Real-time data acquisition and transmission method in intelligent pulping station

Abstract

The invention relates to the technical field of data acquisition and transmission, in particular to a real-time data acquisition and transmission method in an intelligent pulping station, which comprises the following steps of: accessing multi-protocol compatible equipment and initializing parameters, establishing a unified communication interface of all equipment of the pulping station, and finishing initial parameter configuration of the equipment; multi-source heterogeneous data are collected in real time and preliminarily aggregated, multi-device data are synchronously collected through an edge gateway, protocol differences are eliminated, and standardized data streams are generated. Through edge layer intelligent filtering and self-adaptive threshold cleaning technologies, noise interference is effectively filtered out, abnormal sprouts are accurately captured, steam consumption is remarkably reduced on the premise that the concentration precision is guaranteed to be + / -0.1%, a data value chain from real-time control to remote operation and maintenance is opened through a full-life-cycle data engine, the equipment fault diagnosis timeliness is improved, and the equipment fault diagnosis efficiency is improved. And the process switching efficiency is improved, and the synergistic target of reducing the energy consumption and reducing the quality fluctuation is achieved on the whole.

Description

Technical Field

[0001] This invention relates to the field of data acquisition and transmission technology, specifically a method for real-time data acquisition and transmission in an intelligent pulping station. Background Technology

[0002] The pulping station is the core production unit in the papermaking industry. It is primarily responsible for breaking down plant fiber raw materials such as wood and bamboo into pulp using chemical or mechanical methods, providing the basic raw materials for subsequent papermaking processes. Its production process involves multiple complex steps, including raw material pretreatment, cooking, washing, screening, and bleaching. It requires precise control of numerous process parameters such as temperature, pressure, concentration, and pH value, which directly determine the final pulp quality and production cost.

[0003] Typically, traditional pulping plants rely on decentralized, independent control systems for data acquisition. Each process's equipment communicates using proprietary protocols, creating "data silos" that hinder global collaboration. Manual inspections and fixed-threshold alarm mechanisms suffer from response lag; for example, temperature anomalies may not be detected for several minutes. Static sampling frequencies are insufficient to capture data during periods of rapid change in the cooking reaction, often missing critical process fluctuations. Historical data is stored in scattered and inconsistent formats, resulting in a lack of effective data support for production optimization and fault diagnosis, severely restricting improvements in pulping efficiency and quality stability.

[0004] In summary, a method for real-time data acquisition and transmission in intelligent pulping stations needs to be proposed to solve the above problems. Summary of the Invention

[0005] The purpose of this invention is to provide a method for real-time data acquisition and transmission in an intelligent pulping station, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: This invention proposes a method for real-time data acquisition and transmission in an intelligent pulping station, comprising the following steps: S1. Multi-protocol compatible device access and parameter initialization: Establish a unified communication interface for all equipment in the pulping station and complete the initial parameter configuration of the equipment; S2. Real-time acquisition and preliminary aggregation of multi-source heterogeneous data: Simultaneous acquisition of data from multiple devices through edge gateways, elimination of protocol differences, and generation of standardized data streams; S3. Real-time data cleaning and anomaly labeling based on dynamic thresholds: filtering noise, filling in missing values, and labeling abnormal data; S4. Extraction of pulp characteristics based on edge computing: Extract key features reflecting pulping status from cleaning data to reduce data dimensionality; S5. Real-time status modeling of the pulping process driven by digital twins: Constructing a virtual mirror of the physical pulping station, simulating the operating status through real-time data, and predicting future trends; S6. Intelligent optimization control with multi-objective collaboration, based on digital twin prediction and process requirements, dynamically adjusts the parameters of steam valves and dosing pumps to achieve a balance of multiple objectives: quality, energy consumption, and efficiency. S7. Actuator feedback and closed-loop verification: Verify the actual execution effect of control commands, correct model errors, and form a "acquisition-control-feedback" closed loop; S8. Full lifecycle data storage, visualization and remote operation and maintenance, enabling efficient data storage, intuitive presentation and remote monitoring, supporting process optimization and fault diagnosis.

