithm, and the spatial coordinates of the over-standard area are marked. The secondary cleaning instruction generation unit converts the coordinate information into the local path parameters of the execution module and superimposes them on the original cleaning sequence. The control module generates an incremental cleaning instruction based on this, increases the brush head rotation speed and extends the residence time for the target area.

[0074] The maintenance warning module shares the output interface of the quantum neural network with the data processing module. The accumulated blockage clearing times and working condition parameters are asynchronously transmitted to the wear prediction model through the message queue. After the intersection point of the remaining life curve and the safety threshold curve output by the model is time-calibrated, a work order trigger event is generated. The LoRa communication module adopts a star network topology, encapsulates the work order data into a data packet that conforms to the interface protocol of the power plant management system, and optimizes the transmission distance and energy consumption balance through the adaptive power adjustment mechanism to ensure the reliable delivery of the work order instruction.

[0075] The thermocouple array of the high-temperature protection module monitors the cavity temperature distribution in real time. The temperature compensation algorithm accesses the original differential pressure data of the quantum sensor through the shared memory interface. The compensated data is synchronized to the historical database of the data processing module via the data bus to update the baseline differential pressure reference value. The wind speed controller of the heat dissipation channel dynamically adjusts the rotation speed of the micro-turbine according to the cavity temperature gradient, so that the working temperatures of the sensor and the actuator are maintained within the material tolerance threshold, while reducing the impact of heat dissipation energy consumption on the overall power consumption of the system.

[0076] The coordination mechanism between modules achieves millisecond-level time alignment through the central clock synchronization unit, and the clock signal is distributed to all modules via the fiber optic network. The exception handling unit monitors the data flow status between modules. When an instruction loss or data timeout without response is detected, it triggers the redundant instruction retransmission mechanism and recovers the most recent valid data frame from the cache. The system forms a closed-loop control through a hierarchical feedback link: sensor data drives analysis and decision-making, the execution result triggers effect verification, the verification deviation guides path optimization, and the device status triggers maintenance warning, ultimately realizing the self-awareness, self-decision-making, and self-optimization operation of the pressure measurement point system for thermal power generation.

[0077] The present invention provides a specific implementation method and system for on-line blockage clearing control of a pressure sampler. Through the integration of quantum sensing technology and intelligent algorithms, it realizes on-line monitoring and adaptive blockage clearing of the pipelines at key measurement points in thermal power generation. The following is an explanation in combination with the specific steps of the technical solution.

[0078] The quantum differential pressure sensor module is integrated on the inner walls of the fourth and seventh pipe bodies of the sampler pipe body. It uses the quantum tunneling effect to collect the differential pressure fluctuation data on the inner wall of the pipeline in real time and generates a differential pressure sequence containing timestamps. The surface of the sensor is covered with a nano dust-proof coating to inhibit coal powder adhesion, and is encapsulated with a multi-layer ceramic composite material to withstand high-temperature environments. The historical flow field parameter database stores the dust deposition characteristic data under different working conditions, including air flow velocity, dust concentration, and pipe section position information.

[0079] The data processing module receives the differential pressure sequence output by the quantum differential pressure sensor module and inputs it into a pre-trained quantum neural network model. This model extracts the differential pressure fluctuation characteristics through the convolutional layer, combines the attention mechanism to associate with the dust deposition characteristics in the historical database, and outputs the blockage risk levels of the fourth and seventh pipe bodies at multiple future moments. The risk level is mapped to three-level labels of high, medium, and low according to the threshold interval where the probability peak is located in the blockage probability distribution map.

[0080] Based on the blockage risk level, the quantum annealing algorithm generates a dynamic cleaning path within the preset cleaning path search space. The search space is divided according to the risk level, and the high-risk level corresponds to the priority cleaning range of the seventh pipe body. The quantum annealing energy function takes the linear combination of the cleaning energy consumption weight and the dust removal efficiency weight as the optimization goal, traverses the solution space through the tunneling effect of quantum bits, and outputs the alternating cleaning sequence of the fourth and seventh pipe bodies and the cylinder action time window. The genetic algorithm parameter library stores the gene data for different scenarios, including the cylinder thrust coefficient and the spiral brush head rotation speed threshold under the high-sulfur coal working condition. The control module analyzes the dust concentration and air flow velocity parameters of the current working condition, calls the matching scenario gene data, and couples it with the cleaning sequence output by the quantum annealing algorithm in space and time to generate a blockage clearing control instruction including the cylinder jogging frequency and the brush head rotation speed.

