Graphene production multi-zone linkage temperature control method based on edge computing

Abstract

The application discloses a kind of based on edge computing's graphene production multi-zone linkage temperature control method, it is related to industrial process control technical field. Including the following steps: using edge computing node to collect each temperature control area temperature control data and pre-processing, generate standardization multi-zone temperature control data;Association control cycle constructs historical heating input track, historical temperature response track and current control track;According to heat input and temperature response relationship determines directional heat coupling relationship and regional heat response time lag;Based on time lag alignment constructs multi-zone time lag alignment Hankel matrix;According to production condition matching generation historical track weight, constraint DeePC track combination coefficient;Input improved DeePC algorithm prediction temperature track and candidate heating power track;Combining heat coupling relationship collaborative solution linkage temperature control execution amount and closed loop update.The application realizes graphene production multi-temperature zone intelligent linkage temperature control, with efficient collaborative control capability.

Description

Technical Field

[0001] This invention relates to the field of industrial process control technology, and in particular to a multi-zone linkage temperature control method for graphene production based on edge computing. Background Technology

[0002] The graphene production process involves high-temperature reactions, heat transfer, and multi-stage process control. Production equipment typically includes multiple temperature-controlled zones, with heating power adjusted to maintain target temperatures in each zone. Current temperature control methods in graphene production primarily rely on real-time temperature data collected by temperature sensors, combined with preset temperature curves or traditional control algorithms to adjust the heating equipment and achieve stable temperature control during production. However, as production scales up, thermal coupling relationships arise between the multiple temperature-controlled zones due to heat conduction, heat radiation, and material transfer. Independent control of a single zone is insufficient to accurately reflect the thermal influence processes between these zones.

[0003] Existing model-based temperature control methods require the establishment of accurate thermodynamic models. The model parameters depend on the equipment structure, material properties, and production status. Changes in operating conditions, thermal response time delays, and inter-regional thermal influence relationships during graphene production can lead to a decrease in model prediction accuracy. Existing data-driven predictive control methods can use historical production data to predict temperature trends. However, in multi-region temperature control scenarios, historical data from different production stages and operating conditions differ. Directly combining historical trajectories can easily introduce mismatched data, affecting the accuracy of the prediction results.

[0004] Existing technologies lack effective methods for handling the time lag in thermal response across multiple temperature control zones. Changes in heating input and temperature response exist with time delays between multiple zones. When using synchronous data to build predictive models, it is difficult to reflect the directionality and delay characteristics of heat propagation, leading to insufficient coordination between control variables in different zones. Furthermore, existing control methods lack a trajectory update mechanism based on actual temperature response changes during execution, making it difficult to dynamically correct thermal response relationships and historical data contributions according to changes in production status.

[0005] Therefore, how to provide a multi-zone linkage temperature control method for graphene production based on edge computing is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] One objective of this invention is to propose a multi-zone coordinated temperature control method for graphene production based on edge computing. This invention fully utilizes edge computing, data-enabled predictive control, and multi-zone thermal coupling analysis technology to construct a multi-zone time-delay aligned Hankel matrix and optimize the trajectory combination process, thereby achieving coordinated adjustment of multiple temperature control zones during graphene production. It has the advantages of high temperature prediction accuracy, strong regional coordination, and good temperature control stability.

[0007] A multi-zone coordinated temperature control method for graphene production based on edge computing, according to an embodiment of the present invention, includes the following steps:

[0008] Temperature control data from each temperature control zone is collected using edge computing nodes and preprocessed to generate standardized multi-zone temperature control data.

[0009] Standardize multi-zone temperature control data by associating temperature control areas and control cycles to generate historical heating input trajectories, historical temperature response trajectories, and current control trajectories with production condition identifiers;

[0010] Based on the sequential relationship between historical heating input changes and historical temperature response changes, the directed thermal coupling relationship between temperature control zones and the regional thermal response time delay are determined;

[0011] Based on the directed thermal coupling relationship and the regional thermal response time delay, the historical heating input trajectory of the heat source region is time-shifted and aligned with the historical temperature response trajectory of the heated region to construct a multi-region time delay aligned Hankel matrix;

[0012] The weight of the historical trajectory is determined based on the degree of matching between the current control trajectory and the corresponding production conditions of the historical trajectory. The weight of the historical trajectory is used to constrain the DeePC trajectory combination coefficient to generate the weighted trajectory combination coefficient.

[0013] The multi-zone time-delay aligned Hankel matrix, the current control trajectory, and the weighted trajectory combination coefficients are input into the improved DeePC algorithm to generate the predicted temperature trajectory and candidate heating power trajectory for each temperature control zone.

[0014] With the goal of reducing the deviation between the predicted temperature trajectory and the target temperature trajectory and the temperature difference between the associated temperature control regions, based on the directed thermal coupling relationship, the linkage temperature control execution quantity of each temperature control region is solved collaboratively under the constraints of temperature safety boundary and heating power, and the heating power of the corresponding temperature control region is controlled synchronously.

[0015] The actual temperature response after collection and execution is used to generate temperature prediction residuals. Based on the temperature prediction residuals, the regional thermal response time delay and historical trajectory weights are corrected, and the multi-region time delay aligned Hankel matrix is ​​updated. The updated results are used to generate the linkage temperature control execution quantity for the next control cycle.

[0016] Optionally, the preprocessing procedure includes:

[0017] Collect original records of temperature, heating power and production conditions for each temperature control zone, and establish a corresponding relationship according to the zone identifier and the time of collection.

[0018] Perform integrity and uniqueness checks on the original records, and remove missing, duplicate, and abnormal data;

[0019] Divide the time window according to the preset control cycle and align all types of records to the corresponding control time.

[0020] Standardize data fields, units, and dimensions, and arrange them by region identifier, control time, temperature, power, and operating condition to generate standardized multi-zone temperature control data.

[0021] Optionally, the generation process of the historical heating input trajectory, historical temperature response trajectory, and current control trajectory includes:

[0022] Group the temperature control data from multiple zones by region identifier and sort them by control time.

[0023] Select data that are continuous in control time and consistent in production stage, and divide them into several historical trajectory windows according to a preset length;

[0024] Within each historical trajectory window, heating power and its changes are combined in chronological order to form a historical heating input trajectory.

[0025] Within each historical trajectory window, the temperature and its changes are combined in chronological order to form a historical temperature response trajectory.

[0026] The production stage and production condition of the corresponding window are used as identifiers and associated with the historical heating input trajectory and historical temperature response trajectory.

[0027] Select continuous control cycle data before the current control moment, and combine heating power, temperature and production conditions in chronological order to form the current control trajectory.

[0028] Optionally, the process for determining the directed thermal coupling relationship and the regional thermal response time delay includes:

[0029] Read the historical heating input trajectory of each temperature control zone, divide the part in which the power change reaches the preset threshold and the direction is consistent within the continuous control cycle into power change segments, and record the area, start and end time and change direction.

[0030] Using the region as the heat source region, the temperature response segment located after its start time and within the preset response window is retrieved from the historical temperature response trajectory of other regions. The segment whose temperature change reaches the preset threshold and is in the same direction is selected to determine the corresponding heated region.

[0031] The power change segment and the temperature response segment are matched according to the heat source area, the heated area and the direction of change. The number of control cycles between the two at the start time is calculated to obtain the candidate thermal response time delay.

[0032] Candidate thermal response time delays are summarized by combination of heat source region and heated region, abnormal matching results are eliminated, and the number of valid responses, the proportion of consistent directions and the degree of dispersion are statistically analyzed.

[0033] When the indicators meet the preset conditions, a directed thermal coupling relationship is established from the heat source region to the heated region, and the median value of the candidate thermal response time delay is used as the corresponding region's thermal response time delay.

[0034] Optionally, the process of constructing the multi-region time-delay aligned Hankel matrix includes:

[0035] Based on the regional thermal response time lag, the historical heating input trajectory of the heat source area is shifted backward to determine the arrival time of the thermal impact of each heating input record;

[0036] Extract the temperature response record of the heated area at the time of arrival of the heat effect, bind it with the translated heating input record, arrange it according to the time of arrival of the heat effect, and generate a cross-region time delay alignment trajectory.