[0007] Preferably, the implementation process of step S1 is as follows: S1.1. Automatic Equipment Type Identification: The system automatically scans connected equipment and distinguishes between temperature sensors, pressure transmitters, pH meters, online concentration meters, motor frequency converters, and steam valves. S1.2. Load Protocol Driver: Match the communication protocol library according to the device type, support ModbusTCP, MQTT, OPCUA and custom binary protocols, and automatically load the corresponding driver module; S1.3. Set sampling period: Set the sampling frequency to 500ms / time for core parameters, 1s / time for auxiliary parameters, and 200ms / time for device status; S1.4. Generate a unified data template: All data output is in JSON format, with fields including "Device ID", "Timestamp", "Original Value", "Unit" and "Protocol Type".

[0008] Preferably, the implementation process of step S2 is as follows: S2.1. Concurrent Data Acquisition: The edge gateway establishes a maximum of 200 concurrent connections and polls each device at the sampling period set in S1.3; S2.2. Protocol Conversion Processing: ASCII code parsing is performed on custom protocol devices; ModbusTCP devices read register values ​​through function code 0x03 and convert them to engineering values; S2.3. Data caching and fault tolerance: Enable a 1GB local cache to temporarily store data when the network is interrupted, and prioritize retransmission after recovery; S2.4. Aggregation of similar data: Merge data from similar devices in 1-second time windows to shield against the impact of faulty nodes; S2.5. Output standardized data stream: The generated fields include "Device ID", "Timestamp", "Value", "Unit", and "Protocol Type"; The output of S2.5 is directly used as the input of S3; when a new device is added to S1, S2.2 automatically calls the new protocol conversion rules.

[0009] Preferably, the implementation process of step S3 is as follows: S3.1. Sliding window Kalman filter: A filter with a window size of 10 is used to process the input data of S2.5, balancing real-time performance and smoothness; S3.2. Missing value handling: If there are ≤3 consecutive missing points, use forward filling; if there are more than 3 missing points, use linear interpolation to fill them in. S3.3. Dynamic threshold calculation: The mean μ and standard deviation σ are recalculated every 30 minutes based on historical data, and the 3σ interval is updated; S3.4. Anomaly Classification Marking: Data exceeding 3σ is marked as "urgent anomaly", and data approaching the threshold is marked as "early warning"; S3.5. Add Data Quality Label: Add “Data Quality” to the output field, with values ​​including “Qualified”, “Forward Fill”, “Interpolation”, “Urgent Exception”, and “Warning”.

[0010] Preferably, the implementation process of step S4 is as follows: S4.1. Calculate dynamic characteristics: For the data stream output by S3.5, calculate the rate of temperature change and the variance of pressure fluctuation within a 1-second window; S4.2. Extract static features: Take the slurry concentration, pH value, and motor load rate of the 5-second average window; S4.3. Generate feature dataset: Combine dynamic and static feature outputs.

[0011] Preferably, the implementation process of step S5 is as follows: S5.1. Hybrid Model Loading: Based on the pulping chemical reaction mechanism model, an LSTM neural network is superimposed to learn nonlinear features; S5.2. Real-time model synchronization: The digital twin state is updated every second using the feature data output from S4.3; S5.3. Prediction and Error Monitoring: Output the concentration prediction value for the next 5 minutes. If the error of key parameters is >2%, trigger online correction. S5.4. Generate a status report.

[0012] Preferably, the implementation process of step S6 is as follows: S6.1. Target priority setting: The primary target is that the slurry concentration deviation is < ±0.1% and the viscosity is 50-80 mPa·s; the secondary target is that the steam consumption is < 1.2 t / ton of slurry. S6.2. Model Predictive Control: Every 2 seconds, based on the twin state of S5.4, the control sequence for the next 30 seconds is continuously optimized. S6.3. Generate control commands.

[0013] Preferably, the implementation process of step S7 is as follows: S7.1. Real-time feedback acquisition: The actual execution value is acquired within ≤500ms through the device status sensor; S7.2. Execution effect verification: Calculate the absolute deviation between the set value and the actual value, and the dynamic response time; S7.3. Output verification report; S7.4. Model calibration: If the calibration fails, adjust the LSTM model weights or mechanism parameters in S5.

[0014] Preferably, the implementation process of step S8 is as follows: S8.1. Tiered storage architecture: Time-series database stores 30 days of real-time data, while relational database permanently stores process parameters and alarm records; S8.2. Dynamic Visual Dashboard: Displays temperature / pressure trend curves, concentration / energy consumption indicators, and equipment health status; S8.3. Remote Operation and Maintenance Interface: Supports Web / APP adjustment of device parameters in S1, abnormal thresholds in S3, and control targets in S6.

