A method and system for managing energy in papermaking

Abstract

This invention discloses a papermaking energy management method and system. The method includes: real-time acquisition of electricity, steam, and water consumption data, and construction of a multi-dimensional energy data set containing parameters such as flow rate, pressure, and temperature; based on the multi-dimensional energy data set, using a dynamic pattern recognition algorithm to analyze the energy consumption characteristics of each production stage and generate energy consumption characteristics; based on the energy consumption characteristics, combined with real-time production plans and equipment operating status, generating energy-saving control strategies for specific machines and process sections through intelligent optimization algorithms; executing the energy-saving control strategies and continuously monitoring energy consumption data; automatically triggering an early warning mechanism when actual energy consumption deviates from a preset threshold, and simultaneously initiating corresponding energy efficiency correction measures according to the early warning level, thereby achieving refined energy management in the papermaking production process. Using this invention, closed-loop management from energy data acquisition to optimization strategy execution can be achieved, improving the energy utilization efficiency of papermaking production.

Description

Technical Field

[0001] This invention belongs to the field of papermaking technology, and in particular to a papermaking energy management method and system. Background Technology

[0002] The paper industry is a typical energy-intensive industry, with its production process encompassing multiple energy-intensive stages such as pulping, papermaking, and drying. It generally suffers from low energy efficiency and inefficient management. Currently, most paper companies still rely on manual experience and static data for energy management, making it difficult to capture the complex energy consumption dynamics and abnormal fluctuations in each process stage in real time. While some existing automated systems can collect and monitor some data, they lack deep integration and intelligent analysis of multi-source energy data, and cannot generate accurate and forward-looking energy-saving control strategies based on real-time production plans and equipment status. This results in delayed energy dispatch response, and abnormal energy consumption cannot be corrected in a timely manner, hindering further improvement in enterprise energy efficiency and the realization of refined management. Summary of the Invention

[0003] The purpose of this invention is to provide a papermaking energy management method and system to address the shortcomings of existing technologies, enabling closed-loop management from energy data acquisition to optimization strategy execution, thereby improving the energy utilization efficiency of papermaking production.

[0004] One embodiment of this application provides a papermaking energy management method, the method comprising: By deploying a multi-source sensor network in the pulping, papermaking and drying processes, real-time data on the consumption of electricity, steam and water resources are collected, and a multi-dimensional energy data set including flow rate, pressure and temperature parameters is constructed. Based on the aforementioned multidimensional energy data set, a dynamic pattern recognition algorithm is used to analyze the energy consumption characteristics of each production stage, identify energy consumption patterns and abnormal fluctuation patterns, and generate energy consumption characteristics. Based on the energy consumption characteristics, combined with real-time production plans and equipment operating status, an intelligent optimization algorithm is used to generate energy-saving control strategies for specific machines and process sections. The strategies include load adjustment schemes and optimized settings of operating parameters. The energy-saving control strategy is implemented and energy consumption data is continuously monitored. When the actual energy consumption deviates from the preset threshold, an early warning mechanism is automatically triggered. At the same time, corresponding energy efficiency correction measures are initiated according to the early warning level to achieve refined energy management in the paper production process.

[0005] Optionally, the method involves using a multi-source sensor network deployed in the pulping, papermaking, and drying processes to collect real-time data on electricity, steam, and water consumption, and constructing a multi-dimensional energy data set including flow rate, pressure, and temperature parameters, including: Smart meters, steam flow meters, and water flow meters are deployed on key equipment in the pulping process (refining mill), the papermaking process (papermaking machine), and the drying process (drying cylinder), while pressure sensors and temperature sensors are also installed to generate a multi-source sensor network configuration scheme. Sensor data, including three-phase current and voltage of power parameters, mass flow rate of steam, volumetric flow rate of water resources, and pipeline pressure and temperature readings, are acquired in real time through industrial Ethernet and wireless communication protocols to generate raw sensor data streams. The raw sensor data stream is cleaned and format standardized. A sliding window algorithm is used to eliminate instantaneous noise and compensate for sensor drift error, generating a calibrated time-series data sequence. The calibrated time-series data sequences are aligned dimensionally according to production stages and equipment numbers, and the flow, pressure, and temperature parameters of electricity, steam, and water are integrated to construct a time-synchronized multidimensional energy data set.

[0006] Optionally, the step of analyzing the energy consumption characteristics of each production stage based on the multidimensional energy data set using a dynamic pattern recognition algorithm, identifying energy consumption patterns and abnormal fluctuation patterns, and generating energy consumption characteristics includes: Energy consumption time-series data for each stage of pulping, papermaking and drying are extracted from a multidimensional energy dataset. Electricity consumption sequence, steam consumption sequence and water consumption sequence are constructed with time as the axis, and a stage-specific energy consumption data matrix is ​​generated. The dynamic time warping algorithm is applied to perform pattern matching on the energy consumption data matrix of each stage, identify typical energy consumption curves under different production loads, and generate a benchmark energy consumption pattern library. Based on the benchmark energy consumption pattern library, the density clustering algorithm is used to detect deviation patterns in real-time energy consumption data, calculate the similarity distance between each data point and the benchmark pattern, and generate abnormal fluctuation identification results. By combining the results of abnormal fluctuation identification with production process parameters, the correlation between abnormal energy consumption and equipment operating status is analyzed, and key feature indicators are extracted to form an energy consumption feature descriptor.

[0007] Optionally, based on the energy consumption characteristics and combined with real-time production plans and equipment operating status, an energy-saving control strategy for specific machines and process sections is generated through an intelligent optimization algorithm. The strategy includes load adjustment schemes and optimized operating parameter settings, including: Analyze the abnormal patterns and energy efficiency shortcomings in the energy consumption feature descriptors, and combine them with the output requirements of the production planning system and the real-time operating status of the equipment monitoring system to generate a set of optimization target constraints. A multi-objective optimization model is constructed based on the set of optimization objectives and constraints. The model uses energy minimization and output maximization as dual objective functions and employs particle swarm optimization to iteratively solve the model, generating a Pareto optimal solution set. A feasibility assessment is performed on the Pareto optimal solution set, taking into account the maximum load capacity of the equipment, the range of process parameters, and safety production requirements. Feasible solutions that meet the actual working conditions are selected, and a set of candidate control schemes is generated. The candidate control scheme set is transformed into specific executable instructions, including the adjustment value of the pulper motor power, the setting value of the steam valve opening of the drying cylinder, and the frequency control parameters of the water circulation pump, to generate energy-saving control strategies for specific machines and process sections.

[0008] Optionally, the implementation of the energy-saving control strategy and continuous monitoring of energy consumption data, automatically triggering an early warning mechanism when actual energy consumption deviates from a preset threshold, and simultaneously initiating corresponding energy efficiency correction measures according to the early warning level, thereby achieving refined energy management in the paper production process, including: The industrial control system executes equipment control commands in the energy-saving regulation strategy, adjusts operating parameters including motor speed, valve opening and pump station frequency, and generates strategy execution confirmation signals. Real-time collection of energy consumption data after strategy execution, calculation of unit product energy consumption index and comparison with dynamic thresholds set based on historical data, generating energy consumption deviation assessment report; The early warning mechanism is automatically triggered based on the degree of deviation in the energy consumption deviation assessment report. The early warning level is divided into three levels: yellow, orange, and red, according to the percentage of deviation, and a graded early warning instruction is generated. Based on the graded early warning instructions, corresponding energy efficiency correction measures are initiated. During a yellow warning, equipment parameters are automatically fine-tuned; during an orange warning, auxiliary equipment is activated for coordinated adjustment; and during a red warning, production cycle adjustments are triggered and management personnel are notified to intervene, forming a closed-loop refined energy management system.

[0009] Another embodiment of this application provides a papermaking energy management system, the system comprising: The data acquisition module is used to collect real-time data on the consumption of electricity, steam and water resources through a multi-source sensor network deployed in the pulping, papermaking and drying processes, and to construct a multi-dimensional energy data set including flow rate, pressure and temperature parameters. The identification module is used to analyze the energy consumption characteristics of each production link based on the multidimensional energy data set and a dynamic pattern recognition algorithm, identify energy consumption patterns and abnormal fluctuation patterns, and generate energy consumption characteristics. The generation module is used to generate energy-saving control strategies for specific machines and process sections based on the energy consumption characteristics, combined with real-time production plans and equipment operating status, through intelligent optimization algorithms. The strategies include load adjustment schemes and optimized settings of operating parameters. The execution module is used to execute the energy-saving control strategy and continuously monitor energy consumption data. When the actual energy consumption deviates from the preset threshold, the early warning mechanism is automatically triggered. At the same time, the corresponding energy efficiency correction measures are initiated according to the early warning level to achieve refined energy management in the paper production process.

[0010] Another embodiment of this application provides a storage medium storing a computer program, wherein the computer program is configured to execute the method described in any of the preceding claims when running.

[0011] Another embodiment of this application provides an electronic device including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the method described in any of the preceding claims.

[0012] Compared with existing technologies, the present invention provides a papermaking energy management method that collects real-time data on the consumption of electricity, steam, and water resources, and constructs a multi-dimensional energy data set including parameters such as flow rate, pressure, and temperature. Based on the multi-dimensional energy data set, a dynamic pattern recognition algorithm is used to analyze the energy consumption characteristics of each production stage and generate energy consumption characteristics. According to the energy consumption characteristics, combined with real-time production plans and equipment operating status, an intelligent optimization algorithm is used to generate energy-saving control strategies for specific machines and process sections. The energy-saving control strategies are executed and energy consumption data is continuously monitored. When the actual energy consumption deviates from the preset threshold, an early warning mechanism is automatically triggered. At the same time, corresponding energy efficiency correction measures are initiated according to the early warning level, realizing refined energy management of the papermaking production process. This enables closed-loop management from energy data collection to optimization strategy execution, thereby improving the energy utilization efficiency of papermaking production. Attached Figure Description

[0013] Figure 1 A hardware structure block diagram of a computer terminal for a papermaking energy management method provided in an embodiment of the present invention; Figure 2 A schematic flowchart of a papermaking energy management method provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of a papermaking energy management system provided in an embodiment of the present invention. Detailed Implementation

[0014] The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0015] The present invention first provides a papermaking energy management method, which can be applied to electronic devices, such as computer terminals, specifically ordinary computers.

[0016] The following detailed explanation uses a computer terminal as an example. Figure 1 This is a hardware structure block diagram of a computer terminal for a papermaking energy management method provided in an embodiment of the present invention. Figure 1 As shown, the computer device includes a processor, memory, and network interface connected via a system bus, wherein the memory may include non-volatile storage media and internal memory.

[0017] Non-volatile storage media can store operating systems and computer programs. These computer programs include program instructions that, when executed, cause the processor to perform any papermaking energy management method.

[0018] The processor provides computing and control capabilities, supporting the operation of the entire computer device.

[0019] Internal memory provides an environment for the execution of computer programs in non-volatile storage media. When these computer programs are executed by a processor, the processor can execute any papermaking energy management method.