[0081] The execution module responds to the control instruction and drives the double-cylinder device and the spiral brush head with adjustable depth to perform the blockage clearing operation. The data of the dust accumulation thickness on the inner wall of the diversion groove is fed back in real time by the quantum sensor. The proportional-integral algorithm dynamically adjusts the depth of the diversion groove according to the accumulation thickness, so that the groove depth is negatively correlated with the accumulation thickness. When vibration signals are detected during the start-up and shutdown stages of the fluidized bed boiler, the control module switches to the alternating action mode of the double cylinders and superimposes high-frequency pulse signals on the instruction to cope with the instantaneous high-dust impact.

[0082] Within the preset time window of the cylinder push rod action, the quantum sensor and the controller establish a quantum entanglement channel through the quantum key distribution protocol, interrupt the transmission link of the pressure sampling data and cache the current data to isolate the backwashing air flow interference. After the blockage clearing operation is completed, the verification module calls the baseline pressure difference data in the historical database and calculates the pressure difference deviation value after cleaning. If the deviation value exceeds the preset threshold, a secondary cleaning instruction containing the coordinates of the target pipe body is sent to the control module to trigger the spiral brush head to perform supplementary cleaning on the local area.

[0083] The maintenance warning module inputs the cumulative blockage clearing times of the spiral brush head into the pre-trained quantum neural network wear prediction model and outputs the remaining life prediction curve. After this curve is superimposed and compared with the safe operation threshold curve, a spare part replacement work order is generated and pushed to the maintenance queue of the power plant management system through the LoRa wireless communication protocol. The high-temperature resistant protection module internally installs a thermocouple to monitor the cavity temperature in real time, and the temperature compensation algorithm corrects the thermal drift error component output by the quantum sensor according to the thermocouple data.

[0084] This embodiment realizes the dynamic prediction and precise cleaning of the blockage risk through the cooperation of quantum sensing, the quantum annealing algorithm and the genetic algorithm parameter library. At the same time, the data reliability is guaranteed through the quantum entanglement channel and the temperature compensation mechanism, forming a non-disassembly and adaptive online blockage clearing closed-loop control.

[0085] The embodiment of the present invention provides an online blockage clearing control method and system for a pressure sampler, which is applied to the online maintenance of the flue gas pressure measuring point of a pulverized coal boiler in a thermal power plant. The implementation process of the technical solution is described below in combination with specific implementation scenarios.

[0086] Embodiment 1: Prediction of pipeline blockage risk and dynamic cleaning. The quantum differential pressure sensor module is integrated on the inner walls of the fourth and seventh pipe bodies of the sampler, and differential pressure data is collected at intervals of 1 second to generate a differential pressure fluctuation sequence with time stamps. The surface of the sensor is coated with a nano dust-proof coating, and the temperature tolerance range is -40°C to 600°C to inhibit dust adhesion. The differential pressure sequence is input into the pre-trained quantum neural network model. The model extracts the differential pressure fluctuation characteristics through the convolutional layer, and combines the dust deposition characteristic data of the fluidized bed boiler in the historical database to output the blockage risk levels of the fourth and seventh pipe bodies within the next 5 minutes. When the blockage probability of the seventh pipe body exceeds 70%, the model marks it as a high-risk level label.

[0087] The quantum annealing algorithm divides the cleaning path search space according to the risk level, constructs an energy function with a linear combination of the energy consumption weight of 0.6 and the cleaning efficiency weight of 0.4, solves the optimal solution through the qubit tunneling effect, and generates a cleaning sequence with the seventh tube body being prioritized. The genetic algorithm parameter library calls the gene data of the high-sulfur coal working condition and matches the cylinder thrust coefficient of 0.8 and the spiral brush head rotation speed threshold of 400 rpm. The control module couples the cleaning sequence with the scenario gene data to generate a plugging removal instruction with a cylinder jogging frequency of 2 Hz and a brush head rotation speed of 380 rpm.

[0088] The double-cylinder device of the execution module drives the spiral brush head into the seventh tube body. According to the dust accumulation thickness on the inner wall of the diversion groove (the detected value is 3 mm), the depth of the diversion groove is dynamically adjusted to 1.5 mm through the proportional-integral algorithm. When the start-stop vibration signal of the fluidized bed boiler is detected (frequency > 20 Hz), the double-cylinder alternating mode is switched, and a high-frequency pulse signal (frequency 5 Hz) is superimposed to enhance the instantaneous cleaning intensity.