[0015] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention effectively filters out noise interference and accurately captures the initial signs of anomalies through edge layer intelligent filtering and adaptive threshold cleaning technology; the digital twin model integrates physical mechanisms and deep learning algorithms to perform second-level dynamic simulation and prediction of the cooking reaction process, providing a priori decision basis for optimized control; the multi-objective collaborative control algorithm incorporates multiple dimensions such as quality, energy consumption, and equipment loss into a unified optimization framework, significantly reducing steam consumption while ensuring concentration accuracy of ±0.1%; and the full life cycle data engine connects the data value chain from real-time control to remote operation and maintenance, improving the timeliness of equipment fault diagnosis and the efficiency of process switching, and achieving the overall goal of synergistic efficiency improvement in reducing energy consumption and quality fluctuations. Attached Figure Description

[0016] Figure 1 A flowchart illustrating the method for real-time data acquisition and transmission in the intelligent pulping station of the present invention is shown. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1, please refer to Figure 1This invention proposes a method for real-time data acquisition and transmission in an intelligent pulping station, comprising the following steps: S1. Multi-protocol compatible device access and parameter initialization: Establish a unified communication interface for all equipment in the pulping station and complete the initial parameter configuration of the equipment; S2. Real-time acquisition and preliminary aggregation of multi-source heterogeneous data: Simultaneous acquisition of data from multiple devices through edge gateways, elimination of protocol differences, and generation of standardized data streams; S3. Real-time data cleaning and anomaly labeling based on dynamic thresholds: filtering noise, filling in missing values, and labeling abnormal data; S4. Extraction of pulp characteristics based on edge computing: Extract key features reflecting pulping status from cleaning data to reduce data dimensionality; S5. Real-time status modeling of the pulping process driven by digital twins: Constructing a virtual mirror of the physical pulping station, simulating the operating status through real-time data, and predicting future trends; S6. Intelligent optimization control with multi-objective collaboration, based on digital twin prediction and process requirements, dynamically adjusts the parameters of steam valves and dosing pumps to achieve a balance of multiple objectives: quality, energy consumption, and efficiency. S7. Actuator feedback and closed-loop verification: Verify the actual execution effect of control commands, correct model errors, and form a "acquisition-control-feedback" closed loop; S8. Full lifecycle data storage, visualization and remote operation and maintenance, enabling efficient data storage, intuitive presentation and remote monitoring, supporting process optimization and fault diagnosis.

[0019] In this embodiment, it should also be noted that the implementation process of step S1 is as follows: S1.1. Automatic Equipment Type Identification: The system automatically scans connected equipment and distinguishes between temperature sensors, pressure transmitters, pH meters, online concentration meters, motor frequency converters, and steam valves. S1.2. Load Protocol Driver: Match the communication protocol library according to the device type, support ModbusTCP, MQTT, OPCUA and custom binary protocols, and automatically load the corresponding driver module; S1.3. Set sampling period: Set the sampling frequency to 500ms / time for core parameters, 1s / time for auxiliary parameters, and 200ms / time for device status; S1.4. Generate a unified data template: All data output is in JSON format, with fields including "Device ID", "Timestamp", "Original Value", "Unit" and "Protocol Type".

[0020] In this embodiment, it should also be noted that the implementation process of step S2 is as follows: S2.1. Concurrent Data Acquisition: The edge gateway establishes a maximum of 200 concurrent connections and polls each device at the sampling period set in S1.3; S2.2. Protocol Conversion Processing: ASCII code parsing is performed on custom protocol devices; ModbusTCP devices read register values ​​through function code 0x03 and convert them to engineering values; S2.3. Data caching and fault tolerance: Enable a 1GB local cache to temporarily store data when the network is interrupted, and prioritize retransmission after recovery; S2.4. Aggregation of similar data: Merge data from similar devices in 1-second time windows to shield against the impact of faulty nodes; S2.5. Output standardized data stream: The generated fields include "Device ID", "Timestamp", "Value", "Unit", and "Protocol Type"; The output of S2.5 is directly used as the input of S3; when a new device is added to S1, S2.2 automatically calls the new protocol conversion rules.