[0020] This network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 1 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0021] It should be understood that the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0022] See Figure 2 The present invention provides a papermaking energy management method, which may include the following steps: S201 uses a multi-source sensor network deployed in the pulping, papermaking and drying processes to collect real-time data on the consumption of electricity, steam and water resources, and constructs a multi-dimensional energy data set including flow rate, pressure and temperature parameters. Specifically, smart meters, steam flow meters, and water flow meters can be deployed on key equipment in the pulping process (refining mill), the papermaking process (papermaking machine), and the drying process (drying cylinder), while pressure sensors and temperature sensors are also installed to generate a multi-source sensor network configuration scheme. This step forms the foundation for building a papermaking energy data acquisition system. Its core is to achieve comprehensive coverage of energy consumption and process parameters across the three core stages of pulping, papermaking, and drying through targeted sensor deployment. The generated configuration plan must clearly define sensor types, installation locations, measurement parameters, and communication compatibility requirements to ensure the accuracy and completeness of data acquisition. The specific implementation method is as follows: First, clarify the monitoring requirements and sensor selection logic for key equipment in each production stage. The pulping mill is a core power-consuming device, requiring close monitoring of power consumption and operating load parameters. Therefore, deploy a three-phase smart meter with a measurement range adapted to the pulping mill motor power (e.g., 50-500kW), a three-phase current range of 0-600A, a voltage range of 380V±10%, and a measurement accuracy of 0.5 class to ensure accurate power data measurement. Simultaneously, the pulping process consumes a large amount of process water; therefore, deploy an electromagnetic water flow meter near the equipment end of the pulping mill's inlet pipe, with a measurement range of 0-100m³. 3 The flow rate is measured at 0.3 kg / h, with an accuracy of 0.3%. The installation location should avoid areas prone to turbulence, such as pipe bends and valves, and should be at least 5 times the pipe diameter away from bends to ensure the stability of the flow measurement. The paper machine in the papermaking process involves the coordinated consumption of electricity, steam, and water resources. A smart meter of the same specifications as the refiner is installed at the main drive motor. A vortex steam flow meter is installed on the steam supply pipeline (before connecting to the steam inlet of the drying cylinder), with a measurement range of 0-5000 kg / h, an accuracy of 0.5, a steam temperature range of 150-250℃, and a pressure range of 0.3-1.0 MPa. Simultaneously, water flow meters with a range of 0-200 m are installed on the process water inlet and outlet pipelines of the paper machine's wire section and press section. 3 / h; The drying cylinder in the drying process is the core equipment for steam consumption. In addition to deploying steam flow meters in the steam inlet and outlet pipes, sensors also need to be deployed on the surface of the drying cylinder and key nodes of the steam pipe to ensure monitoring of steam utilization efficiency.

[0023] The deployment of pressure and temperature sensors needs to cover key nodes in energy transmission and equipment operation: In the pulping process, pressure sensors are deployed on the refiner inlet pipe and steam heating pipe (if any), with ranges of 0-1.6MPa and 0.3-1.0MPa, respectively, and temperature sensors with ranges of 0-100℃ and 100-250℃. In the papermaking process, pressure sensors are deployed on the main steam pipe and process water return pipe of the papermaking machine, and temperature sensors simultaneously monitor the steam temperature and process water temperature. In the drying process, pressure and temperature sensors are deployed on the steam inlet and outlet pipes of the drying cylinder, respectively. The inlet pressure sensor has a range of 0.3-1.0MPa, and the outlet pressure sensor has a range of 0-0.5MPa. The temperature sensor has a range of 150-250℃ at the inlet and 100-200℃ at the outlet. At the same time, a contact temperature sensor is deployed on the drying cylinder body to monitor the uniformity of the cylinder surface temperature, with a range of 80-200℃.

[0024] Multi-source sensor network configuration schemes need to be documented in a standardized manner, including core information such as unique sensor number, deployment process, device ownership, installation location, measurement parameters, measurement range, accuracy class, communication protocol, and power supply method. For example, the smart meter numbered SP-001 is deployed in the main drive motor control cabinet of the No. 1 refiner in the pulping stage. Its measurement parameters are three-phase current (I_a, I_b, I_c), three-phase voltage (U_ab, U_bc, U_ca), and total active power (P_total). Its range is 0-600A for current and 380V±10% for voltage, with an accuracy of 0.5 class. It uses the Profinet communication protocol and is powered via industrial Ethernet. The steam flow meter numbered SS-003 is deployed in the steam inlet pipe of the No. 2 drying cylinder in the drying stage (3 meters from the valve). Its measurement parameter is steam mass flow rate (Q_steam), with a range of 0-3000kg / h, an accuracy of 0.5 class, and is suitable for temperatures of 150-250℃ and pressures of 0.3-1.0MPa. It uses the Modbus RTU communication protocol and is powered by 24V. A DC temperature sensor, designated ST-005, is deployed in the main steam pipeline of the paper machine. It measures steam temperature (T_steam), with a range of 100-250℃ and an accuracy of ±0.3℃. The communication protocol is LoRa, and it is wirelessly powered. The configuration plan also needs to specify the sensor's installation specifications, such as horizontal installation of the flow meter, vertically upward pressure of the pressure sensor's tap, and an insertion depth of the temperature sensor of at least 1 / 3 of the pipe's inner diameter, to ensure measurement accuracy.

[0025] Sensor data, including three-phase current and voltage of power parameters, mass flow rate of steam, volumetric flow rate of water resources, and pipeline pressure and temperature readings, are acquired in real time through industrial Ethernet and wireless communication protocols to generate raw sensor data streams. This step is crucial for achieving real-time energy data acquisition. It requires selecting communication technologies suitable for the industrial environment to ensure data transmission stability, real-time performance, and interference resistance. Simultaneously, the specific content and format of the acquired data must be clearly defined to lay the foundation for subsequent processing. The specific implementation method is as follows: The choice of communication protocol must be based on both sensor deployment location and data transmission requirements: Profinet is the preferred industrial Ethernet protocol, suitable for areas with wired network coverage such as pulp and paper mills. It boasts a transmission rate of up to 100Mbps, communication latency ≤10ms, supports parallel communication from a large number of sensors, and can meet the transmission needs of high-frequency data such as power parameters and steam flow. LoRa is the preferred wireless communication protocol, suitable for areas where wired deployment is inconvenient, such as around drying cylinders in drying workshops and outdoor pipelines. It offers a communication distance of up to 3km, strong anti-interference capabilities, supports long-term operation of low-power sensors, has a transmission latency ≤50ms, and is suitable for acquiring medium- and low-frequency data such as pressure and temperature. All sensors must support the corresponding communication protocol and connect to a unified data acquisition platform via a gateway. The gateway is deployed in the control room of each workshop and is responsible for data aggregation and protocol conversion.

[0026] The data acquisition frequency is dynamically set according to the parameter type to balance real-time performance and system load: power parameters (three-phase current, voltage, active power) directly reflect the equipment's operating load, and the acquisition frequency is set to 1Hz (once every 1 second) to ensure the capture of load fluctuation details; steam mass flow rate and water volume flow rate are directly related to energy consumption, and the acquisition frequency is set to 0.2Hz (once every 5 seconds); pipeline pressure and temperature parameters change relatively smoothly, and the acquisition frequency is set to 0.1Hz (once every 10 seconds). During the acquisition process, the data acquisition platform sends data requests to each sensor through polling. After the sensor responds, it returns a data packet containing complete information, which the platform automatically receives and stores. If no response is received from a sensor for three consecutive times, it is marked as a communication anomaly and a fault timestamp is recorded.

[0027] The collected raw data must contain complete identification and parameter information. The format of each data record is "sensor number - collection timestamp - parameter type - parameter value - unit - data status", where the collection timestamp is accurate to milliseconds (format is YYYY-MM-DD HH:MM:SS.sss), and the data status is divided into "normal (0), abnormal (1), missing (2)". The specific data content and parameter ranges collected are as follows: In power parameters, the three-phase current I_a, I_b, I_c range is 0-600A, and the value is retained to 1 decimal place; the voltage U_ab, U_bc, U_ca range is 350-410V, and the value is retained to 1 decimal place; the total active power P_total range is 0-500kW, and the value is retained to 1 decimal place; the steam mass flow rate Q_steam range is 0-5000kg / h, and the value is retained to the integer, with the unit being kg / h; the water resource volume flow rate Q_water range is 0-200m 3 / h, rounded to one decimal place, unit: m 3 / h; Pipeline pressure P range 0-1.6MPa, retain 2 decimal places, unit MPa; Temperature T range 50-250℃, retain 1 decimal place, unit ℃.

[0028] The generated raw sensor data stream is a continuous time-series data sequence. For example, a power data record is "SP-001-2025-07-01 08:30:00.123-I_a-320.5-A-0", indicating that the smart meter of the #1 refiner collected a phase A current of 320.5A at that time point, which is normal. A steam flow data record is "SS-003-2025-07-01 08:30:05.456-Q_steam-1850-kg / h-0", indicating that the mass flow rate of the steam inlet of the #2 drying cylinder is 1850kg / h, which is normal. A temperature data record is "ST-005-2025-07-01 08:30:10.789-T_steam-198.2-℃-0", indicating that the temperature of the main steam pipeline of the paper machine is 198.2℃, which is normal. The raw data stream needs to be written to the local data acquisition server in real time, using a circular buffer to store the raw data of the most recent 72 hours, while a backup mechanism is used to prevent data loss.

[0029] The raw sensor data stream is cleaned and format standardized. A sliding window algorithm is used to eliminate instantaneous noise and compensate for sensor drift error, generating a calibrated time-series data sequence. This step is crucial for improving data quality. By targeting and eliminating interfering factors in the raw data, standardizing the data format, and ensuring data reliability and consistency, it provides high-quality input for subsequent energy consumption analysis. The specific implementation method is as follows: Data cleaning mainly targets missing values, outliers, and duplicate values ​​in the original data stream. Missing values ​​are handled using the "nearest neighbor interpolation method". If a data point is marked as "missing (2)", the mean of the three normal data points before and after the missing data is used as the supplement. For example, the flow rate data of a certain sensor at 08:30:05 is missing. The normal data points before and after are 1845 kg / h, 1850 kg / h, 1852 kg / h, 1848 kg / h, 1855 kg / h, and 1853 kg / h, respectively. The interpolation result is (1845+1850+1852+1848+1855+1853) / 6≈1850.5 kg / h, and the integer is 1851 kg / h. Outlier values ​​are handled using the "3σ criterion". First, the mean μ and standard deviation σ of the most recent 100 normal data points of a certain parameter of a certain sensor are calculated. If the value of a certain data point exceeds the range of [μ-3σ, μ+3σ], it is judged as an outlier. For example, the steam flow rate of the most recent 100 normal data points is... The mean of the data is μ=1800kg / h, and the standard deviation is σ=50kg / h. A certain data is 2000kg / h, which is outside the range of [1650, 1950] and is judged as abnormal. Duplicate values ​​are handled by deduplication using timestamps. If multiple data exist for the same parameter from the same sensor at the same timestamp (error ≤1ms), only the first normal data is retained, and the rest are marked as duplicates and deleted.

[0030] The sliding window algorithm is used to eliminate transient noise. The window size is dynamically set according to the data acquisition frequency: the power parameter acquisition frequency is 1Hz, and the window size is set to 10 data points (corresponding to 10 seconds); the steam and water flow acquisition frequency is 0.2Hz, and the window size is set to 5 data points (corresponding to 25 seconds); the pressure and temperature acquisition frequency is 0.1Hz, and the window size is set to 3 data points (corresponding to 30 seconds). The algorithm replaces the original data by calculating the median of the data within the window. The median is more resistant to extreme value interference than the mean. For example, if the 10 consecutive data points of the A-phase current of the pulper are 320.5A, 321.2A, 319.8A, 325.6A (instantaneous noise), 320.1A, 319.9A, 321.5A, 320.8A, 319.7A, and 320.3A, the median of the window is (320.3+320.5) / 2=320.4A. This value is used to replace the instantaneous noise of 325.6A in the original data to eliminate interference.