[0089] Example 2: Pressure data synchronization and secondary cleaning verification Within the 0.2-second time window of the cylinder push rod action, the quantum sensor and the control module establish an entangled channel through the quantum key distribution protocol, interrupt the pressure data transmission, and cache the current data. After the plugging removal is completed, the verification module calls the baseline pressure difference data before cleaning (12.5 Pa) and calculates the pressure difference deviation value after cleaning (0.3 Pa). When the deviation value exceeds the threshold of 0.5 Pa, a local cleaning instruction for the coordinates of the seventh tube body (X = 15, Y = 30) is sent to the control module, triggering the spiral brush head to perform supplementary cleaning on the target area 2 times.

[0090] Example 3: Maintenance warning and high-temperature environment adaptation The maintenance warning module inputs the cumulative plugging removal times of the spiral brush head (1200 times) into the quantum neural network wear prediction model and outputs the remaining life prediction curve. When the predicted remaining life is 15 days, a spare part replacement work order is generated and pushed to the power plant MIS system through the LoRa wireless communication protocol (frequency band 433 MHz). The high-temperature resistant protection module encapsulates the sensor and the execution unit in a multi-layer ceramic composite cavity (temperature resistance 200 °C), and a built-in thermocouple monitors the cavity temperature in real time (the detected value is 150 °C). The temperature compensation algorithm corrects the thermal drift error component of the pressure difference sensor according to the thermocouple data (correction amount ±0.05 Pa).

[0091] Example 4: The system collaboratively controls the process. The quantum differential pressure sensor monitors the differential pressure data of the fourth pipe body in real time. The data processing module outputs a medium-risk level label, and the control module generates a priority cleaning instruction for the fourth pipe body. The execution module adjusts the depth of the diversion channel to 2.0 mm, and the cylinder drives the brush head to perform cleaning at a frequency of 1 Hz. During the cleaning process, the quantum entanglement channel freezes the pressure sampling data, and the verification module confirms that the differential pressure has recovered to the baseline value of 12.5 Pa (deviation < 0.1 Pa). The maintenance warning module updates the cumulative blockage cleaning times of the brush head, predicts the remaining life, and pushes a maintenance work order.

[0092] In the embodiment of the present invention, through the collaboration of quantum sensing, dynamic path planning, and sub-scenario parameter invocation, real-time response to blockage risks and precise cleaning are achieved. At the same time, the measurement reliability is ensured through data synchronization and high-temperature compensation mechanisms, forming an online blockage cleaning solution with closed-loop control.

[0093] Through the collaborative optimization of quantum sensing and intelligent algorithms, the technical solution of the present invention effectively solves the problem of static clustering deviation in the existing multi-level index structure. Based on the dynamic differential pressure data collected in real time by the quantum differential pressure sensor, combined with the multi-moment prediction of pipeline blockage risks by the quantum neural network model, dynamic clustering update of the index structure is realized. The cleaning path search space is generated by the quantum annealing algorithm, and the priority is dynamically divided according to the real-time risk level, replacing the traditional static clustering rule, avoiding the deviation of the clustering center caused by environmental changes, and thus eliminating the influence of static clustering deviation on the index accuracy.

[0094] Aiming at the insufficient cross-modal feature fusion, the technical solution adopts the multi-layer feature extraction mechanism of the quantum neural network model. The convolutional layer extracts the differential pressure fluctuation features, the attention mechanism correlates cross-modal parameters such as dust deposition, and the fully connected layer outputs the probability distribution map to achieve the deep fusion of multi-source data. The genetic algorithm parameter library calls the scenario gene data, spatiotemporally couples the cleaning sequence with the working condition parameters, forms an adaptive matching of cross-modal features, and enhances the collaborative analysis ability of multi-dimensional data during the indexing process.

[0095] In terms of hierarchical semantic association, the system establishes a hierarchical cleaning priority through the blockage risk level label, and combines the alternating cleaning sequence output by the quantum annealing algorithm to construct a semantic association network between multi-level pipe bodies. The verification module triggers a secondary cleaning instruction based on the differential pressure deviation value, forming a closed-loop feedback mechanism to ensure data consistency between levels. The maintenance warning module establishes a semantic mapping between the equipment state and the maintenance strategy by superimposing and comparing the remaining life prediction curve with the safety threshold, completing the missing hierarchical semantic chain in the traditional index.