[0021] In this embodiment, it should also be noted that the implementation process of step S3 is as follows: S3.1. Sliding window Kalman filter: A filter with a window size of 10 is used to process the input data of S2.5, balancing real-time performance and smoothness; S3.2. Missing value handling: If there are ≤3 consecutive missing points, use forward filling; if there are more than 3 missing points, use linear interpolation to fill them in. S3.3. Dynamic threshold calculation: The mean μ and standard deviation σ are recalculated every 30 minutes based on historical data, and the 3σ interval is updated; S3.4. Anomaly Classification Marking: Data exceeding 3σ is marked as "urgent anomaly", and data approaching the threshold is marked as "early warning"; S3.5. Add Data Quality Label: Add “Data Quality” to the output field, with values ​​including “Qualified”, “Forward Fill”, “Interpolation”, “Urgent Exception”, and “Warning”.

[0022] In this embodiment, it should also be noted that the implementation process of step S4 is as follows: S4.1. Calculate dynamic characteristics: For the data stream output by S3.5, calculate the rate of temperature change and the variance of pressure fluctuation within a 1-second window; S4.2. Extract static features: Take the slurry concentration, pH value, and motor load rate of the 5-second average window; S4.3. Generate feature dataset: Combine dynamic and static feature outputs.

[0023] In this embodiment, it should also be noted that the implementation process of step S5 is as follows: S5.1. Hybrid Model Loading: Based on the pulping chemical reaction mechanism model, an LSTM neural network is superimposed to learn nonlinear features; S5.2. Real-time model synchronization: The digital twin state is updated every second using the feature data output from S4.3; S5.3. Prediction and Error Monitoring: Output the concentration prediction value for the next 5 minutes. If the error of key parameters is >2%, trigger online correction. S5.4. Generate a status report.

[0024] In this embodiment, it should also be noted that the implementation process of step S6 is as follows: S6.1. Target priority setting: The primary target is that the slurry concentration deviation is < ±0.1% and the viscosity is 50-80 mPa·s; the secondary target is that the steam consumption is < 1.2 t / ton of slurry. S6.2. Model Predictive Control: Every 2 seconds, based on the twin state of S5.4, the control sequence for the next 30 seconds is continuously optimized. S6.3. Generate control commands.

[0025] In this embodiment, it should also be noted that the implementation process of step S7 is as follows: S7.1. Real-time feedback acquisition: The actual execution value is acquired within ≤500ms through the device status sensor; S7.2. Execution effect verification: Calculate the absolute deviation between the set value and the actual value, and the dynamic response time; S7.3. Output verification report; S7.4. Model calibration: If the calibration fails, adjust the LSTM model weights or mechanism parameters in S5.

[0026] In this embodiment, it should also be noted that the implementation process of step S8 is as follows: S8.1. Tiered storage architecture: Time-series database stores 30 days of real-time data, while relational database permanently stores process parameters and alarm records; S8.2. Dynamic Visual Dashboard: Displays temperature / pressure trend curves, concentration / energy consumption indicators, and equipment health status; S8.3. Remote Operation and Maintenance Interface: Supports Web / APP adjustment of device parameters in S1, abnormal thresholds in S3, and control targets in S6.