[0031] Sensor drift error compensation addresses the systematic deviations that occur in sensors after long-term operation by employing a combination of "periodic calibration + real-time compensation". Regular calibration is performed every 3 months. The sensor is calibrated using standard instruments to establish a drift compensation curve. For example, if the temperature sensor is found to have linear drift in the range of 80-200℃ after calibration, the compensation formula is T_calibration = T_original + 0.005 × (T_original - 100). If the original temperature is 150℃, the calibrated temperature will be 150 + 0.005 × (150 - 100) = 152.5℃. Real-time compensation is achieved by comparing data with redundant sensors in the same location. If two temperature sensors are deployed on a pipeline, and the difference between their data exceeds 0.5℃ for 5 consecutive times, the data from the sensor with higher accuracy is used as the benchmark to correct the deviation sensor in real time. For example, if sensor A's data is 198.2℃ and sensor B's data (accuracy ±0.2℃) is 198.7℃, with a difference of 0.5℃, then sensor A's data will be corrected to 198.7℃.

[0032] The format standardization process unifies the format of all cleaned and calibrated data analysis. Specific requirements are: timestamps must be in the format "YYYY-MM-DD HH:MM:SS.sss"; parameter names must use standardized abbreviations (e.g., I_a, U_ab, Q_steam, Q_water, P, T); numerical values ​​must be retained to a uniform number of decimal places (current, voltage, and power retain 1 decimal place; flow rate retains 1 decimal place; pressure retains 2 decimal places; temperature retains 1 decimal place); and units must use international standard symbols (A, V, kW, kg / h, m³). 3 / h, MPa, ℃). The final calibrated time-series data sequence has a uniform format and reliable data for each record. For example, the corrected power data is "SP-001-2025-07-01 08:30:00.123-I_a-320.4-A-0", the steam flow data is "SS-003-2025-07-01 08:30:05.456-Q_steam-1851-kg / h-0", and the temperature data is "ST-005-2025-07-01 08:30:10.789-T_steam-198.7-℃-0", laying the foundation for subsequent dimensional alignment.

[0033] The calibrated time-series data sequences are aligned dimensionally according to production stages and equipment numbers, and the flow, pressure, and temperature parameters of electricity, steam, and water are integrated to construct a time-synchronized multidimensional energy data set.

[0034] This step is the core of achieving multi-source data fusion. Through dimensional alignment and time synchronization, scattered single-parameter data are integrated into a unified dataset containing multi-dimensional information, which facilitates subsequent energy consumption characteristic analysis. The specific implementation method is as follows: The core of dimensional alignment is to group and categorize the calibrated time-series data according to "production stage - equipment number" to clarify the data's attribution. First, three major production stages are defined: pulping (code P), papermaking (code M), and drying (code D). Each stage is further subdivided by equipment number (e.g., pulping stage: refiner #1, refiner #2; papermaking stage: paper machine #1, paper machine #2; drying stage: dryer group #1, dryer group #2). Each equipment group corresponds to a complete set of energy and process parameters, including electrical parameters (three-phase current, three-phase voltage, total active power), steam parameters (mass flow rate, inlet pressure, inlet temperature, outlet pressure, outlet temperature), and water resource parameters (influent volumetric flow rate, return volumetric flow rate, influent pressure, influent temperature, return pressure, return temperature), ensuring full coverage of energy consumption and process status parameters for each piece of equipment.

[0035] Time synchronization is crucial for dimensional alignment. Network Time Protocol (NTP) is used to synchronize the clocks of all sensors and the data acquisition platform, with synchronization accuracy controlled within ±10ms, ensuring accurate correlation of parameter data at the same time point. Using the system clock of the data acquisition platform as a reference, the time-series data of each device group is timestamped. A "time slice" is set to 10 seconds, meaning that every 10 seconds is a time node, and all parameter data within 5ms before and after that node is extracted. If a parameter has no collected data within a time slice, the most recently calibrated data is used to fill it, ensuring no missing data for each device group within each time slice. For example, if the time slice is set to 08:30:00.000-08:30:10.000, all parameter data for mill #1 within this interval are extracted. If there are 6 electrical parameters (every 1 second), 1 steam flow parameter (every 5 seconds), and 1 pressure and temperature parameter (every 10 seconds), then 08:30:10.000 is used as the timestamp to integrate all parameter data.

[0036] The construction of the multidimensional energy data set adopts a four-level structure of "timestamp-process-equipment-parameter set". Each timestamp corresponds to a complete multidimensional data record, which includes six core parts: timestamp, process code, equipment number, power parameter subset, steam parameter subset, and water resource parameter subset. Each parameter subset contains all specific parameters and values ​​under that category. For example, the multidimensional data record for refiner #1 (stage P, equipment 01) at timestamp 2025-07-01 08:30:10.000 is: "Timestamp: 2025-07-01 08:30:10.000; Stage: P; Equipment: 01; Power parameters: {I_a:320.4A, I_b:318.7A, I_c:321.2A, U_ab:380.5V, U_bc:381.2V, U_ca:379.8V, P_total:185.6kW}; Steam parameters: {Q_steam:0.0kg / h (no steam consumption in the pulping stage), P_in:0.0MPa, T_in:0.0℃, P_out:0.0MPa, T_out:0.0℃};Water resource parameters: {Q_water_in:45.2m 3 / h, Q_water_out:42.8m 3 / h, P_water_in:0.85MPa, T_water_in:35.2℃, P_water_out:0.62MPa, T_water_out:42.5℃}".

[0037] For scenarios involving multiple devices working collaboratively (such as multiple drying cylinders sharing a single steam pipeline in a drying process), a "spreading coefficient" needs to be added to the parameter subset to distribute the common energy consumption according to the equipment load ratio. For example, the No. 2 drying cylinder group contains 3 drying cylinders with a total steam flow of 1851 kg / h. Based on the temperature setting and operating load of each drying cylinder, the spreading coefficients are 0.3, 0.4, and 0.3, respectively. Then, the No. 1 drying cylinder shares a flow of 555.3 kg / h, the No. 2 drying cylinder shares 740.4 kg / h, and the No. 3 drying cylinder shares 555.3 kg / h, ensuring that the energy consumption data of each device is accurate and traceable.

[0038] The resulting multidimensional energy dataset is stored in a time-series database (such as InfluxDB), supporting data querying and filtering by time range, process, and equipment number. The data retention period is one year, meeting the needs of energy consumption pattern analysis and historical data tracing. This dataset ensures the time synchronization of all parameters and clearly defines the dimensional attribution of the data, providing a structured and high-quality data foundation for subsequent dynamic pattern recognition algorithms to analyze energy consumption characteristics.

[0039] S202, Based on the multidimensional energy data set, a dynamic pattern recognition algorithm is used to analyze the energy consumption characteristics of each production link, identify energy consumption patterns and abnormal fluctuation patterns, and generate energy consumption characteristics. Specifically, energy consumption time-series data for each stage of pulping, papermaking, and drying can be extracted from a multi-dimensional energy dataset to construct a time-based sequence of electricity consumption, steam consumption, and water consumption, generating a sub-stage energy consumption data matrix. This step is the foundational data preparation stage for energy consumption characteristic analysis. Its core is to extract key energy consumption time-series data from a structured, multi-dimensional energy dataset, breaking it down by production stage. A standardized matrix is ​​then constructed by aligning the time axis, providing well-structured input data for subsequent pattern recognition. The specific implementation method is as follows: The multidimensional energy data set contains comprehensive information including timestamps, processes, equipment, electricity, steam, and water resources. The extraction process requires filtering data according to a three-tiered logic: "production process - energy consumption type - time". First, the core energy consumption parameters of the three major production processes are clarified: the pulping process is mainly characterized by electricity (energy consumption for refining machine drive) and water resources (process water), with less steam consumption (only in some heating scenarios); the papermaking process requires simultaneous extraction of electricity (paper machine drive, vacuum pump, etc.), steam (wire section dehumidification, press section heating), and water resources (wire section spraying, press section cleaning) consumption data; the drying process is mainly characterized by steam (drying cylinder heating) and electricity (induced draft fan, exhaust fan), with water resource consumption being negligible.

[0040] The timeline for time-series data is uniformly set to "minute level," meaning that raw data is integrated at one data point per minute (average values ​​are taken for raw data collection frequencies higher than minute level, and interpolation is used to supplement those lower). The time span covers a continuous 24-hour production cycle to ensure that the daily energy consumption fluctuation pattern can be captured. For example, when extracting the power consumption data of the No. 1 refiner in the pulping process, the arithmetic mean of the 1Hz data collected every minute is taken, resulting in 24×60=1440 data points, forming the power consumption sequence; the steam consumption sequence is generated by averaging the 0.2Hz data collected every minute, also generating 1440 data points; the water consumption sequence is generated similarly by integrating the net consumption of inlet and outlet flow rates (inlet flow rate - return flow rate) to generate time-series data.

[0041] The energy consumption data matrix for each stage is constructed using a three-dimensional matrix structure, defined as "stage × energy consumption type × time point". The stage dimension includes three categories: pulping (P), papermaking (M), and drying (D); the energy consumption type dimension includes three categories: electricity (E), steam (S), and water (W); and the time point dimension contains 1440 minute-level data points. Each element in the matrix represents the energy consumption value for the corresponding stage and energy consumption type at a specific time point, maintaining the same precision as the calibrated data (one decimal place for electricity, steam, and water). For example, matrix element PE-30 represents the electricity consumption value at the 30th minute of the pulping stage (e.g., 182.5kW), MS-120 represents the steam consumption value at the 120th minute of the papermaking stage (e.g., 1680.3kg / h), and DW-600 represents the water consumption value at the 600th minute of the drying stage (e.g., 0.0m). 3 / h, with no process water consumption in the drying process).

[0042] To ensure the integrity of the matrix, for stages without corresponding energy consumption types (such as water consumption in the drying stage), 0.0 is used as a placeholder. For missing data during equipment failures or downtime, the average of the five nearest time points is used to fill in the missing data, avoiding empty values ​​in the matrix. For example, in the pulping stage, there is no power data at the 150th minute due to a temporary sensor failure. The five data points before and after this point are 178.2kW, 179.5kW, 180.1kW, 181.3kW, 182.0kW, 181.5kW, 180.8kW, 179.9kW, 178.7kW, and 177.9kW, respectively. The average is (178.2+179.5+180.1+181.3+182.0+181.5+180.8+179.9+178.7+177.9) / 10≈179.9kW, which is then filled into the corresponding position in the matrix. The final generated energy consumption data matrix for each stage not only maintains the temporal continuity of energy consumption in each stage, but also achieves dimensional differentiation of different energy consumption types, providing a well-organized dataset for subsequent pattern matching.

[0043] The dynamic time warping algorithm is applied to perform pattern matching on the energy consumption data matrix of each stage, identify typical energy consumption curves under different production loads, and generate a benchmark energy consumption pattern library. This step is the core of energy consumption pattern extraction. The core advantage of the Dynamic Time Warping (DTW) algorithm is its ability to handle time series data matching with different lengths and time offsets, accurately identify common energy consumption characteristics under different production loads, and then build a benchmark pattern library covering common production scenarios. The specific implementation method is as follows: The core logic of the dynamic time warping algorithm is to calculate the minimum cumulative distance between two sequences by stretching or compressing the time axis of time-series data, thereby achieving nonlinear alignment and similarity matching. In energy consumption analysis, it is first necessary to classify common production load levels (according to the production patterns of the paper industry, divided into four load categories: 50%, 70%, 85%, and 100%, covering the entire scenario from low load to full load). Then, for each link and each load level, time-series data for the corresponding period are selected from the energy consumption data matrix (e.g., for the 100% load period, process conditions such as paper machine speed ≥ 95% of design speed and refiner speed ≥ 90% of rated speed must be met), forming multiple sets of energy consumption sequence samples corresponding to the load.