Claims

1. A pressure sampler online clearing control method, characterized in that: include: Based on the quantum differential pressure sensor, real-time differential pressure data of the inner wall of the sampler pipeline is collected to generate a differential pressure fluctuation sequence including a timestamp; The pressure difference fluctuation sequence is input into a pre-trained quantum neural network model, and combined with the dust deposition characteristic data in the historical flow field parameter database, the pipeline blockage risk levels at multiple future moments are output; According to the blockage risk level, a dynamic cleaning path is generated in a preset cleaning path search space by using a quantum annealing algorithm, and scene gene data matching the current working condition in a genetic algorithm parameter library is called to determine the cleaning action priorities of the fourth pipe body and the seventh pipe body; Based on the dynamic cleaning path and cleaning action priority, a control instruction is sent to the cylinder drive module to drive the spiral brush head to perform a clearing operation according to the set inching frequency and rotation speed; In the preset time window of the cylinder push rod action, the transmission of pressure sampling data is synchronously frozen through the quantum entanglement channel of the quantum sensor and the controller to isolate the interference of the back-blowing airflow on the pressure data; The real-time differential pressure data after the clearing operation is completed is collected through the quantum differential pressure sensor, and the deviation value between the real-time differential pressure data and the baseline differential pressure data before clearing is calculated; If the deviation value exceeds a preset threshold, a secondary local cleaning instruction is sent to the control module to trigger the spiral brush head to perform additional cleaning on the local area of ​​the target tube body; Based on the prediction results of the wear rate of the spiral brush head using the quantum neural network model and the cumulative number of clearing times, a remaining life prediction curve is constructed to generate a maintenance cycle warning signal and push it to the power plant management system through the wireless communication module.

2. The pressure sampler online blockage clearing control method according to claim 1 is characterized in that: The method of predicting the pipeline blockage risk level by using a quantum neural network includes: Inputting the real-time pressure difference change and the historical flow field parameters into a pre-trained quantum neural network model, extracting the pressure difference fluctuation characteristics through the convolution layer of the pre-trained quantum neural network model, and associating the dust deposition characteristics through the attention mechanism; Outputting a blockage probability distribution map at multiple future moments based on the fully connected layer of the quantum neural network model, wherein the probability distribution map represents the dust accumulation probability of different pipe sections; According to the preset threshold interval where the probability peak in the congestion probability distribution graph is located, it is mapped to a high risk, medium risk or low risk level label.

3. The pressure sampler online blockage clearing control method according to claim 1 is characterized in that: The method of generating a dynamic cleaning path by using a quantum annealing algorithm comprises: Divide the cleaning path search space of the fourth tube body and the seventh tube body according to the risk level label, wherein a high risk level corresponds to a priority cleaning range of the seventh tube body; Constructing a quantum annealing energy function in the search space, wherein the energy function takes a linear combination of a cleaning energy consumption weight and a dust removal efficiency weight as an optimization target; The solution space of the energy function is traversed through the tunneling effect of quantum bits, and the alternating cleaning sequence of the fourth tube body and the seventh tube body and the corresponding cylinder action time window are output.

4. The pressure sampler online blockage clearing control method according to claim 1 is characterized in that: The scene gene data in the matching genetic algorithm parameter library includes: Analyze the dust concentration and airflow velocity parameters in the current working environment characteristics, and call the scene gene data matching the parameters from the genetic algorithm parameter library, wherein the scene gene data includes the cylinder thrust coefficient and the spiral brush head speed threshold under the high-sulfur coal working condition; The cylinder thrust parameters in the scene gene data are coupled in time and space with the cleaning sequence output by the quantum annealing algorithm to generate a clearing control instruction including the cylinder jog frequency and the brush head rotation speed.

5. The pressure sampler online blockage clearing control method according to claim 1 is characterized in that: The control cylinder drives the spiral brush head to perform a high-frequency inching clearing operation, including: According to the dust accumulation thickness data on the inner wall of the guide groove fed back by the quantum sensor, the depth parameter of the spiral guide groove is dynamically adjusted through the proportional integral algorithm, so that the groove depth is negatively correlated with the accumulation thickness; Based on the adjusted guide groove depth, the corresponding cylinder inching frequency upper limit value and brush head rotation speed reference value are matched from the genetic algorithm parameter library; When the vibration signal of the fluidized bed boiler in the start-up and shutdown stages is detected, the double-cylinder alternating action mode is switched, and a high-frequency pulse signal is superimposed in the clearing control instruction to cope with instantaneous high dust impact.