[0037] For each heated area, the historical heating input trajectory, historical temperature response trajectory, and all cross-regional time-delay alignment trajectories pointing to this area are combined according to the same control time and fixed heat source area sequence to generate a multi-regional joint trajectory.

[0038] A sliding window is set according to the sum of the current control trajectory length and the preset prediction time domain length. Trajectory segments are extracted from the multi-zone joint trajectory for each control cycle. The first and last segments of each trajectory segment are arranged as historical intervals and prediction intervals, respectively. Each trajectory segment is used as a matrix column. The matrix columns are spliced ​​according to the heated area identifier to construct a multi-zone time-delay aligned Hankel matrix.

[0039] Optionally, the process of using historical trajectory weights to constrain the DeePC trajectory combination coefficients includes:

[0040] Read the historical production stage and historical operating status carried by the historical trajectory corresponding to each matrix column in the multi-zone time-delay aligned Hankel matrix, and perform consistency matching between the current production stage and the historical production stage.

[0041] For historical trajectories that are consistent with the production stage, the normalized difference between the corresponding fields in the current working condition and the historical working condition is calculated, the difference between each field is summarized to obtain the working condition difference, and the historical trajectories are arranged in order of working condition difference from smallest to largest.

[0042] According to the preset decreasing weight sequence, the historical trajectories are assigned weights from high to low in order of arrangement, and the historical trajectories with inconsistent production stages are assigned weights lower than those with consistent production stages.

[0043] Each historical trajectory weight is bound to the corresponding matrix column in the multi-zone time-delay aligned Hankel matrix and the DeePC trajectory combination coefficient;

[0044] The allowable range of values ​​for the corresponding DeePC trajectory combination coefficients is set based on the historical trajectory weights;

[0045] Under the condition that the current control trajectory is consistent with the historical interval of the multi-zone time delay alignment Hankel matrix, the DeePC trajectory combination coefficient constrained by the historical trajectory weight is solved to generate the weighted trajectory combination coefficient.

[0046] Optionally, the process of predicting the temperature trajectory and candidate heating power trajectory includes:

[0047] The multi-zone time-delay aligned Hankel matrix is ​​divided into historical intervals and prediction intervals;

[0048] The current control trajectory is matched with the historical intervals according to the temperature control area identifier and control time. The matrix columns of the historical intervals are combined using weighted trajectory combination coefficients to obtain the reconstruction result of the current control trajectory.

[0049] When the deviation between the reconstruction result and the current control trajectory does not exceed the preset trajectory deviation limit, the matrix columns of the prediction interval are combined using the weighted trajectory combination coefficient to obtain the temperature prediction sequence and heating power prediction sequence of each temperature control area in the preset prediction time domain.

[0050] The temperature prediction sequence and heating power prediction sequence are arranged according to the prediction control time and bound according to the temperature control area identifier to generate the predicted temperature trajectory and candidate heating power trajectory.

[0051] Optionally, the process of collaboratively solving the linkage temperature control execution quantities of each temperature control zone under the constraints of temperature safety boundaries and heating power includes:

[0052] Calculate the temperature deviation between the predicted temperature trajectory and the target temperature trajectory of each temperature control zone at each predicted control time.

[0053] Based on the directional thermal coupling relationship, the heat source region and the heated region are determined, the predicted temperature difference at the corresponding predictive control time is calculated, and the temperature deviation, the predicted temperature difference and the candidate heating power trajectory are associated.

[0054] When the increase in candidate heating power in the heat source region causes the predicted temperature of the heated region to rise, the candidate heating power of the heated region at the corresponding time is reduced; when the decrease in candidate heating power in the heat source region causes the predicted temperature of the heated region to fall, the candidate heating power of the heated region at the corresponding time is increased, thus obtaining the multi-zone coordinated heating power trajectory.

[0055] Temperature safety boundary verification is performed on the power trajectory of multi-zone coordinated heating, and the heating power that causes the predicted temperature of any temperature control zone to exceed the limit is corrected, thereby limiting the heating power of the relevant heat source zone.

[0056] The power range and the variation amplitude of adjacent control cycles are constrained and verified on the verified heating power trajectory to obtain the multi-zone linkage heating power trajectory that meets the constraints.

[0057] Extract the heating power adjustment amount of the current control cycle and generate the corresponding linkage temperature control execution amount according to the temperature control zone.

[0058] Optionally, the process of synchronously controlling the heating power of the corresponding temperature control zone includes:

[0059] Bind each linked temperature control execution quantity to the corresponding temperature control area identifier and the current control cycle;

[0060] At the control moment of the current control cycle, the execution quantities of each linkage temperature control are simultaneously sent to the corresponding temperature control area;

[0061] The heating power of the corresponding temperature control zone is adjusted according to the execution amount of each linkage temperature control, and the heating power of each temperature control zone is adjusted synchronously according to the multi-zone linkage heating power trajectory.

[0062] Optionally, the process of using the updated result for the next control cycle includes:

[0063] According to the temperature control zone identification and control time, the actual temperature response is matched with the predicted temperature trajectory, and the temperature prediction residual between the actual temperature and the predicted temperature is calculated.

[0064] When the occurrence time of the temperature prediction residual is inconsistent with the predicted temperature change time, the regional thermal response time delay corresponding to the directed thermal coupling relationship is corrected according to the control cycle difference between the two.

[0065] When the temperature prediction residual under the same production condition continuously exceeds the preset residual limit, the weight of the corresponding historical trajectory is reduced, and the weight of the historical trajectory that matches the actual temperature response is increased.

[0066] Based on the corrected regional thermal response time delay, the historical heating input trajectory is re-translated, and the heating input and actual temperature response of the current control cycle are written into the corresponding trajectory window, and the multi-region time delay aligned Hankel matrix is ​​updated;

[0067] The updated multi-zone time-delay aligned Hankel matrix and historical trajectory weights are used to generate the predicted temperature trajectory, candidate heating power trajectory, and linkage temperature control execution quantity for the next control cycle.

[0068] The beneficial effects of this invention are:

[0069] (1) This invention establishes a directed thermal coupling relationship and shifts the historical heating input trajectory of the heat source region according to the regional thermal response time delay, so that the heating power change and the temperature response of the heated region correspond in the time dimension, constructs a multi-region time delay aligned Hankel matrix, solves the problem of mismatch of predictive control data caused by the thermal propagation delay between multiple temperature control regions, and improves the accuracy of multi-region temperature prediction.

[0070] (2) This invention matches historical trajectories with the current production stage and working conditions to generate historical trajectory weights, and uses the historical trajectory weights to constrain the DeePC trajectory combination coefficients, so that historical trajectories with a high degree of matching with the current graphene production state are given priority to participate in temperature prediction and control solutions, thereby reducing prediction deviations caused by the mixing of data from different production stages and improving the adaptability of the temperature control process to changes in production conditions.

[0071] (3) This invention solves the candidate heating power trajectory of each temperature control region in a coordinated manner based on the directed thermal coupling relationship, and performs constraint verification by combining the temperature safety boundary, heating power range and power change amplitude, so as to realize the linkage adjustment of heating power in multiple temperature control regions. At the same time, it uses the temperature prediction residual to update the regional thermal response time delay and historical trajectory weight to form a closed-loop optimization control, thereby improving the temperature stability and temperature control reliability of the graphene production process. Attached Figure Description

[0072] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0073] Figure 1 This is a flowchart of a multi-zone coordinated temperature control method for graphene production based on edge computing proposed in this invention;

[0074] Figure 2 This is a schematic diagram of the DeePC predictive control process of a multi-zone linkage temperature control method for graphene production based on edge computing proposed in this invention.

[0075] Figure 3 This is a thermal map showing the temperature distribution of multiple temperature control zones in a multi-zone linkage temperature control method for graphene production based on edge computing proposed in this invention. Detailed Implementation

[0076] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0077] Example 1: Reference Figures 1-2 A multi-zone coordinated temperature control method for graphene production based on edge computing includes the following steps:

[0078] Temperature control data from each temperature control zone is collected using edge computing nodes and preprocessed to generate standardized multi-zone temperature control data.