[0027] Example 2, please refer to Figure 1 In practical applications, the method for real-time data acquisition and transmission in the intelligent pulping station of this invention specifically includes the following steps: S1. Multi-protocol compatible device access and parameter initialization: S1.1. Automatically identify device type: The system automatically scans connected devices and distinguishes between device types such as temperature sensors (PT100, range 0-200℃), pressure transmitters (4-20mA, range 0-1.6MPa), pH meters (0-14), online concentration meters (electromagnetic, range 0-5%), motor frequency converters (Modbus RTU), and steam valves (electric adjustment, 4-20mA control signal); S1.2. Load protocol driver: It matches the communication protocol library according to the device type, supports ModbusTCP, MQTT, OPCUA and custom binary protocols, and automatically loads the corresponding driver module; S1.3. Set the sampling period: Set the sampling frequency to 500ms / time for core parameters (temperature, pressure), 1s / time for auxiliary parameters (pH, concentration), and 200ms / time for equipment status (motor current, valve opening). S1.4. Generate a unified data template: All data output is in JSON format, with fields including "Device ID", "Timestamp", "Original Value", "Unit", and "Protocol Type". This step provides the foundation for device communication for all subsequent steps; if S2 discovers a new, undefined device, it will trigger S1 to rescan and update the protocol library; S2. Real-time acquisition and preliminary aggregation of multi-source heterogeneous data: S2.1. Concurrent Data Acquisition: The edge gateway establishes a maximum of 200 concurrent connections and polls each device at the sampling period set in S1.3; S2.2. Protocol Conversion Processing: ASCII code parsing is performed on custom protocol devices (such as old instruments); ModbusTCP devices read register values ​​through function code 0x03 and convert them into engineering values; S2.3. Data Caching and Fault Tolerance: Enable a 1GB local cache to temporarily store data when the network is interrupted, and prioritize retransmission after network recovery; S2.4. Aggregation of similar data: Data from similar devices (e.g., averaging the values ​​of 4 temperature sensors) are merged within a 1-second time window to shield the impact of faulty nodes. S2.5. Output standardized data stream: The generated fields include "Device ID", "Timestamp", "Value", "Unit", and "Protocol Type" (e.g., {"device_id":"T-01","timestamp":"2024-03-10T10:00:00.500","value":185.6,"unit":"℃","protocol":"Modbus_TCP"}). The output of S2.5 is directly used as the input of S3; when a new device is added to S1, S2.2 automatically calls the new protocol conversion rules; S3. Real-time data cleaning and anomaly labeling based on dynamic thresholds: S3.1. Sliding window Kalman filter: A filter with a window size of 10 is used to process the S2.5 input data to balance real-time performance and smoothness (applicable to data with a 500ms period). S3.2. Handling Missing Values: If there are ≤3 consecutive missing points, use forward filling; if there are more than 3 missing points, use linear interpolation to fill them in. S3.3. Dynamic threshold calculation: Every 30 minutes, the mean μ and standard deviation σ are recalculated based on historical data, and the 3σ interval (μ±3σ) is updated. S3.4. Anomaly Classification Marking: Data exceeding 3σ is marked as "urgent anomaly" (e.g., temperature > 190℃), and data approaching the threshold (e.g., 185-190℃) is marked as "warning". S3.5. Additional Data Quality Labels: The output field now includes "Data Quality," with values ​​such as "Qualified," "Forward Fill," "Interpolation," "Urgent Exception," and "Warning." After cleaning, the data is sent to S4 for feature extraction; when marked as "urgent anomaly", an S8 real-time alarm is automatically triggered; when S5 detects a decrease in model confidence, it will request S3 to review the data quality for a specific period. S4. Slurry characteristic feature extraction based on edge computing: S4.1. Calculate dynamic characteristics: For the data stream output by S3.5, calculate the temperature change rate (ΔT / Δt) and pressure fluctuation variance (σ) within a 1-second window. 2 _p); S4.2. Extracting static features: Take the average slurry concentration, pH value, and motor load rate (current / rated current × 100%) within a 5-second window. S4.3. Generate feature dataset: Combining dynamic and static features, the output is: {"time":"2024-03-10T10:00:01","dT_dt":4.0,"σ 2 _p":0.02,"conc":3.2,"ph":10.5,"load":65}; The feature dataset is directly input into the S5-driven digital twin model; when the S5 model detects the pressure fluctuation variance σ 2 When _p > 0.1, S4.1 is automatically required to adjust the size of the dynamic feature window; S5. Real-time status modeling of the pulping process driven by digital twin: S5.1. Hybrid Model Loading: Based on the pulping chemical reaction mechanism model, an LSTM neural network is superimposed to learn nonlinear features; S5.2. Real-time model synchronization: The digital twin state is updated every second using the feature data output by S4.3; S5.3. Prediction and Error Monitoring: Output the concentration prediction value for the next 5 minutes. If the error of key parameters is >2%, trigger online correction. S5.4. Generate a status report: Output: {"time":"2024-03-10T10:00:01","sim_conc":3.18,"sim_temp":185.2,"pred_conc_5min":3.3,"warning":"pH deviates from expected value by mechanism +0.3"}; The prediction results of S5.3 serve as prior knowledge for the optimized control of S6; when the error exceeds the limit, a "data quality verification" command is sent to S3; feedback data from S7 is received to correct the model time constant. S6. Intelligent optimization control with multi-objective collaboration: S6.1. Target Priority Setting: The primary objective is a slurry concentration deviation of <±0.1% and a viscosity of 50-80 mPa·s; the secondary objective is a steam consumption of <1.2 t / ton of slurry. S6.2. Model Predictive Control: Every 2 seconds, based on the twin state of S5.4, the control sequence for the next 30 seconds is continuously optimized. S6.3. Generate control commands: Output: {"device_id":"V-01","action":"open","value":75,"reason":"Increase the temperature to 188℃ to meet the reaction rate","timestamp":"2024-03-10T10:00:02"}; The instruction is issued to S7 for execution; when S8 switches the pulp type, the pH target range in S6.1 is automatically updated; S7. Actuator Feedback and Closed-Loop Verification: S7.1. Real-time feedback collection: The actual execution value is obtained within ≤500ms using equipment status sensors (such as valve opening encoders); S7.2. Execution effect verification: Calculate the absolute deviation between the set value and the actual value (required to be <5%), and the dynamic response time (required to be <10s). S7.3 Output Verification Report: generate: {"device_id":"V-01","set_value":75,"actual_value":73,"delay":300,"status":"qualified"}; S7.4. Model calibration: If the verification fails, adjust the weights or mechanism parameters of the S5 LSTM model; The verification report is input in reverse to S5 to correct the digital twin; if the verification fails three times in a row, a "device maintenance reminder" is sent to S8. S8. Full lifecycle data storage, visualization, and remote operation and maintenance: S8.1. Tiered Storage Architecture: The time-series database (InfluxDB) stores 30 days of real-time data, while the relational database (MySQL) permanently stores process parameters and alarm records. S8.2. Dynamic Visual Kanban: Display temperature / pressure trend curves, concentration / energy consumption indicators, and equipment health status (red / yellow / green lights); S8.3. Remote Operation and Maintenance Interface: It supports adjusting device parameters in S1, abnormal thresholds in S3, and control targets in S6 via Web / APP. Historical data is used for S5 model training; long-term running data is used to optimize the 3σ range of S3; remote commands are issued through the S2 communication link.