6. The pressure sampler online blockage clearing control method according to claim 1 is characterized in that: The verification of the effectiveness of the clearing operation includes: The real-time differential pressure data after the clearing operation is completed is collected through the quantum differential pressure sensor, and the baseline differential pressure data before clearing is called from the historical database; Input the real-time pressure difference data and the baseline pressure difference data into a differential calculation module to generate a pressure difference deviation value and mark the corresponding pipe body position; If the pressure difference deviation value exceeds the preset secondary cleaning threshold, a local cleaning instruction including the position coordinates of the tube body is sent to the control module to trigger the spiral brush head to perform additional cleaning on the target area.

7. The pressure sampler online blockage clearing control method according to claim 1 is characterized in that: The generating of the maintenance cycle warning signal comprises: The cumulative number of times the spiral brush head clears blockages is input into the pre-trained quantum neural network wear prediction model, and the remaining life prediction curve is output; The remaining life prediction curve is superimposed and compared with a preset safe operation threshold curve to determine a spare parts replacement trigger time node; The spare parts replacement trigger time node is encapsulated as a work order instruction through the LoRa wireless communication protocol and pushed to the maintenance task queue of the power plant management system.

8. The pressure sampler online blockage clearing control method according to claim 1 is characterized in that: Also includes: Before the cylinder performs the blockage clearing operation, a quantum entanglement channel between the sensor and the controller is established through the quantum key distribution protocol; When the start signal of the cylinder push rod action is triggered, the transmission link of the pressure sampling data is interrupted through the channel, and the pressure data in the current time window is cached; When the cylinder reset completion signal is received, the real-time transmission of the pressure sampling data is resumed and the data cache state is released.

9. The pressure sampler online blockage clearing control method according to claim 1, characterized in that: Also includes: Under high temperature conditions, the quantum sensor and the cylinder drive module are encapsulated in a high temperature resistant isolation cavity made of multilayer ceramic composite materials, and a thermocouple is built into the cavity to monitor the cavity temperature in real time; The temperature data collected by the thermocouple is input into a temperature compensation algorithm to dynamically correct the thermal drift error component in the pressure difference data output by the quantum sensor.

10. An online blockage clearing control system for a pressure sampler, applied to the online blockage clearing control method for a pressure sampler as claimed in any one of claims 1 to 9, characterized in that: include: A quantum pressure difference sensor module is integrated in the inner wall of the fourth tube body and the seventh tube body of the sampler tube body, collects pipeline pressure difference fluctuation data in real time through quantum tunneling effect, and outputs a pressure difference sequence containing a timestamp; a data processing module is connected to the quantum pressure difference sensor module, receives the pressure difference sequence and inputs the pre-trained quantum neural network model, combines the dust deposition characteristics in the historical flow field parameter database, outputs the blockage risk level and dynamic cleaning path of the fourth tube body and the seventh tube body, and calls the scene gene data matching the current working condition in the genetic algorithm parameter library; The control module receives the dynamic cleaning path and scene gene data, generates a clearing control instruction including the cylinder jog frequency, the spiral brush head rotation speed and the cleaning sequence, and establishes a quantum entanglement channel with the quantum pressure difference sensor module through the quantum key distribution protocol, and freezes the pressure sampling data transmission within the cylinder action window; the execution module includes a dual-cylinder drive device and a spiral brush head with adjustable depth, responds to the clearing control instruction, dynamically adjusts the brush head rotation speed according to the dust accumulation thickness on the inner wall of the guide groove, and switches to the alternating action mode during the start-up and shutdown stage of the fluidized bed boiler; the verification module is connected to the quantum pressure difference sensor module, and retrieves the information from the historical database. The baseline pressure difference data before cleaning is used to calculate the pressure difference deviation value after cleaning. If the deviation value exceeds the threshold, a secondary cleaning instruction containing the coordinates of the target pipe body is sent to the control module; the maintenance warning module generates a spare parts replacement work order based on the remaining life prediction curve of the spiral brush head output by the quantum neural network model and superimposes the cumulative number of clearing times, and pushes it to the maintenance queue of the power plant management system through the LoRa wireless communication protocol; the high temperature protection module encapsulates the quantum pressure difference sensor module and the execution module in a multi-layer ceramic composite cavity, uses a built-in thermocouple to collect cavity temperature data, and corrects the thermal drift error in the pressure difference data through a temperature compensation algorithm.

Images