[0079] Standardize multi-zone temperature control data by associating temperature control areas and control cycles to generate historical heating input trajectories, historical temperature response trajectories, and current control trajectories with production condition identifiers;

[0080] Based on the sequential relationship between historical heating input changes and historical temperature response changes, the directed thermal coupling relationship between temperature control zones and the regional thermal response time delay are determined;

[0081] Based on the directed thermal coupling relationship and the regional thermal response time delay, the historical heating input trajectory of the heat source region is time-shifted and aligned with the historical temperature response trajectory of the heated region to construct a multi-region time delay aligned Hankel matrix;

[0082] The weight of the historical trajectory is determined based on the degree of matching between the current control trajectory and the corresponding production conditions of the historical trajectory. The weight of the historical trajectory is used to constrain the DeePC trajectory combination coefficient to generate the weighted trajectory combination coefficient.

[0083] The multi-zone time-delay aligned Hankel matrix, the current control trajectory, and the weighted trajectory combination coefficients are input into the improved DeePC algorithm to generate the predicted temperature trajectory and candidate heating power trajectory for each temperature control zone.

[0084] With the goal of reducing the deviation between the predicted temperature trajectory and the target temperature trajectory and the temperature difference between the associated temperature control regions, based on the directed thermal coupling relationship, the linkage temperature control execution quantity of each temperature control region is solved collaboratively under the constraints of temperature safety boundary and heating power, and the heating power of the corresponding temperature control region is controlled synchronously.

[0085] The actual temperature response after collection and execution is used to generate temperature prediction residuals. Based on the temperature prediction residuals, the regional thermal response time delay and historical trajectory weights are corrected, and the multi-region time delay alignment Hankel matrix is ​​updated. The updated results are used to generate the linkage temperature control execution quantity for the next control cycle.

[0086] In this embodiment, the preprocessing process includes:

[0087] Collect raw records of temperature, heating power, and production conditions for each temperature control zone, and establish a correspondence based on zone identifier and collection time. Specifically, read the zone identifier and timestamp carried by each raw record, and bind raw records of temperature, heating power, and production conditions with the same zone identifier and collection time within the same preset time interval to the same data group. When the collection times of various raw records within the same data group are inconsistent, use the collection time of the temperature raw record as the reference collection time, and select the heating power raw record and production condition raw record with the smallest time difference from the reference collection time to establish a correspondence.

[0088] The original records are checked for completeness and uniqueness, and missing, duplicate, and abnormal data are removed. Each original record is examined for its region identifier, acquisition time, and corresponding acquisition value. Records with any necessary field empty and whose corresponding values ​​cannot be determined from adjacent records are marked as missing data and removed. A joint verification identifier is generated based on the region identifier, acquisition time, and record type. When two or more original records have the same joint verification identifier, the original record with complete acquisition fields and valid acquisition values ​​is retained, while other original records are marked as duplicate data and removed. Temperature, heating power, and production condition acquisition values ​​are compared with their corresponding preset acquisition ranges, and continuity is verified by considering the change amplitude at adjacent control times. Original records exceeding the preset acquisition range or whose adjacent change amplitude exceeds the preset change limit are marked as abnormal data and removed.

[0089] Divide the time window according to the preset control cycle and align all types of records to the corresponding control time.

[0090] Standardize data fields, units, and dimensions, and arrange them by region identifier, control time, temperature, power, and operating condition to generate standardized multi-zone temperature control data.

[0091] In this embodiment, the generation process of the historical heating input trajectory, the historical temperature response trajectory, and the current control trajectory includes:

[0092] Group the temperature control data from multiple zones by region identifier and sort them by control time.

[0093] The area identifier is a unique code pre-assigned to each temperature control area within the graphene production device. Each area identifier is bound to the temperature acquisition location and heating execution location of the corresponding temperature control area. When collecting temperature, heating power, and production condition raw records, the corresponding temperature control area identifier is written into the raw records. When grouping, the temperature control area identifier carried by the standardized multi-zone temperature control data is read one by one. Data with the same temperature control area identifier are grouped into the same data group. Within each data group, the data are arranged from earliest to latest according to the control time. The temperature, heating power, and production condition corresponding to the same control time are bound as area control records.

[0094] Select data that are continuous in control time and consistent in production stage, and divide them into several historical trajectory windows according to a preset length;

[0095] Within each historical trajectory window, heating power and its changes are combined in chronological order to form a historical heating input trajectory.

[0096] Read the heating power corresponding to each control moment within the historical trajectory window and arrange them from earliest to latest according to the control moment; subtract the heating power of the previous control moment from the heating power of the current control moment to obtain the heating power change corresponding to the current control moment; calculate the difference between the heating power at the start of the window and the heating power before the immediately preceding start of the window to obtain the heating power change corresponding to the start of the window; when there is no heating power before the immediately preceding start of the window, set the heating power change corresponding to the start of the window to zero; bind the heating power and heating power change of the same control moment as a heating input record, and arrange the heating input records sequentially according to the control moment to form a historical heating input trajectory;

[0097] Within each historical trajectory window, the temperature and its changes are combined in chronological order to form a historical temperature response trajectory.

[0098] Read the temperature corresponding to each control moment within the historical trajectory window and arrange them from earliest to latest. Subtract the temperature of the previous control moment from the temperature of the current control moment to obtain the temperature change corresponding to the current control moment. Calculate the difference between the temperature at the start of the window and the temperature immediately preceding the start of the window's control moment to obtain the temperature change corresponding to the start of the window's control moment. If there is no temperature immediately preceding the start of the window's control moment, set the temperature change corresponding to the start of the window's control moment to zero. Bind the temperature and temperature change of the same control moment into a temperature response record, and arrange the temperature response records sequentially according to the control moment to form a historical temperature response trajectory.

[0099] The production stage and production condition of the corresponding window are used as identifiers and associated with the historical heating input trajectory and historical temperature response trajectory.

[0100] Select continuous control cycle data before the current control moment, and combine heating power, temperature and production conditions in chronological order to form the current control trajectory.

[0101] In this embodiment, the process of determining the directed thermal coupling relationship and the regional thermal response time delay includes:

[0102] Read the historical heating input trajectory of each temperature control zone, divide the part in which the power change reaches the preset threshold and the direction is consistent within the continuous control cycle into power change segments, and record the area, start and end time and change direction.

[0103] The difference between the heating power at the current control time and the heating power at the previous control time is calculated sequentially according to the control time. A difference greater than zero is determined as the direction of power increase, and a difference less than zero is determined as the direction of power decrease. Starting from the control time when the absolute value of the first power change reaches the preset power change threshold, the power change of subsequent control times is checked sequentially. When the power change of consecutive control times all reach the preset power change threshold and maintain the same direction of change, the corresponding trajectory parts are merged into the same power change segment. When the power change is lower than the preset power change threshold, the direction of power change changes, or the historical heating input trajectory is interrupted, the current power change segment ends, the first control time that meets the conditions is taken as the start time, and the last control time that meets the conditions before the end is taken as the end time.

[0104] Using the region as the heat source region, retrieve the temperature response segment located after its start time and within the preset response window from the historical temperature response trajectory of other regions, filter the segment whose temperature change reaches the preset threshold and whose direction is consistent, and determine the corresponding heated region.

[0105] The power change segment and the temperature response segment are matched according to the heat source area, the heated area and the direction of change. The number of control cycles between the two at the start time is calculated to obtain the candidate thermal response time delay.

[0106] A thermal response retrieval record is established for each power change segment. The temperature control area to which the power change segment belongs is written into the heat source area field, and the temperature control area to which the retrieved temperature response segment belongs is written into the heated area field. The change direction of the power change segment and the temperature response segment are compared. When the power increases and the temperature increases, and the power decreases and the temperature decreases, the change direction is determined to be matched. The start time of the temperature response segment is further checked to see if it is later than the start time of the power change segment, and whether the interval between the two is within the preset response window. When both the change direction and the time sequence meet the conditions, the two segments are bound into a matching record. When the same power change segment corresponds to more than two temperature response segments that meet the conditions in the same heated area, the temperature response segment with the earliest start time is selected to establish a match. If the start times are the same, the temperature response segment with the larger temperature change amplitude is selected. The control cycle difference between the start time of the power change segment and the start time of the temperature response segment in the matching record is calculated, and the obtained control cycle number is written into the candidate thermal response time delay field.