[0044] Taking the power consumption sequence matching of the pulping process at 100% load as an example, three sets of sample sequences for two consecutive hours (120 data points in each set) were selected, and the DTW algorithm was applied for pairwise matching: First, the DTW distance between the first and second sets of sequences was calculated (the cumulative distance threshold was set to 100), and the optimal alignment path was found through dynamic programming. The time axes of the two sequences were synchronized, and common features (such as mean, peak value, and fluctuation range) were extracted. Then, the matched sequences were matched with the third set of samples for a second time to further optimize feature consistency. Finally, the typical power consumption curve of the pulping process at 100% load was determined: mean 180kW, peak value 200kW, fluctuation range ±5kW, valley value 170kW. The sequence as a whole showed a "stable to slightly fluctuating" trend with no obvious abrupt changes.

[0045] Following the same logic, the steam consumption sequences and water consumption sequences for each stage and load level are matched, typical curve features are extracted, and a baseline energy consumption model library is finally generated. The model library adopts a four-level index structure of "stage - load level - energy consumption type - characteristic parameters," with each model containing a complete description of curve features and numerical range: In the 50% load model of the pulping stage, the average power consumption is 90kW, the peak is 100kW, and the fluctuation is ±3kW; the average steam consumption is 800kg / h, and the fluctuation is ±20kg / h; the average water consumption is 25m³ / h. 3 / h, fluctuation ±1m 3 / h; In the papermaking process at 85% load, the average power consumption is 220kW, the peak is 240kW, and the fluctuation is ±8kW; the average steam consumption is 2200kg / h, and the fluctuation is ±50kg / h; the average water consumption is 55m³. 3 / h, fluctuation ±2m 3 / h; In the drying process at 100% load, the average power consumption is 150kW, the peak is 160kW, and the fluctuation is ±4kW; the average steam consumption is 3000kg / h, and the fluctuation is ±30kg / h; the average water consumption is 0.5m3 / h (cleaning water only), fluctuation ±0.1m 3 The pattern library contains 12 core patterns (3 stages × 4 load levels), covering the mainstream operating conditions of papermaking production, providing a clear benchmark for subsequent anomaly identification.

[0046] Based on the benchmark energy consumption pattern library, the density clustering algorithm is used to detect deviation patterns in real-time energy consumption data, calculate the similarity distance between each data point and the benchmark pattern, and generate abnormal fluctuation identification results. This step is the core execution step for abnormal energy consumption detection. The density clustering algorithm (DBSCAN) can automatically identify abnormal clusters based on the spatial density of data points, and combine similarity distance to quantify the degree of deviation, accurately locating energy consumption anomalies. The specific implementation method is as follows: First, the real-time data processing workflow is clearly defined: the minute-level energy consumption data (electricity, steam, and water) collected in real time is categorized by stage, forming an analysis window (containing 30 data points) every 30 minutes, with a window sliding step of 10 minutes to ensure real-time capture of energy consumption fluctuations. For the energy consumption sequence of each analysis window, its corresponding baseline pattern is first determined through pattern matching (e.g., the average power sequence of the real-time pulping stage is 178kW, with fluctuations of ±4kW, matched to the 100% load baseline pattern), and then the similarity distance between each data point within the window and the corresponding time point of the baseline pattern is calculated.

[0047] The similarity distance is calculated using a combination of DTW distance and Euclidean distance: DTW distance measures the deviation of the overall sequence shape (e.g., whether the curve trend is consistent), while Euclidean distance quantifies the absolute deviation of the values ​​(e.g., the difference between real-time values ​​and baseline values). The comprehensive distance D = 0.6 × DTW distance + 0.4 × Euclidean distance (weights are determined based on experimental verification, prioritizing morphological matching accuracy). Distance thresholds are set as follows: for the pulping stage, the comprehensive distance threshold is 50 for power consumption, 80 for steam consumption, and 10 for water consumption; for the papermaking stage, it is 60 for power, 100 for steam, and 15 for water consumption; and for the drying stage, it is 40 for power, 120 for steam, and 5 for water consumption. If the distance exceeds the threshold, it is considered an anomaly for a single data point.

[0048] The parameter configuration of the density clustering algorithm needs to be adapted to the characteristics of the energy consumption data: the neighborhood radius ε is set to 30 (corresponding to a reasonable clustering range of comprehensive distance), and the minimum number of points MinPts is set to 5 (to ensure the stability of clustering and avoid misclassification of isolated points). The algorithm clusters 30 data points within the analysis window, classifying regions with a density higher than MinPts within the ε range as normal clusters, and identifying points that deviate from normal clusters or have a density lower than the threshold as outliers. For example, in a 30-minute analysis window during the pulping process, the power consumption at the 15th minute was 220kW. The distance between this point and the DTW distance of 62 and the Euclidean distance of 58 from the 100% load baseline model was 62. The combined distance D = 0.6 × 62 + 0.4 × 58 = 37.2 + 23.2 = 60.4, which exceeded the threshold of 50. Furthermore, there were only 2 data points within a 30-minute neighborhood of this point, which did not reach MinPts = 5, so it was identified as an outlier. In the same window, there was another data point at the 20th minute with a combined distance of 55, which was also identified as an outlier. The remaining 28 data points formed a normal cluster, and the combined distance between the cluster center and the baseline model was only 12.

[0049] The generated abnormal fluctuation identification results contain structured information: abnormal occurrence time (accurate to the minute), the relevant process, energy consumption type, real-time value, corresponding baseline mode, overall distance, deviation threshold, and abnormality type (single data point abnormality / continuous abnormality cluster). For example, an abnormal record is "Time: 2025-07-02 10:15; Process: Pulping; Energy consumption type: Electricity; Real-time value: 220.0kW; Baseline mode: Pulping - 100% load; Overall distance: 60.4; Threshold: 50; Abnormality type: Single data point abnormality". If three consecutive data points are identified as abnormal (e.g., 10:15, 10:16, 10:17), they are marked as "continuous abnormality cluster", indicating a possible equipment failure or process abnormality.

[0050] By combining the results of abnormal fluctuation identification with production process parameters, the correlation between abnormal energy consumption and equipment operating status is analyzed, and key feature indicators are extracted to form an energy consumption feature descriptor.

[0051] This step is the core of energy consumption feature extraction. By linking production processes and equipment status, it uncovers energy consumption patterns and root causes of anomalies, extracts representative feature indicators, and forms standardized energy consumption feature descriptors. This provides a basis for generating subsequent optimization strategies. The specific implementation method is as follows: First, production process parameters and equipment operating status data are integrated. Process parameters include refiner speed (pulping stage), paper machine speed (papermaking stage), dryer set temperature (drying stage), and pulp concentration (pulping / papermaking stage). Equipment operating status data includes motor current, bearing temperature, valve opening, and vacuum pump pressure. These data are aligned with energy consumption data using timestamps to construct a three-dimensional correlation dataset of "energy consumption-process-equipment". When analyzing the correlation, the Pearson correlation coefficient is used to quantify the linear correlation between energy consumption parameters and process / equipment parameters (range -1 to 1, with the absolute value closer to 1 indicating a stronger correlation). Simultaneously, causal relationships are determined by combining process logic.

[0052] Taking the abnormal power supply in the pulping process as an example, the correlation data shows that during the abnormal period, the speed of the pulper increased from the rated 1480 r / min to 1550 r / min (change in process parameters), and the temperature of the motor bearing increased from 45℃ to 58℃ (change in equipment status). The correlation coefficient between power consumption and the pulper speed is 0.85 (strong positive correlation), and the correlation coefficient with the bearing temperature is 0.7 (moderate positive correlation). Therefore, it is determined that the core reason for the abnormal power supply is that the pulper speed is too high, resulting in increased load and energy consumption. During a certain period in the drying process, the steam consumption was abnormally high (2800 kg / h, compared to 2500 kg / h in the baseline mode). The correlation data shows that the insulation layer of the drying cylinder was not properly sealed (equipment status), with a steam leakage rate of 8%, and the set temperature of the drying cylinder was not adjusted (process parameters were stable). Therefore, it is determined that the abnormality originated from a defect in the equipment insulation.

[0053] The extraction of key feature indicators needs to cover two major categories: "energy consumption pattern characteristics" and "anomaly correlation characteristics," totaling 12 core indicators. Energy consumption pattern characteristics include the 24-hour average, peak, trough, and fluctuation coefficient (fluctuation coefficient = standard deviation / mean × 100%) of each energy consumption type, and the distribution ratio during different time periods (such as the proportion of energy consumption during peak periods). Anomaly correlation characteristics include the frequency of anomalies (number of anomalies per unit time), the duration of anomalies, the proportion of anomaly types (single point anomaly / continuous cluster anomalies), the correlation coefficient with core process parameters, the correlation coefficient with equipment status parameters, the root cause label of anomalies (process adjustment / equipment failure / environmental impact), and the energy consumption optimization potential value (energy saving space calculated based on the baseline model).

[0054] Taking the pulping process as an example, the final energy consumption characteristic descriptor is as follows: "Process: Pulping; Electricity consumption - 24-hour average 185.2kW, peak 220.0kW, valley 170.5kW, fluctuation coefficient 5.8%, peak period (8:00-18:00) accounts for 72%; abnormal frequency 3 times / day, abnormal duration ≤1 minute, abnormal type percentage: single point abnormality 100%, correlation coefficient with pulper speed 0.85, correlation coefficient with motor bearing temperature 0.7, abnormal root cause label: process adjustment (excessive speed), optimization potential value 15.2kW (energy saving space based on baseline model); Steam consumption - 24-hour average 1520.3kg / h, peak 1600.5kg / h, valley 1450.8kg / h, fluctuation coefficient 2.3%, no abnormality, correlation coefficient with pulp heating temperature 0.65; Water consumption - 24-hour average 45.6m..." 3 / h, peak value 50.2m 3 / h, valley value 42.1m 3 / h, fluctuation coefficient 2.1%, no abnormality, correlation coefficient with paper machine speed 0.7; comprehensive characteristics: energy consumption generally shows the pattern of "concentrated peak and stable low peak", power consumption is significantly affected by the speed of the refiner, there are short-term abnormalities caused by process adjustment, and it has an energy saving potential of about 15%.

[0055] This feature descriptor comprehensively integrates the operating patterns, abnormal characteristics, and related factors of various energy consumption types, providing complete feature support for subsequent intelligent optimization algorithms to accurately locate energy-saving directions and generate targeted control strategies.

[0056] S203, based on the energy consumption characteristics and combined with the real-time production plan and equipment operating status, an energy-saving control strategy for specific machines and process sections is generated through an intelligent optimization algorithm. The strategy includes load adjustment schemes and optimized settings of operating parameters. Specifically, it can analyze abnormal patterns and energy efficiency shortcomings in energy consumption feature descriptors, and combine them with the output requirements of the production planning system and the real-time operating status of the equipment monitoring system to generate a set of optimization target constraints. This step is a prerequisite for building an optimization model. Its core is to extract the core problem from energy consumption characteristics, clarify the optimization boundaries based on actual production and equipment conditions, and ensure that subsequent optimizations both meet production needs and are engineering feasible. The specific implementation method is as follows: First, we conducted in-depth analysis of energy consumption characteristic descriptors to pinpoint abnormal patterns and energy efficiency shortcomings. The analysis of abnormal patterns focused on three dimensions: "abnormality type - frequency of occurrence - degree of impact." For example, the descriptor for the pulping process showed "abnormal power consumption frequency 3 times / day, each lasting 1 minute, the root cause being excessively high refiner speed (1550 r / min > rated 1480 r / min), resulting in an over-consumption of 15.2 kW of power," clearly indicating that this abnormality belongs to the "process parameter exceeding limit type abnormality," accounting for approximately 8% of the total energy consumption. The descriptor for the papermaking process mentioned "steam consumption fluctuation coefficient 8% > benchmark value 5%, correlation coefficient with wire dehumidification efficiency 0.78," classifying it as an "energy efficiency shortcoming type problem," with steam utilization rate 12% lower than the benchmark mode. The descriptor for the drying process indicated "drying cylinder steam leakage rate 8%, resulting in an over-consumption of 200 kg / h of steam," belonging to the "equipment defect type abnormality." Energy efficiency shortcomings also need to be addressed by quantifying energy-saving potential, such as a 15% potential for power optimization in the pulping process, a 10% potential for steam optimization in the papermaking process, and a 12% potential for steam optimization in the drying process, to provide a basis for setting optimization targets.