[0028] Through the above steps, this invention effectively filters out noise interference and accurately captures early signs of anomalies by using edge-layer intelligent filtering and adaptive threshold cleaning technology. The digital twin model integrates physical mechanisms and deep learning algorithms to perform second-level dynamic simulation and prediction of the cooking reaction process, providing a priori decision-making basis for optimized control. The multi-objective collaborative control algorithm incorporates multiple dimensions such as quality, energy consumption, and equipment loss into a unified optimization framework, significantly reducing steam consumption while ensuring concentration accuracy of ±0.1%. The full life-cycle data engine connects the data value chain from real-time control to remote operation and maintenance, improving the timeliness of equipment fault diagnosis and the efficiency of process switching, and achieving the overall goal of synergistic efficiency improvement in reducing energy consumption and quality fluctuations.

[0029] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their likenesses.

Claims

1. A method for real-time data acquisition and transmission in an intelligent pulping station, characterized by, Comprise the following steps: S1. Multi-protocol compatible device access and parameter initialization, establish unified communication interface of all devices of pulp station, complete initial parameter configuration of device; S2. Multi-source heterogeneous data real-time acquisition and preliminary aggregation, synchronously collect multi-device data through edge gateway, eliminate protocol difference, generate standardized data stream; S3. Real-time data cleaning and abnormal marking based on dynamic threshold, filter noise, fill in missing values, and mark abnormal data; S4. Feature extraction of pulp characteristics based on edge computing, extract key features reflecting the state of pulping from cleaned data, and reduce data dimension; S5. Real-time state modeling of pulping process driven by digital twin, build a virtual mirror of the physical pulping station, drive the simulation running state through real-time data, and predict future trends; S6. Multi-objective collaborative intelligent optimization control, dynamically adjust the steam valve, dosing pump device parameters based on digital twin prediction and process requirements, and realize multi-objective balance of "quality-energy efficiency"; S7. Actuator feedback and closed loop verification, verify the actual execution effect of the control instruction, correct the model error, and form a "collection-control-feedback" closed loop; S8. Full life cycle data storage, visualization and remote operation and maintenance, realize efficient storage, intuitive presentation and remote monitoring of data, and support process optimization and fault diagnosis.

2. The method for real-time data acquisition and transmission in an intelligent pulping station according to claim 1, characterized in that, The implementation process of step S1 is: S1.

1. Automatically identify device type: the system automatically scans the connected devices, and distinguishes temperature sensors, pressure transmitters, pH meters, online concentration meters, motor frequency converters, and steam valve device types; S1.

2. Load protocol driver: match the communication protocol library according to the device type, support ModbusTCP, MQTT, OPCUA and custom binary protocol, and automatically load the corresponding driver module; S1.

3. Set the sampling period: set the sampling frequency of the core parameters to 500ms / second, the auxiliary parameters to 1s / second, and the device state to 200ms / second; S1.