[0107] Candidate thermal response time delays are summarized by combination of heat source region and heated region, abnormal matching results are eliminated, and the number of valid responses, the proportion of consistent directions and the degree of dispersion are statistically analyzed.

[0108] The heat source region identifier and the heated region identifier are combined to form a region combination identifier. All matching records are read, and matching records with the same region combination identifier are grouped into the same thermal response record group, arranged sequentially according to the start time of the power change segment. Within each thermal response record group, the candidate thermal response delays corresponding to each matching record are summarized into a candidate thermal response delay sequence, retaining the power change direction and temperature change direction corresponding to each candidate thermal response delay. The candidate thermal response delay sequence is verified, and matching records whose temperature response segment start time is no later than the power change segment start time, matching records whose candidate thermal response delay exceeds the preset response window, matching records whose power change direction is inconsistent with the temperature change direction, matching records where the heat source region and the heated region are the same, and matching records repeatedly generated from the same power change segment and the same temperature response segment are removed. The matching records retained after removing abnormal matching results are determined as valid matching results. The number of valid matching results is counted to obtain the number of valid responses. The proportion of valid matching results with consistent change directions to the number of valid responses is calculated to obtain the direction consistency ratio. The degree of dispersion is determined based on the deviation of each candidate thermal response delay from the median value of the candidate thermal response delay.

[0109] When the indicators meet the preset conditions, a directed thermal coupling relationship is established from the heat source region to the heated region, and the median value of the candidate thermal response time delay is used as the corresponding region's thermal response time delay.

[0110] That is, when the number of effective responses reaches the preset minimum number of responses, the proportion of consistent directions reaches the preset proportion of consistent directions, and the dispersion of candidate thermal response time delay does not exceed the preset dispersion limit, a directed thermal coupling relationship is established from the heat source region to the heated region, and the median value of the candidate thermal response time delay corresponding to the effective matching result is used as the thermal response time delay of the corresponding region.

[0111] In this embodiment, the process of constructing a multi-region time-delay aligned Hankel matrix includes:

[0112] Read the starting temperature control region identifier, the pointing temperature control region identifier, and the regional thermal response time delay carried by the directed thermal coupling relationship. Determine the temperature control region corresponding to the starting temperature control region identifier as the heat source region. Determine the temperature control region corresponding to the pointing temperature control region identifier as the heated region. Establish a one-to-one correspondence record between the heat source region, the heated region, and the regional thermal response time delay.

[0113] Based on the regional thermal response time lag, the historical heating input trajectory of the heat source area is shifted backward to determine the arrival time of the thermal impact of each heating input record;

[0114] Keeping the heating power and heating power change in the historical heating input trajectory unchanged, the original control time of each heating input record is increased by the number of control cycles contained in the corresponding region's thermal response time delay to obtain the arrival time of the thermal effect; the backward shift only changes the time corresponding position of the heating input record, without changing the heating power and heating power change carried by the heating input record;

[0115] Extract the temperature response record of the heated area at the time of arrival of the heat effect, bind it with the translated heating input record, arrange it according to the time of arrival of the heat effect, and generate a cross-region time delay alignment trajectory.

[0116] Specifically, using the unified control time after preprocessing as the matching benchmark, temperature response records whose control time is equal to the arrival time of the heat effect are retrieved from the historical temperature response trajectory of the heated area. The translated heating input record, the temperature of the heated area, and the temperature change of the heated area are bound into a cross-regional time-delay alignment record. Heating input records that do not have a corresponding temperature response record or that exceed the historical trajectory window after translation are not written. The retained cross-regional time-delay alignment records are arranged in ascending order according to the arrival time of the heat effect to form a cross-regional time-delay alignment trajectory from the heat source area to the heated area.

[0117] For each heated area, the historical heating input trajectory, historical temperature response trajectory, and all cross-regional time-delay alignment trajectories pointing to this area are combined according to the same control time and fixed heat source area sequence to generate a multi-regional joint trajectory.

[0118] Specifically, the process involves selecting control times from the historical temperature response trajectories of the heated areas one by one, extracting the heating input record, temperature response record, and cross-regional time delay alignment record from each cross-regional time delay alignment trajectory corresponding to that control time; setting fixed arrangement positions for heat source areas pointing to the same heated area according to the temperature control area identifier in ascending order, and maintaining these arrangement positions unchanged at different control times; combining the heating input record, temperature response record, and cross-regional time delay alignment record of each heat source area in the order of the local area heating input record, local area temperature response record, and cross-regional time delay alignment record of each heat source area into a multi-regional joint record, and then arranging the multi-regional joint record in ascending order according to the control time to generate a multi-regional joint trajectory for the corresponding heated area; and using only continuous multi-regional joint records with corresponding records at each predetermined arrangement position for matrix construction.

[0119] Set a sliding window according to the sum of the current control trajectory length and the preset prediction time domain length, extract trajectory segments from the multi-zone joint trajectory for each control cycle, arrange the first and last segments of each trajectory segment as historical intervals and prediction intervals respectively, treat each trajectory segment as a matrix column, and splice the matrix columns according to the heated area identifier to construct a multi-zone time delay aligned Hankel matrix.

[0120] Specifically, the current control trajectory length is added to the preset prediction time domain length to obtain the trajectory segment length. Starting from the first control moment of the multi-zone joint trajectory, continuous multi-zone joint records reaching the trajectory segment length are extracted to form the first trajectory segment. The starting control moment of the subsequent trajectory segment is shifted one control cycle relative to the previous trajectory segment until the remaining multi-zone joint records are insufficient to reach the trajectory segment length. Within each trajectory segment, the multi-zone joint records that have reached the current control trajectory length are arranged according to the control moment to form a historical interval, and the multi-zone joint records that have reached the preset prediction time domain length are arranged according to the control moment to form a prediction interval. The historical intervals and prediction intervals are arranged vertically to form a matrix column. Then, the matrix columns corresponding to different trajectory segments are arranged horizontally according to the starting control moment to construct a multi-zone time delay aligned Hankel matrix. The same row position in the matrix corresponds to the same temperature control area, the same data type, and the same relative control moment, and each column in the matrix corresponds to a continuous trajectory segment.

[0121] In this embodiment, the process of constraining the DeePC trajectory combination coefficients using historical trajectory weights includes:

[0122] Extract the current production stage and current operating status from the production condition identifier carried by the current control trajectory;

[0123] Read the historical production stage and historical operating status carried by the historical trajectory corresponding to each column of the multi-zone time-delay aligned Hankel matrix, and perform consistency matching between the current production stage and the historical production stage; specifically including:

[0124] The system reads the production stage identifier corresponding to the current control trajectory, and sequentially reads the historical production stage identifiers corresponding to each column of the multi-zone time-delay aligned Hankel matrix. When the current production stage identifier is the same as the historical production stage identifier, the corresponding historical trajectory is marked as a production stage consistent trajectory. When the current production stage identifier is different from the historical production stage identifier, the corresponding historical trajectory is marked as a production stage inconsistent trajectory. The production stage identifier is pre-written according to the stage division results of the graphene production process when the standardized multi-zone temperature control data is generated. Only data with the same production stage identifier are retained in the same historical trajectory window.

[0125] For historical trajectories that are consistent with the production stage, the normalized difference between the corresponding fields in the current working condition and the historical working condition is calculated, the difference between each field is summarized to obtain the working condition difference, and the historical trajectories are arranged in order of working condition difference from smallest to largest.

[0126] Specifically, the same fields in the current operating condition and the historical operating condition are mapped to a unified numerical range. The absolute difference between the normalized values ​​of each corresponding field is calculated. Then, the absolute difference is weighted and summed according to the preset operating condition weights of each field to obtain the operating condition difference between the current control trajectory and the corresponding historical trajectory. The smaller the operating condition difference, the closer the production operating conditions of the corresponding historical trajectory and the current control trajectory are. When the operating condition differences of two historical trajectories are the same, they are arranged from the nearest to the furthest starting control time of the historical trajectory.