[0057] Secondly, integrate the production planning system's output requirements and clearly define the core target baseline for optimization. The production plan must encompass both short-term output targets and long-term capacity planning. For example, a short-term requirement might be "daily output ≥ 50 tons, paper type: linerboard, basis weight 80g / m³". 2 The long-term requirement is "energy consumption per ton of paper ≤ 800 kWh / ton". Production requirements need to be converted into quantifiable process parameter constraints, such as paper machine speed ≥ 800 m / min (corresponding to a daily output of 50 tons) and pulp concentration stable at 3.5%-4.0% (to ensure uniform basis weight). These parameters are directly related to equipment load and energy consumption and need to be included as hard constraints in the optimization conditions.

[0058] Finally, the constraints at the equipment level are supplemented by combining the real-time operating status of the equipment monitoring system. The real-time equipment status includes key parameters such as motor load rate, bearing temperature, valve opening, and vacuum pump pressure. For example, the current load rate of the pulper is 85% (rated load rate ≤ 90%), and the motor bearing temperature is 58℃ (safety threshold ≤ 70℃); the current steam valve opening of the drying cylinder is 75% (maximum opening 90%), and the cylinder surface temperature is 185℃ (process range 150-200℃); the current frequency of the water circulation pump is 50Hz (rated frequency 50Hz, adjustable range 30-50Hz). Based on the above information, an optimization objective constraint set is generated, comprising four categories: output constraints, equipment load constraints, process parameter constraints, and safety threshold constraints. The output constraints are: "Daily output ≥ 50 tons, paper machine speed ≥ 800 m / min, pulp concentration ∈ [3.5%, 4.0%]"; the equipment load constraints are: "Refiner motor power ≤ 200 kW (corresponding to 90% load rate), water circulation pump frequency ∈ [30 Hz, 50 Hz], dryer cylinder steam valve opening ∈ [50%, 90%]"; the process parameter constraints are: "Dryer cylinder temperature ∈ [150℃, 200℃], refiner speed ≤ 1480 r / min, wire section dehumidification pressure ∈ [0.3 MPa, 0.5 MPa]"; and the safety threshold constraints are: "Motor bearing temperature ≤ 70℃, steam pipeline pressure ≤ 1.0 MPa, motor current ≤ 600 A". This constraint set clearly defines the optimization objective direction and delineates insurmountable engineering boundaries, providing clear input for subsequent model construction.

[0059] A multi-objective optimization model is constructed based on the set of optimization objectives and constraints. The model uses energy minimization and output maximization as dual objective functions and employs particle swarm optimization to iteratively solve the model, generating a Pareto optimal solution set. This step is the core computational step in the optimization. By constructing a scientific mathematical model and utilizing the global search capability of the particle swarm optimization algorithm, the optimal solution set that balances energy consumption and output is found within the constraint boundary. The specific implementation method is as follows: First, the objective function and decision variables of the multi-objective optimization model are constructed. The decision variables are selected from key operating parameters that directly affect energy consumption and output, including: refiner motor power (P1, kW) and refiner speed (n1, r / min) in the pulping stage; paper machine speed (v, m / min) and pulp concentration (c, %) in the papermaking stage; and steam valve opening (k, %) and dryer temperature (T, °C) in the drying stage, and water circulation pump frequency (f, Hz), totaling 7 decision variables. The variable range strictly follows the constraint set. The dual objective functions are the energy consumption minimization function and the output maximization function: Energy consumption minimization function E_min = a×P1 + b×Q_steam (k,T) + d×Q_water (f), where a is the unit cost coefficient of electricity (0.15 yuan / kWh), b is the unit cost coefficient of steam (0.3 yuan / kg), and d is the unit cost coefficient of water resources (5 yuan / m³). 3 Q_steam(k,T) represents steam consumption (positively correlated with valve opening k and drying cylinder temperature T, with the fitting formula Q_steam=1500×k + 5×T - 50000, unit kg / h), and Q_water(f) represents water consumption (positively correlated with pump frequency f, with the fitting formula Q_water=2×f + 10, unit m). 3 / h); the production maximization function Y_max = k1×v×c, where k1 is the production coefficient (0.001 tons / (m·%), based on a paper basis weight of 80g / m). 2 (Derivation) That is, the output is positively correlated with the vehicle speed and the slurry concentration.

[0060] Secondly, configure the particle swarm optimization (PSO) algorithm parameters and start the iterative solution. The core of the PSO algorithm is to find the optimal solution by simulating the flight and cooperation of particles in the solution space. The parameter configuration needs to balance search efficiency and solution accuracy: the particle population size is set to 50 (i.e., searching 50 potential solutions at the same time), the maximum number of iterations is set to 100 (to ensure sufficient search), the inertia weight ω is initially 0.9, and linearly decreases to 0.4 during the iteration process (global search in the early stage, local optimization in the later stage), the learning factors c1 (individual cognitive factor) and c2 (social cognitive factor) are both set to 2.0 (balancing individual experience and group experience), and the boundaries of particle position and velocity are determined by the range of decision variables (e.g., the position boundary of the grinder power P1 [150kW, 200kW], velocity boundary [-10kW / iteration, 10kW / iteration]).

[0061] The iterative solution process is executed according to the following logic: Initialize 50 particles, each particle corresponding to a set of random values ​​of decision variables (e.g., the decision variables of an initial particle are P1=180kW, n1=1400r / min, v=820m / min, c=3.8%, k=65%, T=175℃, f=42Hz), calculate the biobjective function value of each particle (energy consumption E=180×0.15 + (1500×65+ 5×175 - 50000)×0.3 / 60 + (2×42 + 10)×5 / 60≈27 + (97500 + 875 - 50000)×0.005 + (94)×0.083≈27 + 48375×0.005 + 7.8≈27 + 241.9 + 7.8≈276.7 yuan / The hourly output is Y = 0.001 × 820 × 3.8 ≈ 3.116 tons / hour, and the daily output is approximately 74.8 tons. Then, the individual optimal position (pbest, the decision variable corresponding to its own historical optimal objective function) and the global optimal position (gbest, the decision variable corresponding to the population's historical optimal objective function) of each particle are updated. The update formula is: particle velocity v_i (t+1) = ω × v_i (t) + c1 × r1 × (pbest_i - x_i (t)) + c2 × r2 × (gbest - x_i (t)) (r1 and r2 are random numbers in [0,1]), particle position x_i (t+1) = x_i (t) + v_i (t+1). After updating, it is necessary to check whether the particle position exceeds the boundary of the decision variable. If it does, it is truncated to the boundary value. This iteration is repeated 100 times, and the dual objective function value of the particle is recorded after each iteration. Finally, the Pareto optimal solution is selected. This means there is no other set of solutions that are better than both objectives. For example, the decision variables for a Pareto optimal solution are P1=175kW, n1=1450r / min, v=850m / min, c=3.7%, k=60%, T=180℃, and f=40Hz, corresponding to an energy consumption of E≈262.5 yuan / hour and a daily output of ≈78.3 tons. Another optimal solution is P1=190kW, n1=1480r / min, v=880m / min, c=3.9%, k=70%, T=185℃, and f=45Hz, corresponding to an energy consumption of E≈302.8 yuan / hour and a daily output of ≈85.6 tons. These solutions constitute the Pareto optimal solution set, each representing a different energy consumption-output trade-off, providing diverse options for subsequent selection.

[0062] A feasibility assessment is performed on the Pareto optimal solution set, taking into account the maximum load capacity of the equipment, the range of process parameters, and safety production requirements. Feasible solutions that meet the actual working conditions are selected, and a set of candidate control schemes is generated. This step is a crucial screening process for optimizing solutions. Through rigorous feasibility verification, solutions that are theoretically feasible but not practically feasible are eliminated, ensuring that candidate solutions can be directly applied to production. The specific implementation method is as follows: The feasibility assessment employs a "three-layer verification" logic, sequentially verifying equipment load feasibility, process parameter feasibility, and safe production feasibility. Only the Pareto optimal solution that passes all verifications can enter the candidate solution set. The first layer, equipment load feasibility verification, focuses on confirming that the equipment load corresponding to the decision variable does not exceed the rated capacity. For example, a Pareto solution with a refiner power of 195kW corresponds to a load rate of 195 / 217≈89.8% (rated load rate 90%), which is within the limit and passes the verification. Another solution with a refiner power of 205kW has a load rate of 94.5%, exceeding the rated value and is directly eliminated. A solution with a water circulation pump frequency of 52Hz exceeds the adjustable range of 30-50Hz and is also eliminated. The second layer of process parameter feasibility verification ensures that decision variables meet the technical requirements of the production process. For example, a solution with a drying cylinder temperature of 145℃ is below the process range of 150℃, resulting in insufficient paper drying, and is therefore rejected; a pulp concentration of 3.3% is below the lower limit of the process of 3.5%, which will affect the basis weight uniformity of the paper, and is also rejected; a solution with a paper machine speed of 900m / min, a drying cylinder temperature of 190℃, and a steam valve opening of 80% meets the process requirements and passes the verification. The third layer of safety production feasibility verification focuses on checking whether the equipment operating parameters are within the safety thresholds. For example, a solution with a calculated motor bearing temperature of 68℃ (derived from power and speed) is below the safety threshold of 70℃ and passes the verification; another solution with a calculated steam pipeline pressure of 1.05MPa exceeds the safety threshold of 1.0MPa, posing a leakage risk, and is therefore rejected.

[0063] After three layers of verification, five feasible solutions that meet the actual working conditions were selected from the Pareto optimal solution set, generating a candidate control scheme set. Each scheme clearly includes the specific values ​​of all decision variables, the corresponding energy consumption and output indicators, and the core advantages of the scheme. Scheme 1: Refiner power 175kW, speed 1450r / min, paper machine speed 850m / min, pulp concentration 3.7%, dryer valve opening 60%, temperature 180℃, water circulation pump frequency 40Hz; energy consumption 262.5 yuan / hour, daily output 78.3 tons; core advantage: lowest energy consumption, suitable for low-capacity demand scenarios. Option 2: Refiner power 180kW, speed 1460r / min, paper machine speed 860m / min, pulp concentration 3.8%, dryer valve opening 65%, temperature 182℃, water circulation pump frequency 42Hz; energy consumption 278.3 yuan / hour, daily output 80.5 tons; core advantage: balanced energy consumption and output, strong versatility. Option 3: Refiner power 185kW, speed 1470r / min, paper machine speed 870m / min, pulp concentration 3.8%, dryer valve opening 68%, temperature 183℃, water circulation pump frequency 43Hz; energy consumption 290.5 yuan / hour, daily output 82.7 tons; core advantage: moderate output, controllable energy consumption. Option 4: Refiner power 190kW, speed 1480r / min, paper machine speed 880m / min, pulp concentration 3.9%, dryer valve opening 70%, temperature 185℃, water circulation pump frequency 45Hz; energy consumption 302.8 yuan / hour, daily output 85.6 tons; core advantage: high output, meeting peak capacity demand. Option 5: Refiner power 195kW, speed 1480r / min, paper machine speed 890m / min, pulp concentration 3.9%, dryer valve opening 72%, temperature 188℃, water circulation pump frequency 47Hz; energy consumption 315.2 yuan / hour, daily output 88.4 tons; core advantage: highest output, suitable for urgent order scenarios. The candidate control scheme set covers different capacity requirements, and each scheme has undergone full-dimensional feasibility verification to ensure risk-free implementation.