4. Generate a unified data template: all data outputs are in JSON format, and the fields include "device ID", "timestamp", "original value", "unit", and "protocol type".

3. The method for real-time data acquisition and transmission in an intelligent pulping station according to claim 2, characterized in that, The implementation process of step S2 is: S2.

1. Concurrent data acquisition: the edge gateway establishes a maximum of 200 concurrent connections to poll each device at the sampling period set in S1.3; S2.

2. Protocol conversion processing: ASCII code analysis is performed on custom protocol devices; ModbusTCP devices read register values through function code 0x03 and convert them into engineering values; S2.

3. Data buffering and fault tolerance: a 1GB local buffer area is enabled to temporarily store data during network interruption, and the data is preferentially transmitted after recovery; S2.

4. Aggregation of same type data: merge the same device data according to a 1s time window, and shield the influence of faulty nodes; S2.

5. Output standardized data stream: generate fields including "device ID", "timestamp", "value", "unit", and "protocol type"; The output of S2.5 is directly used as the input of S3; when a new device is added in S1, S2.2 automatically calls the new protocol conversion rule.

4. The method for real-time data acquisition and transmission in an intelligent pulping station according to claim 3, characterized in that, The implementation process of step S3 is: S3.

1. Sliding window Kalman filter: filter with window size 10 is used to process S2.5 input data, balancing real-time performance and smoothness; S3.

2. Missing value processing: if consecutive missing points ≤ 3, use forward filling; if missing > 3, use linear interpolation to fill in; S3.

3. Dynamic threshold calculation: every 30 minutes, recalculate mean μ and standard deviation σ based on historical data, update 3σ interval; S3.

4. Abnormal classification labeling: data exceeding 3σ is labeled as "emergency exception", data close to threshold is labeled as "warning"; S3.

5. Additional data quality label: output field adds "data quality", values include "qualified", "forward filling", "interpolation", "emergency exception", "warning".

5. The method for real-time data acquisition and transmission in an intelligent pulping station according to claim 4, characterized in that, The implementation process of the step S4 is: S4.

1. Calculate dynamic features: calculate temperature change rate and pressure fluctuation variance within 1s window for the data stream output by S3.5; S4.

2. Extract static features: Take 5s average window of slurry concentration, pH value, motor load rate; S4.

3. Generate feature dataset: combine dynamic and static features output.

6. The method for real-time data acquisition and transmission in an intelligent pulping station according to claim 5, characterized in that, The implementation process of the step S5 is: S5.

1. Mixed model loading: use the mechanism model of pulping chemical reaction as the basis, superimpose LSTM neural network to learn nonlinear features; S5.

2. Real-time model synchronization: update the digital twin state every 1 second with the feature data output by S4.3; S5.

3. Prediction and error monitoring: output 5-minute concentration prediction value, if key parameter error > 2%, trigger online correction; S5.

4. Generate state report.

7. The method for real-time data acquisition and transmission in an intelligent pulping station according to claim 6, characterized in that, The implementation process of the step S6 is: S6.

1. Target priority setting: the main target is slurry concentration deviation < ±0.1% and viscosity 50-80 mPa·s, the secondary target is steam specific consumption < 1.2 t / ton of pulp; S6.

2. Model predictive control: based on the twin state of S5.4, roll optimization of control sequence for future 30 seconds every 2 seconds; S6.

3. Generate control instructions.

8. The method for real-time data acquisition and transmission in an intelligent pulping station according to claim 7, characterized in that, The implementation process of the step S7 is: S7.

1. Real-time feedback collection: actual execution value is obtained within ≤500ms through device state sensors; S7.

2. Execution effect verification: calculate the absolute deviation of set value and actual value, dynamic response time; S7.

3. Output verification report; S7.

4. Model correction: if verification fails, adjust the LSTM model weight or mechanism parameters of S5.

9. The method for real-time data acquisition and transmission in an intelligent pulping station according to claim 8, characterized in that, The implementation process of the step S8 is: S8.

1. Hierarchical storage architecture: time series database stores 30 days of real-time data, relational database permanently stores process parameters and alarm records; S8.

2. Dynamic visualization dashboard: display temperature / pressure trend curve, concentration / energy consumption index, device health status; S8.

3. Remote operation interface: support Web / APP adjustment of device parameters of S1, abnormal threshold of S3, control target of S6.

Images