[0127] According to the preset decreasing weight sequence, the historical trajectories are assigned weights from high to low in order of arrangement, and the historical trajectories with inconsistent production stages are assigned weights lower than those with consistent production stages.

[0128] The preset decreasing weight sequence is a numerical sequence pre-stored within the edge computing nodes that monotonically decreases as the historical trajectory sorting position increases; let the number of consistent trajectories in the production stage be... The sorting position is Then the first The weight of each historical trajectory is determined according to the following formula: ;

[0129] In the formula, Indicates the sorting position is The historical trajectory weights corresponding to the historical trajectories; This indicates the number of consistent trajectories during the production phase; This indicates the sorting position of the corresponding historical trajectories after being arranged in ascending order of working condition differences; the first historical trajectory in the sorting position receives the maximum historical trajectory weight, and the last historical trajectory in the sorting position receives the minimum historical trajectory weight. The sum of the historical trajectory weights corresponding to the consistent trajectories in each production stage is 1; the historical trajectory weights of inconsistent trajectories in the production stage are set to a preset low weight value that is less than the minimum historical trajectory weight of the consistent trajectories in the production stage, so that inconsistent trajectories in the production stage only participate in the solution when the consistent trajectories in the production stage are insufficient to reconstruct the current control trajectory.

[0130] Each historical trajectory weight is bound to the corresponding matrix column in the multi-zone time-delay aligned Hankel matrix and the DeePC trajectory combination coefficient;

[0131] The DeePC trajectory combination coefficient is the coefficient used in the DeePC algorithm to combine the columns of the multi-zone time-delay aligned Hankel matrix. Each DeePC trajectory combination coefficient corresponds to a column of the multi-zone time-delay aligned Hankel matrix. Each column of the matrix is ​​multiplied by the corresponding DeePC trajectory combination coefficient and then accumulated to reconstruct the current control trajectory and generate a predicted trajectory that is continuous with the current control trajectory.

[0132] The allowable range of values ​​for the corresponding DeePC trajectory combination coefficient is set based on the historical trajectory weight, so that the higher the historical trajectory weight, the larger the allowable range of values ​​for the corresponding DeePC trajectory combination coefficient.

[0133] Specifically, the normalized historical trajectory weights are obtained by dividing each historical trajectory weight by the maximum value among all historical trajectory weights: ;

[0134] In the formula, Indicates the first The weight of each historical trajectory; Indicates the first Normalized historical trajectory weights corresponding to each historical trajectory; This indicates the number of historical trajectories involved in the trajectory combination; the normalized historical trajectory weights are multiplied by the preset upper limit of the combination coefficients to obtain the upper limit of the absolute value of the corresponding DeePC trajectory combination coefficients, and the allowed value range is set according to the following formula:

[0135] ;

[0136] In the formula, Indicates the first The DeePC trajectory combination coefficients corresponding to each historical trajectory; This indicates the upper limit of the preset combination coefficients; Indicates the first The normalized historical trajectory weights corresponding to each historical trajectory; thus, historical trajectories with a high degree of working condition matching have a wider range of allowed values ​​for combination coefficients, while historical trajectories with a low degree of working condition matching have a narrower range of allowed values ​​for combination coefficients.

[0137] Under the condition that the current control trajectory is consistent with the historical interval of the multi-zone time delay alignment Hankel matrix, the DeePC trajectory combination coefficient constrained by the historical trajectory weight is solved to generate the weighted trajectory combination coefficient.

[0138] The trajectory consistency condition is that after weighting and superimposing the historical intervals of each column of the multi-zone time-delay aligned Hankel matrix according to the DeePC trajectory combination coefficient, the resulting historical heating input sequence corresponds to the heating input sequence in the current control trajectory according to the control time, and the resulting historical temperature response sequence corresponds to the temperature response sequence in the current control trajectory according to the control time, and the reconstruction deviation of the historical heating input sequence and the historical temperature response sequence does not exceed the preset trajectory deviation limit.

[0139] The solution objective is to minimize the reconstruction deviation between the current control trajectory and the historical intervals obtained by weighted superposition. Simultaneously, a significant penalty is applied to the DeePC trajectory combination coefficients, which have lower weights in the historical trajectories. The solution is obtained using the following formula: ;

[0140] satisfy: ;

[0141] In the formula, This represents the historical interval of a multi-zone time-delay aligned Hankel matrix; Indicates the current control trajectory; This represents the coefficient vector composed of the combination coefficients of each DeePC trajectory; Indicates the first The DeePC trajectory combination coefficients corresponding to each historical trajectory; This indicates the number of historical trajectories involved in the trajectory combination; Indicates the constraint strength of the combination coefficients; Indicates the first The weight of each historical trajectory; This represents a preset positive number to prevent the denominator from being zero; the calculated DeePC trajectory combination coefficients form the weighted trajectory combination coefficients;

[0142] In a specific example, after arranging the three historical trajectories in ascending order of working condition differences, the corresponding normalized historical trajectory weights are as follows: When the preset upper limit of the combination coefficients is satisfied

[0143] At that time, the allowed value ranges for the DeePC trajectory combination coefficients corresponding to the three historical trajectories are as follows:

[0144] ;

[0145] ;

[0146] ;

[0147] Under the condition of satisfying the current control trajectory reconstruction deviation limit, the solution is as follows:

[0148] ;

[0149] Therefore, the first historical trajectory, which is closest to the current production conditions, bears the main combined contribution, while the third historical trajectory, which differs significantly from the current production conditions, bears only a smaller combined contribution.

[0150] In this embodiment, the process of predicting the temperature trajectory and candidate heating power trajectory includes:

[0151] The multi-zone time-delay aligned Hankel matrix is ​​divided into historical intervals and prediction intervals;

[0152] The current control trajectory length and preset prediction time domain length used when constructing the multi-zone time-delay aligned Hankel matrix are read. According to the relative control time of the multi-zone joint records in each matrix column, the matrix rows corresponding to the multi-zone joint records located at the beginning and with a number equal to the current control trajectory length are assigned to the historical interval, and the matrix rows corresponding to the multi-zone joint records located at the end and with a number equal to the preset prediction time domain length are assigned to the prediction interval. The historical interval and the prediction interval maintain the same number of matrix columns. The historical interval and the prediction interval of each matrix column originate from the same continuous trajectory segment, and the arrangement position of each temperature control zone, data type, and relative control time in different matrix columns remains consistent.

[0153] The current control trajectory is mapped to the historical interval according to the temperature control area identifier and control time. The matrix columns of the historical interval are combined using weighted trajectory combination coefficients to obtain the reconstruction result of the current control trajectory. According to the temperature control area identifier, data type and relative control time, the heating input record and temperature response record in the current control trajectory are mapped to fixed matrix rows in the historical interval. Each matrix column of the historical interval is multiplied by the weighted trajectory combination coefficient corresponding to that matrix column, and the products are accumulated according to the same row position to obtain the reconstruction result with the same data arrangement order and trajectory length as the current control trajectory.

[0154] When the deviation between the reconstruction result and the current control trajectory does not exceed the preset trajectory deviation limit, the matrix columns of the prediction interval are combined using weighted trajectory combination coefficients to obtain the temperature prediction sequence and heating power prediction sequence of each temperature control area within the preset prediction time domain. Specifically, keeping the weighted trajectory combination coefficients and matrix column correspondences used in the historical interval reconstruction process unchanged, each matrix column of the prediction interval is multiplied by the corresponding weighted trajectory combination coefficient, and the product of each matrix column is accumulated according to the same row position to obtain the prediction interval combination result. According to the preset temperature control area identifier, data type, and relative control time arrangement position of the prediction interval, the temperature record and heating power record of each temperature control area are extracted from the prediction interval combination result. The temperature records of the same temperature control area are arranged from near to far according to the relative control time to generate the temperature prediction sequence, and the heating power records of the same temperature control area are arranged from near to far according to the relative control time to generate the heating power prediction sequence.