[0064] The candidate control scheme set is transformed into specific executable instructions, including the adjustment value of the pulper motor power, the setting value of the steam valve opening of the drying cylinder, and the frequency control parameters of the water circulation pump, to generate energy-saving control strategies for specific machines and process sections.

[0065] This step is the final stage of implementing the optimization results. It transforms the abstract scheme parameters into operational instructions that the equipment can directly execute, clarifies the control objects, adjustment values, execution timing, and verification standards, and generates a standardized energy-saving control strategy. The specific implementation method is as follows: First, candidate solutions are broken down logically into "machine - process section - parameter type," and the decision variables of each solution are transformed into operation instructions for specific equipment. The operation instructions must clearly define the control object (specific machine number), current parameter value, target parameter value, adjustment range, adjustment duration (to avoid equipment shock caused by sudden parameter changes), and verification indicators (energy consumption and output targets to be achieved after adjustment). Taking the most widely used solution two (energy consumption and output balance) as an example, the specific instructions after transformation are as follows: Pulping stage #1 refiner: Current power 185kW → target power 180kW, adjustment range -5kW, adjustment duration 3 minutes (divided into 3 steps, each step -1.7kW, 1 minute interval); Current speed 1480r / min → target speed 1460r / min, adjustment range -20r / min, adjustment duration 2 minutes (divided into 2 steps, each step -10r / min, 1 minute interval); Verification indicators: After adjustment, motor load rate 83%, power consumption ≤180kW, bearing temperature ≤65℃. Papermaking process, No. 2 paper machine: Current speed 820m / min → target speed 860m / min, adjustment increment +40m / min, adjustment time 4 minutes (in 4 steps, each step +10m / min, 1 minute interval); Current pulp concentration 3.6% → target concentration 3.8%, adjustment increment +0.2%, achieved by adjusting the pulp delivery pump frequency, adjustment time 1 minute; Verification indicators: stable machine speed 860±5m / min, pulp concentration 3.8±0.1%, daily output ≥80 tons. Drying Stage #3 Drying Cylinder Group: Current steam valve opening 75% → Target opening 65%, adjustment range -10%, adjustment time 2 minutes (in 2 steps, -5% each, 1 minute interval); Current temperature 178℃ → Target temperature 182℃, adjustment range +4℃, achieved through steam pressure fine-tuning, adjustment time 1 minute; Verification indicators: Drying cylinder temperature 182±2℃, steam consumption ≤1800kg / h. Water Circulation System #4 Circulation Pump: Current frequency 45Hz → Target frequency 42Hz, adjustment range -3Hz, adjustment time 1 minute; Verification indicators: Water supply flow rate stabilized at 54m³ / h. 3 / h, pipeline pressure 0.4MPa±0.05MPa.

[0066] Secondly, all equipment operation instructions are integrated, and energy-saving control strategies are generated for specific machines and process sections according to process segment. Each strategy includes four parts: control objectives, execution procedures, verification standards, and emergency plans. Control objectives: Daily output ≥ 80.5 tons, unit product energy consumption ≤ 780 kWh / ton, refiner power consumption ≤ 180 kW, dryer steam consumption ≤ 1800 kg / h, and water circulation pump energy consumption reduced by 5%. Execution procedures: Step 1 (0-1 minute): Adjust water circulation pump frequency to 42 Hz; Step 2 (1-3 minutes): Adjust dryer steam valve opening to 65% and temperature to 182℃; Step 3 (3-6 minutes): Adjust refiner speed to 1460 r / min and power to 180 kW; Step 4 (6-10 minutes): Adjust paper machine speed to 860 m / min and pulp concentration to 3.8%; Equipment status is monitored in real-time throughout the process, and parameters are recorded every minute. Verification Standards: After execution, the machine must operate stably for 30 minutes, during which energy consumption must meet the following requirements: E ≤ 278.3 yuan / hour, output must meet the following requirement: Y ≥ 80.5 tons / day, and all equipment status parameters must be within the constraints. Emergency Plan: If the motor bearing temperature exceeds 68℃ during adjustment, immediately suspend the mill power adjustment, maintain the current speed, and continue only after the temperature drops below 65℃; if the drying cylinder temperature fluctuates by more than ±5℃, suspend temperature adjustment, check the steam pipeline pressure, and eliminate any leaks before resuming execution.

[0067] For other candidate solutions, the same logic is used to transform them into corresponding control strategies. For example, the strategy of Solution 1 (low energy consumption) focuses on reducing the power of the refiner and the opening of the dryer valve, while the strategy of Solution 4 (high output) focuses on increasing the paper machine speed and pulp concentration. This forms a series of control strategies that cover different needs. The final energy-saving control strategy is both specific and executable, and has flexible adaptability. The corresponding solution can be selected and implemented according to actual production needs.

[0068] S204, execute the energy-saving control strategy and continuously monitor energy consumption data. When the actual energy consumption deviates from the preset threshold, an early warning mechanism is automatically triggered. At the same time, corresponding energy efficiency correction measures are initiated according to the early warning level to achieve refined energy management in the paper production process.

[0069] Specifically, the energy-saving control strategy can be executed through the industrial control system to adjust operating parameters including motor speed, valve opening and pump station frequency, and generate a strategy execution confirmation signal. This step is the core execution stage for implementing energy-saving control strategies. As the core carrier of instruction execution, the industrial control system needs to accurately parse the operation instructions in the control strategy, drive equipment parameter adjustments through standardized communication protocols, and generate execution confirmation signals to ensure that the instructions are implemented. The specific implementation method is as follows: The industrial control system adopts a distributed control system (DCS). This system establishes real-time communication with the field controllers (PLCs) in the pulping, papermaking, and drying processes via industrial Ethernet. The communication protocol adopts Profinet, with a transmission rate of 100Mbps and a latency of ≤10ms, ensuring the real-time performance and stability of command transmission. The system first parses the equipment control commands in the energy-saving control strategy, breaking them down logically into "machine number - parameter type - target value - adjustment rate - execution duration". For example, the command for the #1 refiner in the pulping stage is parsed as "machine: #1 refiner; parameter type: motor speed; current value: 1480 r / min; target value: 1460 r / min; adjustment rate: 10 r / min / second; execution duration: 2 seconds". The command for the #3 drying cylinder in the drying stage is parsed as "machine: #3 drying cylinder; parameter type: steam valve opening; current value: 75%; target value: 65%; adjustment rate: 5% / second; execution duration: 2 seconds". The command for the #4 circulating pump in the water circulation system is parsed as "machine: #4 circulating pump; parameter type: operating frequency; current value: 45 Hz; target value: 42 Hz; adjustment rate: 1.5 Hz / second; execution duration: 2 seconds".

[0070] During command execution, the system drives the equipment actuators by outputting analog signals (4-20mA current signals) or digital signals (switching signals) through the field controller: motor speed adjustment is achieved by controlling the output frequency of the frequency converter. For example, the frequency converter of the pulping machine receives a 12mA current signal (corresponding to 1460r / min) and gradually reduces the output frequency from 50Hz (1480r / min) to 49.3Hz (1460r / min) at a rate of 10r / min / second, avoiding sudden speed changes that could impact the equipment load; steam valve opening adjustment is achieved through an electric actuator. The actuator receives a 10mA current signal (corresponding to 65% opening) from the controller and drives the valve core to move at a rate of 5% / second, providing real-time feedback of the opening signal to the system; pump station frequency adjustment is achieved through a frequency converter control cabinet. The controller sends a 42Hz frequency command to the frequency converter, and the frequency converter adjusts the output at a rate of 1.5Hz / second, causing the circulating pump motor speed to change synchronously.

[0071] The generation of a strategy execution confirmation signal must meet two conditions: "parameters meet target + stable status." The system collects the adjusted operating parameters of the equipment in real time. When the parameters reach the target value and remain stable for 3 consecutive seconds (fluctuation ≤ ±1%), the command is considered successfully executed. The confirmation signal includes core information such as execution timestamp, machine number, parameter type, target value, actual stable value, and execution status (success / failure). For example, a confirmation signal might be: "Timestamp: 2025-07-03 14:30:05.123; Machine: #1 grinding machine; Parameter type: motor speed; Target value: 1460 r / min; Actual stable value: 1459.8 r / min; Execution status: Success." If the parameter adjustment fails to reach the target value (e.g., valve jamming causing the opening to only reach 68%), the execution status is marked as "failure," accompanied by a fault code (e.g., "E001 - Valve actuator jamming"). Simultaneously, the system retry mechanism is triggered, retrying twice within 30 seconds. If the retry still fails, a fault alarm is pushed to the maintenance terminal. All confirmation signals are uploaded to the energy management platform in real time, forming a traceable record of command execution.

[0072] Real-time collection of energy consumption data after strategy execution, calculation of unit product energy consumption index and comparison with dynamic thresholds set based on historical data, generating energy consumption deviation assessment report; This step is the core of verifying the effectiveness of energy consumption control. By collecting real-time data, calculating indicators, and comparing thresholds, the energy consumption deviation after the strategy is implemented is quantified, providing data support for subsequent early warning and correction. The specific implementation method is as follows: After the strategy is implemented, the multi-source sensor network continues to collect energy consumption data at the original acquisition frequency: electricity data is collected at 1Hz, and steam and water resource data are collected at 0.2Hz. The collected data is transmitted in real time to the energy management platform via industrial Ethernet. The platform performs real-time analysis and integration of the data, extracting the electricity consumption (kW), steam consumption (kg / h), and water consumption (m³) for each stage. 3 Core data ( / h), such as the average power consumption of #1 refiner within 10 minutes after strategy execution: 179.5kW; the average steam consumption of #3 drying cylinder: 1780kg / h; and the average water consumption of #4 circulating pump: 53.8m³ / h. 3 / h.

[0073] The calculation of unit product energy consumption index adopts the core logic of "time period energy consumption / time period output", with the time period set to 1 hour, meaning that unit product energy consumption is calculated once per hour. First, the total energy consumption within 1 hour is calculated: total electrical energy consumption is the sum of electricity consumption in each stage (e.g., pulping 179.5kW + papermaking 218.3kW + drying 149.2kW = 547kW), total steam energy consumption is the sum of steam consumption in each stage (e.g., papermaking 1650kg / h + drying 1780kg / h = 3430kg / h), and total water energy consumption is the sum of water consumption in each stage (e.g., pulping 42.5m... 3 / h + Papermaking 53.8m 3 / h + drying 0.5m 3 / h=96.8m 3 / h); then the total output within 1 hour is calculated using the paper machine's basis weight sensor (80g / m³). 2 Based on the vehicle speed sensor (860m / min), the production output = vehicle speed × width × quantity × time = 860m / min × 2.5m × 80g / m 2 ×60min=860×2.5×80×60×10^-6 tons= 10.32 tons; the final unit product energy consumption indicators are: unit electricity energy consumption=547kWh / 10.32 tons≈53.0kWh / ton, unit steam energy consumption=3430kg / 10.32 tons≈332.4kg / ton, unit water energy consumption= 96.8m 3 / 10.32 tons ≈ 9.4m 3 / ton.

[0074] The dynamic threshold is set based on a comprehensive analysis of historical data and production plans, and adopts the calculation logic of "average of the past 30 days ± dynamic deviation". The dynamic deviation is adjusted according to the production load: the deviation is ±5% when the production is at full load (output ≥ 50 tons / day), ±8% when the production is at medium load (30-50 tons / day), and ±10% when the production is at low load (< 30 tons / day). For example, the average unit electricity consumption at full load over the past 30 days is 55.0 kWh / ton, with a dynamic threshold range of 55.0 × (1-5%) = 52.25 kWh / ton to 55.0 × (1+5%) = 57.75 kWh / ton; the average unit steam consumption at full load over the past 30 days is 345.0 kg / ton, with a dynamic threshold range of 345.0 × (1-5%) = 327.75 kg / ton to 345.0 × (1+5%) = 362.25 kg / ton; and the average unit water consumption at full load over the past 30 days is 9.8 m³. 3 / ton, the dynamic threshold range is 9.8×(1-5%)=9.31m 3 / tons to 9.8 × (1 + 5%) = 10.29 m 3 / ton.

[0075] The energy consumption deviation assessment report is generated based on the core logic of "indicator comparison - deviation calculation - preliminary cause analysis". The deviation calculation uses the percentage deviation formula: Deviation percentage = (actual unit energy consumption - median dynamic threshold) / median dynamic threshold × 100%, where the median is the average of the upper and lower limits of the threshold (e.g., median electricity consumption = (52.25 + 57.75) / 2 = 55.0 kWh / ton). For example, if the actual unit electricity consumption is 53.0 kWh / ton, the deviation percentage = (53.0 - 55.0) / 55.0 × 100% ≈ -3.6% (negative deviation indicates energy saving effect, positive deviation indicates excessive consumption); if the actual unit steam energy consumption is 332.4 kg / ton, the deviation percentage = (332.4 - 345.0) / 345.0 × 100% ≈ -3.7%; if the actual unit water energy consumption is 9.4 m³ / ton... 3 / ton, deviation percentage = (9.4-9.8) / 9.8×100%≈-4.1%. The report includes the assessment period, actual values ​​of energy consumption indicators for each unit, dynamic threshold range, deviation percentage, production links involved, and preliminary reasons for the deviation (e.g., negative deviation indicates effective control strategies, positive deviation may be due to equipment leakage, parameter drift, etc.). For example, a report excerpt reads: "Assessment period: 2025-07-03 15:00-16:00; Unit power consumption: 53.0 kWh / ton, threshold range 52.25-57.75 kWh / ton, deviation percentage -3.6%; Unit steam consumption: 332.4 kg / ton, threshold range 327.75-362.25 kg / ton, deviation percentage -3.7%; Links involved: pulping, drying; Preliminary reasons: adjustment of pulper speed reduced power consumption, optimization of drying cylinder valve opening reduced steam waste, and control strategies were effectively implemented."

[0076] The early warning mechanism is automatically triggered based on the degree of deviation in the energy consumption deviation assessment report. The early warning level is divided into three levels: yellow, orange, and red, according to the percentage of deviation, and a graded early warning instruction is generated. This step is the rapid response stage for abnormal energy consumption. The early warning mechanism is based on the quantification and classification of deviation percentage, and is triggered through audible and visual alarms, system pop-ups, SMS push notifications, etc., to ensure that abnormal energy consumption is noticed in a timely manner. The specific implementation method is as follows: The warning levels are strictly based on the absolute value of the deviation percentage, combined with the energy consumption control precision requirements of the paper industry, setting three warning thresholds: A yellow warning corresponds to an absolute deviation percentage between 5% and 10%, indicating a slight abnormality in energy consumption, within the safe operating range, but requiring monitoring of the trend; an orange warning corresponds to an absolute deviation percentage between 10% and 15%, indicating a significant abnormality in energy consumption, which may affect energy-saving targets or equipment operating efficiency, requiring timely intervention; a red warning corresponds to an absolute deviation percentage exceeding 15%, indicating a serious abnormality in energy consumption, with a risk of equipment failure or process loss of control, requiring emergency handling. A positive deviation percentage (excessive energy consumption) directly triggers a warning, while a negative deviation percentage exceeding the threshold (excessive energy saving may affect output or product quality) also triggers a warning. For example, if the actual unit power consumption is 49.0 kWh / ton, and the deviation percentage is -10.9%, an orange warning is triggered.

[0077] The early warning mechanism is triggered as follows: the energy management platform analyzes the energy consumption deviation assessment report in real time, extracts the deviation percentage of each unit's energy consumption indicators, and immediately activates the corresponding level of early warning if any indicator reaches the early warning threshold. The triggering method for a yellow alert is as follows: a yellow alert pop-up window appears on the system interface, displaying the alert level, involved indicators, deviation percentage, and assessment period. At the same time, the audible and visual alarm in the on-site control room emits a low-frequency buzzer (frequency 500Hz, interval 2 seconds) for 3 minutes. The triggering method for an orange alert is as follows: a system pop-up window + a high-frequency buzzer (frequency 1000Hz, interval 1 second) + a text message push to the mobile phone of maintenance personnel. The text message contains the message "Orange alert: 2025-07-03 16:00-17:00, unit steam energy consumption 378.5kg / ton, deviation percentage 9.7%, involving the drying process, please check in time." The triggering method for a red alert is as follows: a system pop-up window + a continuous buzzer + a text message push + a telephone notification to management personnel. The telephone notification automatically dials a preset number through the system's integrated voice module and plays the alert voice message "Emergency notification: A red alert has been triggered. Unit power energy consumption is 66.2kWh / ton, deviation percentage 20.4%, involving the pulping process. There may be equipment failure. Please handle it immediately."

[0078] The generation of tiered early warning instructions includes core information such as "early warning level - involved indicators - deviation data - handling suggestions - response time limit". For example, a yellow early warning instruction would be "early warning level: yellow; involved indicators: unit water energy consumption 10.8m". 3 / ton; Deviation percentage: 10.2%; Handling suggestion: Check the operating status of the water circulation pump and check for pipeline leaks; Response time limit: within 1 hour; Orange warning instruction is "Warning level: Orange; Involved indicator: Unit steam energy consumption 392.0 kg / ton; Deviation percentage: 13.6%; Handling suggestion: Check the sealing of the drying cylinder insulation layer and adjust the steam valve opening to the optimized range; Response time limit: within 30 minutes"; Red warning instruction is "Warning level: Red; Involved indicator: Unit power energy consumption 66.2 kWh / ton; Deviation percentage: 20.4%; Handling suggestion: Immediately reduce the load of the grinding machine, stop the machine and check whether the motor is stuck or overloaded, and start the backup equipment simultaneously; Response time limit: within 10 minutes". The warning instructions are synchronized to the industrial control system and operation and maintenance management platform in real time, providing clear guidance for subsequent correction measures.

[0079] Based on the graded early warning instructions, corresponding energy efficiency correction measures are initiated. During a yellow warning, equipment parameters are automatically fine-tuned; during an orange warning, auxiliary equipment is activated for coordinated adjustment; and during a red warning, production cycle adjustments are triggered and management personnel are notified to intervene, forming a closed-loop refined energy management system.

[0080] This step is the core of the closed-loop handling of energy consumption anomalies. It accurately addresses different degrees of energy consumption deviations through tiered correction measures, and establishes a closed-loop mechanism of "measure implementation - effect monitoring - data feedback - threshold optimization" to achieve refined energy management. The specific implementation method is as follows: The automatic fine-tuning measures during a yellow alert focus on "parameter micro-correction," requiring no manual intervention. The industrial control system automatically adjusts the operating parameters of relevant equipment based on the indicators and handling suggestions in the alert command, with the adjustment range controlled within ±3%, avoiding any impact on production. For example, if a yellow alert is triggered by unit water energy consumption (10.8m...),... 3 If the unit steam energy consumption triggers a yellow alert (365.0 kg / ton, deviation 5.8%), the system automatically adjusts the frequency of the No. 4 circulating pump from 42Hz to 41Hz to reduce the water supply flow, while simultaneously monitoring the pipeline pressure (maintaining it at 0.4MPa±0.05MPa) to ensure production water needs are met. If the unit steam energy consumption triggers a yellow alert (365.0 kg / ton, deviation 5.8%), the system automatically adjusts the opening of the drying cylinder steam valve from 65% to 63%, while simultaneously monitoring the drying cylinder temperature (maintaining it at 182℃±2℃) to prevent temperature fluctuations from affecting the paper drying effect. After fine-tuning, energy consumption data is continuously monitored, and a micro-assessment report is generated every 5 minutes. If the deviation percentage drops below 5% within 30 minutes, the alert is automatically lifted; if there is no improvement, it is upgraded to an orange alert.

[0081] During an orange alert, the coordinated adjustment of auxiliary equipment focuses on "main equipment + auxiliary equipment linkage." This is achieved by activating energy-saving auxiliary equipment or adjusting auxiliary parameters to optimize energy utilization efficiency. For example, if the unit steam energy consumption triggers an orange alert (392.0 kg / ton, deviation 13.6%), the system, in addition to fine-tuning the opening of the drying cylinder valves, activates the steam waste heat recovery device to guide the waste steam discharged from the drying cylinder into the pulp preheating system, replacing some of the fresh steam consumption. Simultaneously, the induced draft fan frequency is adjusted from 50Hz to 48Hz to reduce heat loss. If the unit electricity energy consumption triggers an orange alert (60.5 kWh / ton, deviation 10.0%), the system activates the reactive power compensation device to improve the power factor from 0.88 to 0.95, reducing ineffective power consumption. Simultaneously, the grinding gap of the pulper is fine-tuned from 0.2mm to 0.22mm to reduce motor load. After coordinated adjustment, the system generates an evaluation report every 10 minutes. If the deviation percentage drops below 10% within 1 hour, the warning is downgraded to yellow and monitored continuously; if it drops below 5%, the warning is lifted; if there is no improvement, it is upgraded to a red warning.

[0082] During a red alert, production rhythm adjustments focus on "risk control + emergency response," prioritizing equipment safety and product quality while simultaneously notifying management personnel for intervention. For example, if a red alert is triggered by unit power consumption (66.2 kWh / ton, deviation 20.4%), the system immediately reduces the paper machine speed from 860 m / min to 750 m / min (production rhythm adjustment) to lower the overall production load. Simultaneously, the refiner power is reduced from 180 kW to 150 kW to prevent motor overload damage. If an equipment malfunction is detected (e.g., motor bearing temperature rises to 68°C), the system automatically shuts down the machine and starts a backup refiner to ensure production continuity. Upon receiving notification, management personnel must arrive on-site within 10 minutes to organize troubleshooting and repair. After repair, the system adjusts parameters according to a "gradual recovery" principle (e.g., increasing refiner power by 10 kW every 5 minutes) to avoid sudden parameter changes. The conditions for lifting a red alert are: equipment malfunction resolved, energy consumption deviation percentage ≤ 5% for two consecutive hours, and product quality indicators (basis weight, thickness) meeting requirements.

[0083] The construction of a closed-loop energy management system runs through the entire disposal process. The core process is as follows: After the measures are implemented, the system continuously collects energy consumption and production data and provides real-time feedback on the correction effect; every 24 hours, the early warning disposal situation is summarized and analyzed, and dynamic thresholds are optimized (for example, if a certain link frequently triggers a yellow warning, the median threshold of that indicator can be adjusted to the average of the past 7 days); a monthly energy management report is generated, summarizing energy consumption patterns, early warning disposal efficiency, energy-saving effects, and proposing improvement suggestions such as parameter optimization and equipment maintenance; at the same time, all data is recorded in the energy management database and retained for more than one year to provide data support for subsequent strategy optimization and equipment upgrades. For example, through closed-loop management, the power consumption early warning disposal efficiency in the pulping process has been shortened from 30 minutes to 15 minutes, and the comprehensive energy consumption per ton of paper has been reduced from 800 kWh / ton to 760 kWh / ton, achieving precise control and continuous optimization of energy consumption.