[0155] Arrange the temperature prediction sequence and heating power prediction sequence according to the prediction control time and bind them according to the temperature control area identifier to generate the predicted temperature trajectory and candidate heating power trajectory; take the first control cycle after the current control time as the first prediction control time, and configure the prediction control time for each prediction value in the temperature prediction sequence and heating power prediction sequence in turn; establish a correspondence between the temperature prediction value with the same temperature control area identifier and the same prediction control time and the heating power prediction value, and form the predicted temperature trajectory and candidate heating power trajectory for each temperature control area respectively.

[0156] In this embodiment, the improved DeePC algorithm is based on a data-enabled predictive control algorithm. Taking into account the characteristics of heat propagation direction, thermal response time delay, and changes in production conditions among multiple temperature-controlled zones in the graphene production process, the algorithm improves the trajectory data construction method, trajectory combination method, and multi-zone control solution method. The improved DeePC algorithm mainly consists of the following parts.

[0157] The multi-zone control trajectory input structure and the improved DeePC algorithm receive historical heating input trajectories, historical temperature response trajectories, current control trajectories, and production condition identifiers for each temperature control zone.

[0158] Historical heating input trajectory is used to characterize the heating power and heating power change process of each temperature control zone in a continuous control cycle; historical temperature response trajectory is used to characterize the temperature and temperature change process of the corresponding temperature control zone under the action of heating input; current control trajectory is used to characterize the actual heating input and temperature response in a continuous control cycle before the current control moment; production condition identifier is used to distinguish the production stage and production condition corresponding to the historical trajectory and the current control trajectory.

[0159] The multi-zone time-delay aligned Hankel matrix structure and the improved DeePC algorithm determine the heat source region and the heated region based on the directed thermal coupling relationship. According to the regional thermal response time delay, the historical heating input trajectory of the heat source region is shifted backward along the control time, so that the shifted heating input record and the delayed temperature response record corresponding to the heated region are at the same control time.

[0160] The historical heating input trajectory of the heated area, the historical temperature response trajectory of the heated area, and the cross-regional time-delay alignment trajectory corresponding to each heat source area are combined to form a multi-regional joint trajectory. Then, the multi-regional joint trajectory is slidably truncated according to the current control trajectory length and the preset prediction time domain length, and each trajectory segment is arranged into a matrix column to construct a multi-regional time-delay alignment Hankel matrix.

[0161] The multi-zone time-delay aligned Hankel matrix is ​​divided into historical intervals and prediction intervals. The historical interval is used to reconstruct the current control trajectory, and the prediction interval is used to generate future temperature trajectories and candidate heating power trajectories.

[0162] The production condition weighted trajectory combination structure uses the improved DeePC algorithm to read the current production stage and current operating condition state corresponding to the current control trajectory, and matches them with the historical production stage and historical operating condition state corresponding to each historical trajectory in the multi-zone time delay alignment Hankel matrix.

[0163] The weight of each historical trajectory is determined based on the consistency of production stages and the differences in operating conditions. Historical trajectories that are close to the current production conditions receive higher weights, while those that differ significantly from the current production conditions receive lower weights.

[0164] Each historical trajectory weight is bound to the corresponding matrix column in the multi-zone time-delay aligned Hankel matrix and the DeePC trajectory combination coefficient. The historical trajectory weight is used to limit the allowable contribution of the corresponding DeePC trajectory combination coefficient, so that historical trajectories with a high degree of matching with production conditions participate in the reconstruction of the current control trajectory and the prediction of future trajectories first.

[0165] The improved DeePC algorithm uses weighted trajectory combination coefficients to weight the matrix columns of the historical intervals of the multi-zone time-delay aligned Hankel matrix to obtain the reconstruction result of the current control trajectory.

[0166] When the deviation between the reconstruction result and the current control trajectory does not exceed the preset trajectory deviation limit, the weighted trajectory combination coefficients and the correspondence of the matrix columns remain unchanged, and the matrix columns of the prediction interval are weighted and combined.

[0167] From the combined results of the predicted intervals, the predicted temperature value and the predicted heating power value are extracted according to the temperature control area identifier and the predicted control time, and the predicted temperature trajectory and candidate heating power trajectory of each temperature control area are generated respectively.

[0168] The multi-zone linkage control solution structure and the improved DeePC algorithm aim to reduce the deviation between the predicted temperature trajectory and the target temperature trajectory of each temperature control zone. At the same time, the predicted temperature difference between the associated temperature control zones is calculated based on the directional thermal coupling relationship.

[0169] Under the constraints of temperature safety boundary, regional temperature difference limit, heating power limit and heating power change limit, the candidate heating power trajectory of each temperature control region is solved collaboratively to generate the linkage temperature control execution quantity for the corresponding control cycle.

[0170] When changes in the heating power of the heat source region will cause changes in the predicted temperature of the heated region, the improved DeePC algorithm synchronously corrects the candidate heating power trajectory of the heated region, so that the heating power adjustment of each temperature control region maintains a thermally coupled and coordinated relationship.

[0171] The rolling feedback update structure allows edge computing nodes to send the linked temperature control execution data to the corresponding temperature control area and collect the actual temperature response after execution.

[0172] The actual temperature response is compared with the predicted temperature trajectory to generate a temperature prediction residual. The regional thermal response time delay and historical trajectory weights are corrected based on the temperature prediction residual. Cross-regional trajectory alignment is then performed again based on the corrected regional thermal response time delay, and the multi-regional time delay alignment Hankel matrix is ​​updated.

[0173] The updated multi-zone time-delay aligned Hankel matrix and historical trajectory weights are used to generate the predicted temperature trajectory and solve the linkage temperature control execution quantity for the next control cycle, thereby forming a rolling closed-loop control.

[0174] In summary, the improved DeePC algorithm retains the basic structure of the DeePC algorithm for state reconstruction and predictive control based on historical input-output trajectories. It achieves coordinated temperature control of multiple temperature control zones in the graphene production process through multi-zone thermal response time delay alignment, weighted trajectory combination under production conditions, and collaborative solution under directed thermal coupling constraints.

[0175] In this embodiment, the process of collaboratively solving the linkage temperature control execution quantities of each temperature control zone under the constraints of temperature safety boundaries and heating power includes:

[0176] Calculate the temperature deviation between the predicted temperature trajectory and the target temperature trajectory of each temperature control zone at each predicted control time. Specifically, according to the temperature control zone identifier and the predicted control time, match the predicted temperature value in the predicted temperature trajectory with the target temperature value in the target temperature trajectory one by one. Subtract the corresponding predicted temperature value from the target temperature value to obtain the temperature deviation at the corresponding predicted control time. Temperature deviations greater than zero are defined as under-temperature deviations, and temperature deviations less than zero are defined as over-temperature deviations. Arrange the temperature deviations according to the predicted control time to form a temperature deviation sequence for each temperature control zone.

[0177] Based on the directed thermal coupling relationship, the heat source region and the heated region are determined. The predicted temperature difference at the corresponding predictive control time is calculated, and the temperature deviation, predicted temperature difference, and candidate heating power trajectory are associated. The heat source region, heated region, and regional thermal response time delay corresponding to each directed thermal coupling relationship are read. A correspondence is established between the current predicted control time of the heat source region and the predicted control time of the heated region after the regional thermal response time delay. The predicted temperature values ​​of the two corresponding predicted control times are extracted and the difference is calculated. The absolute value of the difference is determined as the predicted temperature difference, while retaining the direction of the difference where the predicted temperature of the heat source region is higher or lower than the predicted temperature of the heated region. The candidate heating power of the heat source region, the candidate heating power of the heated region, the temperature deviation of the heated region, and the predicted temperature difference are written into the same associated record according to the directed thermal coupling relationship and the corresponding predictive control time.

[0178] When the increase in candidate heating power in the heat source region causes the predicted temperature of the heated region to rise, the candidate heating power of the heated region at the corresponding time is reduced; when the decrease in candidate heating power in the heat source region causes the predicted temperature of the heated region to fall, the candidate heating power of the heated region at the corresponding time is increased, thus obtaining the multi-zone coordinated heating power trajectory.