[0084] As can be seen, real-time data collection of electricity, steam, and water consumption is used to construct a multi-dimensional energy data set including flow, pressure, and temperature parameters. Based on this multi-dimensional energy data set, a dynamic pattern recognition algorithm is employed to analyze the energy consumption characteristics of each production stage, generating energy consumption characteristics. According to these energy consumption characteristics, combined with real-time production plans and equipment operating status, an intelligent optimization algorithm generates energy-saving control strategies for specific machines and process sections. The energy-saving control strategies are executed, and energy consumption data is continuously monitored. When actual energy consumption deviates from a preset threshold, an early warning mechanism is automatically triggered, and corresponding energy efficiency correction measures are initiated based on the warning level. This achieves refined energy management in the papermaking process, enabling closed-loop management from energy data collection to optimization strategy execution, thereby improving the energy utilization efficiency of papermaking production.

[0085] Another embodiment of the present invention provides a papermaking energy management system, see [link to relevant documentation]. Figure 3 The system may include: The acquisition module 301 is used to collect real-time data on the consumption of electricity, steam and water resources through a multi-source sensor network deployed in the pulping, papermaking and drying processes, and to construct a multi-dimensional energy data set including flow rate, pressure and temperature parameters. The identification module 302 is used to analyze the energy consumption characteristics of each production link based on the multidimensional energy data set and a dynamic pattern recognition algorithm, identify energy consumption patterns and abnormal fluctuation patterns, and generate energy consumption characteristics. The generation module 303 is used to generate energy-saving control strategies for specific machines and process sections based on the energy consumption characteristics, combined with real-time production plans and equipment operating status, through intelligent optimization algorithms. The strategies include load adjustment schemes and optimized settings of operating parameters. The execution module 304 is used to execute the energy-saving control strategy and continuously monitor energy consumption data. When the actual energy consumption deviates from the preset threshold, the early warning mechanism is automatically triggered. At the same time, the corresponding energy efficiency correction measures are initiated according to the early warning level to achieve refined energy management in the paper production process.

[0086] This invention also provides a storage medium storing a computer program, wherein the computer program is configured to execute the steps in any of the above method embodiments when running.

[0087] This invention also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to perform the steps in any of the above method embodiments.

[0088] Specifically, the aforementioned electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the aforementioned processor, and the input / output device is connected to the aforementioned processor.

[0089] The above description, based on the embodiments shown in the figures, details the structure, features, and effects of the present invention. The above description is only a preferred embodiment of the present invention, but the present invention is not limited to the scope of implementation shown in the figures. Any changes made in accordance with the concept of the present invention, or equivalent embodiments modified to have equivalent changes, that do not exceed the spirit covered by the specification and figures, should be within the protection scope of the present invention.

Claims

1. A method for managing energy in a papermaking process, the method comprising: The method includes: By deploying a multi-source sensor network in the pulping, papermaking and drying processes, real-time data on the consumption of electricity, steam and water resources are collected, and a multi-dimensional energy data set including flow rate, pressure and temperature parameters is constructed. Based on the aforementioned multidimensional energy data set, a dynamic pattern recognition algorithm is used to analyze the energy consumption characteristics of each production stage, identify energy consumption patterns and abnormal fluctuation patterns, and generate energy consumption characteristics. Based on the energy consumption characteristics, combined with real-time production plans and equipment operating status, an intelligent optimization algorithm is used to generate energy-saving control strategies for specific machines and process sections. The strategies include load adjustment schemes and optimized settings of operating parameters. The energy-saving control strategy is implemented and energy consumption data is continuously monitored. When the actual energy consumption deviates from the preset threshold, an early warning mechanism is automatically triggered. At the same time, corresponding energy efficiency correction measures are initiated according to the early warning level to achieve refined energy management in the paper production process.

2. The method of claim 1, wherein, The aforementioned system utilizes a multi-source sensor network deployed in the pulping, papermaking, and drying processes to collect real-time data on electricity, steam, and water consumption, and constructs a multi-dimensional energy data set including flow rate, pressure, and temperature parameters, comprising: Smart meters, steam flow meters, and water flow meters are deployed on key equipment in the pulping process (refining mill), the papermaking process (papermaking machine), and the drying process (drying cylinder), while pressure sensors and temperature sensors are also installed to generate a multi-source sensor network configuration scheme. Sensor data, including three-phase current and voltage of power parameters, mass flow rate of steam, volumetric flow rate of water resources, and pipeline pressure and temperature readings, are acquired in real time through industrial Ethernet and wireless communication protocols to generate raw sensor data streams. The raw sensor data stream is cleaned and format standardized. A sliding window algorithm is used to eliminate instantaneous noise and compensate for sensor drift error, generating a calibrated time-series data sequence. The calibrated time-series data sequences are aligned dimensionally according to production stages and equipment numbers, and the flow, pressure, and temperature parameters of electricity, steam, and water are integrated to construct a time-synchronized multidimensional energy data set.

3. The method of claim 2, wherein, Based on the multidimensional energy data set, a dynamic pattern recognition algorithm is used to analyze the energy consumption characteristics of each production stage, identify energy consumption patterns and abnormal fluctuation patterns, and generate energy consumption characteristics, including: Energy consumption time-series data for each stage of pulping, papermaking and drying are extracted from a multidimensional energy dataset. Electricity consumption sequence, steam consumption sequence and water consumption sequence are constructed with time as the axis, and a stage-specific energy consumption data matrix is ​​generated. The dynamic time warping algorithm is applied to perform pattern matching on the energy consumption data matrix of each stage, identify typical energy consumption curves under different production loads, and generate a benchmark energy consumption pattern library. Based on the benchmark energy consumption pattern library, the density clustering algorithm is used to detect deviation patterns in real-time energy consumption data, calculate the similarity distance between each data point and the benchmark pattern, and generate abnormal fluctuation identification results. By combining the results of abnormal fluctuation identification with production process parameters, the correlation between abnormal energy consumption and equipment operating status is analyzed, and key feature indicators are extracted to form an energy consumption feature descriptor.

4. The method of claim 3, wherein, Based on the energy consumption characteristics and combined with real-time production plans and equipment operating status, an intelligent optimization algorithm generates energy-saving control strategies for specific machines and process sections. These strategies include load adjustment schemes and optimized operating parameter settings, including: Analyze the abnormal patterns and energy efficiency shortcomings in the energy consumption feature descriptors, and combine them with the output requirements of the production planning system and the real-time operating status of the equipment monitoring system to generate a set of optimization target constraints. A multi-objective optimization model is constructed based on the set of optimization objectives and constraints. The model uses energy minimization and output maximization as dual objective functions and employs particle swarm optimization to iteratively solve the model, generating a Pareto optimal solution set. A feasibility assessment is performed on the Pareto optimal solution set, taking into account the maximum load capacity of the equipment, the range of process parameters, and safety production requirements. Feasible solutions that meet the actual working conditions are selected, and a set of candidate control schemes is generated. The candidate control scheme set is transformed into specific executable instructions, including the adjustment value of the pulper motor power, the setting value of the steam valve opening of the drying cylinder, and the frequency control parameters of the water circulation pump, to generate energy-saving control strategies for specific machines and process sections.

5. The method of claim 4, wherein, The process involves executing the energy-saving control strategy and continuously monitoring energy consumption data. When actual energy consumption deviates from a preset threshold, an early warning mechanism is automatically triggered. Simultaneously, corresponding energy efficiency correction measures are initiated based on the early warning level, thereby achieving refined energy management in the paper production process, including: The industrial control system executes equipment control commands in the energy-saving regulation strategy, adjusts operating parameters including motor speed, valve opening and pump station frequency, and generates strategy execution confirmation signals. Real-time collection of energy consumption data after strategy execution, calculation of unit product energy consumption index and comparison with dynamic thresholds set based on historical data, generating energy consumption deviation assessment report; The early warning mechanism is automatically triggered based on the degree of deviation in the energy consumption deviation assessment report. The early warning level is divided into three levels: yellow, orange, and red, according to the percentage of deviation, and a graded early warning instruction is generated. Based on the graded early warning instructions, corresponding energy efficiency correction measures are initiated. During a yellow warning, equipment parameters are automatically fine-tuned; during an orange warning, auxiliary equipment is activated for coordinated adjustment; and during a red warning, production cycle adjustments are triggered and management personnel are notified to intervene, forming a closed-loop refined energy management system.

6. A papermaking energy management system characterized by, The system includes: The data acquisition module is used to collect real-time data on the consumption of electricity, steam and water resources through a multi-source sensor network deployed in the pulping, papermaking and drying processes, and to construct a multi-dimensional energy data set including flow rate, pressure and temperature parameters. The identification module is used to analyze the energy consumption characteristics of each production link based on the multidimensional energy data set and a dynamic pattern recognition algorithm, identify energy consumption patterns and abnormal fluctuation patterns, and generate energy consumption characteristics. The generation module is used to generate energy-saving control strategies for specific machines and process sections based on the energy consumption characteristics, combined with real-time production plans and equipment operating status, through intelligent optimization algorithms. The strategies include load adjustment schemes and optimized settings of operating parameters. The execution module is used to execute the energy-saving control strategy and continuously monitor energy consumption data. When the actual energy consumption deviates from the preset threshold, the early warning mechanism is automatically triggered. At the same time, the corresponding energy efficiency correction measures are initiated according to the early warning level to achieve refined energy management in the paper production process.

7. The system of claim 6, wherein, The acquisition module is specifically used for: Smart meters, steam flow meters, and water flow meters are deployed on key equipment in the pulping process (refining mill), the papermaking process (papermaking machine), and the drying process (drying cylinder), while pressure sensors and temperature sensors are also installed to generate a multi-source sensor network configuration scheme. Sensor data, including three-phase current and voltage of power parameters, mass flow rate of steam, volumetric flow rate of water resources, and pipeline pressure and temperature readings, are acquired in real time through industrial Ethernet and wireless communication protocols to generate raw sensor data streams. The raw sensor data stream is cleaned and format standardized. A sliding window algorithm is used to eliminate instantaneous noise and compensate for sensor drift error, generating a calibrated time-series data sequence. The calibrated time-series data sequences are aligned dimensionally according to production stages and equipment numbers, and the flow, pressure, and temperature parameters of electricity, steam, and water are integrated to construct a time-synchronized multidimensional energy data set.

8. The system of claim 7, wherein, The identification module is specifically used for: Energy consumption time-series data for each stage of pulping, papermaking and drying are extracted from a multidimensional energy dataset. Electricity consumption sequence, steam consumption sequence and water consumption sequence are constructed with time as the axis, and a stage-specific energy consumption data matrix is ​​generated. The dynamic time warping algorithm is applied to perform pattern matching on the energy consumption data matrix of each stage, identify typical energy consumption curves under different production loads, and generate a benchmark energy consumption pattern library. Based on the benchmark energy consumption pattern library, the density clustering algorithm is used to detect deviation patterns in real-time energy consumption data, calculate the similarity distance between each data point and the benchmark pattern, and generate abnormal fluctuation identification results. By combining the results of abnormal fluctuation identification with production process parameters, the correlation between abnormal energy consumption and equipment operating status is analyzed, and key feature indicators are extracted to form an energy consumption feature descriptor.

9. A storage medium, characterized by The storage medium stores a computer program, wherein the computer program is configured to execute the method of any one of claims 1-5 when it is run.

10. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to run the computer program to perform the method of any one of claims 1-5.

Images