[0179] Calculate the change in candidate heating power between adjacent predicted control times in the heat source region, and calculate the change in predicted temperature of the heated region before and after the arrival of the corresponding heat effect. When the change in candidate heating power is greater than zero and the change in predicted temperature of the heated region is greater than zero, it is determined that the power increase of the heat source region has a temperature-raising effect on the heated region. Based on the excess between the predicted temperature of the heated region and the target temperature, the candidate heating power of the heated region at the corresponding predicted control time is gradually reduced. After each reduction, the candidate heating power trajectory is recombined and the predicted temperature trajectory is calculated until the predicted temperature of the heated region returns to the allowable range of the target temperature, or the candidate heating power of the heated region reaches the preset minimum heating power.

[0180] When the change in candidate heating power between adjacent predicted control times in the heat source region is less than zero, and the change in predicted temperature of the heated region at the corresponding heat effect arrival time is less than zero, it is determined that the power reduction in the heat source region has a cooling effect on the heated region. Based on the under-temperature between the target temperature and the predicted temperature of the heated region, the candidate heating power of the heated region at the corresponding predicted control time is gradually increased. After each increase, the predicted temperature trajectory of the heated region and the associated temperature control region is recalculated until the predicted temperature of the heated region returns to the target temperature allowable range, or the candidate heating power of the heated region reaches the preset maximum heating power. The candidate heating power trajectories of each temperature control region after the reduction and increase processing are combined according to the temperature control region identifier to obtain the multi-zone collaborative heating power trajectory.

[0181] Temperature safety boundary verification is performed on the power trajectory of multi-zone coordinated heating, and the heating power that causes the predicted temperature of any temperature control zone to exceed the limit is corrected, thereby limiting the heating power of the relevant heat source zone.

[0182] Specifically, the predicted temperature values ​​of each temperature control region are compared with the corresponding upper and lower safety boundaries of the temperature control system according to the predicted control time to determine the first temperature control region where the predicted temperature exceeds the boundary and the predicted control time. When the predicted temperature is higher than the upper safety boundary, the candidate heating power of the temperature control region exceeding the boundary before the corresponding predicted control time is first reduced. Based on the directed thermal coupling relationship pointing to the temperature control region exceeding the boundary, the heat source region corresponding to the arrival time of the thermal influence and the predicted control time of the boundary is found. The candidate heating power of the relevant heat source region is reduced in descending order of the increase in predicted temperature caused by the corresponding thermal influence. When the predicted temperature is lower than the lower safety boundary, the candidate heating power of the temperature control region exceeding the boundary is increased without causing the predicted temperature of the associated temperature control region to exceed the upper safety boundary. After each heating power correction, the predicted temperature trajectory of each temperature control region is regenerated and the temperature safety boundary comparison is performed until all predicted temperatures are within the temperature safety boundary.

[0183] The power range and the variation amplitude of adjacent control cycles are constrained and verified on the verified heating power trajectory to obtain the multi-zone linkage heating power trajectory that meets the constraints.

[0184] The candidate heating power of each temperature control zone at each predictive control time is compared with the corresponding preset minimum heating power and preset maximum heating power. When the candidate heating power is lower than the preset minimum heating power, the candidate heating power is corrected to the preset minimum heating power. When the candidate heating power is higher than the preset maximum heating power, the candidate heating power is corrected to the preset maximum heating power. Using the actual heating power of the current control cycle as the first comparison benchmark, the heating power change amplitude between adjacent predictive control times is calculated sequentially. When the heating power change amplitude exceeds the preset power change limit, the heating power of the next predictive control time is corrected to the sum or difference of the heating power of the previous control time and the preset power change limit. After completing the power range and change amplitude correction, the temperature safety boundary is re-verified. The trajectory that simultaneously meets the temperature safety boundary, heating power range, and power change amplitude constraints is determined as the multi-zone linkage heating power trajectory.

[0185] Extract the heating power adjustment amount for the current control cycle and generate corresponding linkage temperature control execution amounts according to the temperature control zone. The linkage temperature control execution amount is the heating power adjustment instruction for the current control cycle extracted from the multi-zone linkage heating power trajectory, and consists of the temperature control zone identifier, the current control cycle, and the corresponding heating power adjustment amount; the heating power adjustment amount is the difference between the target heating power and the current actual heating power for the current control cycle, and is used to synchronously increase, decrease, or maintain the heating power of the corresponding temperature control zone.

[0186] In this embodiment, the process of synchronously controlling the heating power of the corresponding temperature control zone includes:

[0187] Bind each linked temperature control execution quantity to the corresponding temperature control area identifier and the current control cycle;

[0188] At the control moment of the current control cycle, the execution quantities of each linkage temperature control are simultaneously sent to the corresponding temperature control area;

[0189] The heating power of the corresponding temperature control zone is adjusted according to the execution amount of each linkage temperature control, and the heating power of each temperature control zone is adjusted synchronously according to the multi-zone linkage heating power trajectory.

[0190] In this embodiment, the process of using the update result for the next control cycle includes:

[0191] According to the temperature control zone identification and control time, the actual temperature response is matched with the predicted temperature trajectory, and the temperature prediction residual between the actual temperature and the predicted temperature is calculated.

[0192] When the occurrence time of the temperature prediction residual is inconsistent with the predicted temperature change time, the corresponding regional thermal response time lag of the directed thermal coupling relationship is corrected according to the control cycle difference between the two; the time of change of heating power in the heat source region that caused the predicted temperature change is read, and the actual response time when the actual temperature of the heated region begins to change in the same direction is determined; the number of control cycles between the actual response time and the time of change of heating power in the heat source region is calculated to obtain the observed thermal response time lag; the observed thermal response time lag is compared with the original regional thermal response time lag, and when the actual temperature response is later than the predicted temperature change, the original regional thermal response time lag is increased by the corresponding control cycle difference; when the actual temperature response is earlier than the predicted temperature change, the original regional thermal response time lag is decreased by the corresponding control cycle difference, and the correction result is written into the directed thermal coupling relationship from the heat source region to the heated region.

[0193] When the temperature prediction residual under the same production condition continuously exceeds the preset residual limit, the weight of the corresponding historical trajectory is reduced, and the weight of the historical trajectory that matches the actual temperature response is increased. The historical trajectories involved in generating the current predicted temperature trajectory and their corresponding weighted trajectory combination coefficients are determined. Historical trajectories with non-zero weighted trajectory combination coefficients and whose predicted temperature change direction is inconsistent with the actual temperature change direction are identified as deviation historical trajectories. The weight of the historical trajectory corresponding to the deviation historical trajectory is reduced according to the preset weight adjustment range. Among the historical trajectories consistent with the production stage, the differences between each historical temperature response trajectory and the actual temperature response in terms of change direction, change magnitude, and response time are compared. Historical trajectories with smaller differences and not exceeding the preset matching limit are identified as matching historical trajectories. The weight of the historical trajectory corresponding to the matching historical trajectory is increased according to the preset weight adjustment range. After the adjustment is completed, the weights of all historical trajectories are normalized to maintain a uniform value scale.

[0194] Based on the corrected regional thermal response time delay, the historical heating input trajectory is re-translated, and the heating input and actual temperature response of the current control cycle are written into the corresponding trajectory window, and the multi-region time delay aligned Hankel matrix is ​​updated;

[0195] Specifically, the arrival time of the thermal impact of each heating input record in the heat source area is redetermined according to the corrected regional thermal response time delay, and the shifted heating input record is re-bound with the temperature response record of the heated area at the corresponding arrival time of the thermal impact, thus updating the cross-regional time delay alignment trajectory. The actual heating input, actual temperature response, and production condition identifier of the current control cycle are appended to the end of the historical trajectory of the corresponding temperature control area, and the trajectory segment is re-slidably extracted according to the sum of the current control trajectory length and the preset prediction time domain length. The first and second segments of each trajectory segment are arranged as historical intervals and prediction intervals, respectively, and the matrix columns are arranged according to the start control time, replacing the matrix columns in the original multi-regional time delay alignment Hankel matrix affected by the regional thermal response time delay correction, thus forming the updated multi-regional time delay alignment Hankel matrix.

[0196] The updated multi-zone time-delay aligned Hankel matrix and historical trajectory weights are used to generate the predicted temperature trajectory, candidate heating power trajectory, and linkage temperature control execution quantity for the next control cycle.

[0197] Example 2: To verify the feasibility of this invention in practice, it was applied to a multi-temperature zone heating and temperature control scenario in a graphene production process. This production process employs a continuous heating structure, with the production device divided into multiple temperature control zones along the heat transfer direction. Each temperature control zone is equipped with a temperature acquisition device and a heating execution device to achieve temperature detection and heating power adjustment in different zones. In actual production, heat transfer influences exist between different temperature control zones. Changes in the heating power of the front-end zone will affect the temperature state of the back-end zone after a certain response time, and temperature changes in each zone are not completely synchronized. When using a traditional independent temperature control method, each temperature control zone only adjusts its heating based on its own real-time temperature, without considering the heat propagation path and thermal response lag between zones. This easily leads to problems such as excessively rapid heating in local areas, temperature lag in adjacent areas, and uneven overall temperature distribution.

[0198] This embodiment employs the edge computing-based multi-zone coordinated temperature control method for graphene production proposed in this invention. During the production process, edge computing nodes collect real-time temperature data, heating power data, and production status data for each temperature control zone. The edge computing nodes perform field standardization, time synchronization, and anomaly data filtering on the collected data, associating data from different temperature control zones within the same control cycle to form standardized multi-zone temperature control data. Data from different temperature control zones is distinguished by zone identifiers, ensuring that each temperature control zone corresponds to independent temperature response records and heating input records.

[0199] During continuous control, edge computing nodes organize standardized multi-zone temperature control data according to temperature control areas and control cycles, constructing historical heating input trajectories, historical temperature response trajectories, and current control trajectories. The historical heating input trajectory records the heating power changes in each temperature control area within the continuous control cycle, while the historical temperature response trajectory records the temperature changes under the corresponding heating action. The current control trajectory describes the recent temperature and heating power changes in each area under the current production state, providing input for subsequent predictive control.

[0200] To address the issue of inter-regional thermal impact delay in graphene production, this embodiment analyzes the heat propagation relationship between different temperature control zones based on the temporal relationship between historical heating input changes and historical temperature response changes. By identifying the response time of temperature changes in other temperature control zones after a change in heating power in one zone, the direction of heat propagation and the regional thermal response time lag are determined. For example, if an increase in heating power in one zone causes a temperature change in the same direction in adjacent zones after several control cycles, a directional thermal coupling relationship is determined between the two zones, and the corresponding regional thermal response time lag is recorded.

[0201] After determining the regional thermal response time lag, the historical heating input trajectory of the heat source region is time-shifted according to the regional thermal response time lag, so that the heating power change time corresponds to the temperature change time of the heated region. The shifted cross-regional heating input data is then correlated with the temperature response data of the heated region to form a cross-regional time-lag aligned trajectory. Furthermore, by combining the local input / output trajectories of each temperature control region and the cross-regional time-lag aligned trajectory, a multi-regional time-lag aligned Hankel matrix is ​​constructed, ensuring that the historical trajectory data includes the heat propagation relationships between different regions.

[0202] During predictive control, edge computing nodes read the production stage and operating status corresponding to the current control trajectory, match the current production status with the corresponding production status of historical trajectories, and generate historical trajectory weights based on the degree of matching. Historical trajectories with production states close to the current control state receive higher weights, while historical trajectories with significantly different production states have reduced participation. These historical trajectory weights are used to constrain the DeePC trajectory combination coefficients, causing the improved DeePC algorithm to prioritize historical trajectories close to the current production state for prediction.

[0203] The improved DeePC algorithm predicts the temperature change trend of each temperature-controlled region within the future control cycle based on the multi-zone time-delay aligned Hankel matrix, the current control trajectory, and the weighted trajectory combination coefficients, while simultaneously generating candidate heating power trajectories. The prediction process comprehensively considers the thermal coupling effects between different temperature-controlled regions, ensuring that the prediction results not only reflect changes in heating input within the region but also the impact of heat propagation on temperature changes in associated regions.

[0204] In this embodiment, the candidate heating power trajectories of each temperature-controlled region are collaboratively optimized based on the deviation between the predicted temperature trajectory and the target temperature trajectory, as well as the predicted temperature difference between associated temperature-controlled regions. When an increase in heating power of a certain region causes a temperature rise in an associated region, the heating power of the heated region in the corresponding control cycle is adjusted downward; when a power decrease in the heat source region causes a temperature drop in the heated region, the heating power of the heated region in the corresponding control cycle is increased. After completing inter-region power coordination, temperature safety boundary check and power change constraint check are performed on the adjusted heating power trajectories, and linked temperature control execution quantities corresponding to each temperature-controlled region are generated.

[0205] According to the generated linked temperature control execution quantities, the heating equipment of each temperature-controlled region is adjusted synchronously, so that multiple temperature-controlled regions perform heating power changes in accordance with the collaborative calculation results. During the execution process, the actual temperature response is continuously collected, and compared with the predicted temperature trajectory to calculate the temperature prediction residual. When there is a deviation between the actual temperature change and the predicted result, the regional thermal response time lag and historical trajectory weight are corrected according to the temperature prediction residual, and the multi-region time-lag aligned Hankel matrix is re-updated, so that the temperature prediction and control results of the next control cycle are more consistent with the current production state.

[0206] In this embodiment, continuous control is performed on multiple temperature-controlled regions through the above method, generating a Figure 3 temperature distribution heatmap of multiple temperature-controlled regions as shown. It can be seen from the heat distribution result that each temperature-controlled region maintains a continuous temperature change trend during the control process, there is no obvious abrupt change region between regions, heat is gradually transferred along the preset direction, and the temperature distribution of each region is more uniform. The temperature gradient change of different regions in the heatmap is continuous, which shows that the present invention can effectively utilize the regional thermal coupling relationship to coordinately control the temperature change between different temperature-controlled regions.

[0207] Figure 3 is a temperature distribution heatmap of multiple temperature-controlled regions in an embodiment of the present invention, which is used to show the spatial temperature distribution state of different temperature-controlled regions inside the graphene production device at a certain control moment after adopting the multi-region linked temperature control method based on edge computing.

[0208] Figure 3 The horizontal direction represents the axial position of the furnace body, the production device is spatially divided along the heat transfer direction, and different positions correspond to different temperature-controlled regions; the vertical direction represents the number of the temperature-controlled region, which is used to distinguish different heating control regions inside the device. Figure 3 The color change in represents the temperature of the corresponding position, the color scale on the right indicates the temperature range, and the color change from dark to light indicates that the temperature gradually increases.

[0209] In this embodiment, each temperature control zone corresponds to an independent temperature acquisition location and heating execution location. The edge computing node analyzes the heat propagation relationship between zones based on the acquired temperature data, heating power data, and production condition data, and aligns the heating input of different zones according to the zone's thermal response time lag. An improved DeePC algorithm is used to predict the future temperature change trend of each zone, and the heating power of each zone is adjusted in conjunction with the directed thermal coupling relationship to keep the temperature changes of different zones coordinated.

[0210] Depend on Figure 3 As can be seen, after multi-zone coordinated temperature control, the internal temperature of the furnace exhibits a continuous gradient distribution, with gentle temperature changes along the furnace's axial direction and no obvious temperature abrupt changes between different temperature control zones. The temperature response is faster near the heat source area, while areas farther from the heat source gradually reach their corresponding temperature states after a thermal response time lag. This indicates that the present invention can identify the heat transfer process between zones and perform coordinated control based on the thermal response delay relationship.

[0211] Figure 3 This invention reflects the spatial temperature control effect after determining the directional thermal coupling relationship, constructing the multi-region time-delay aligned Hankel matrix, and solving the linkage temperature control execution quantity. By uniformly mapping the temperature states of different temperature control regions to a spatial distribution map, the temperature consistency and heat transfer continuity between multiple regions can be intuitively observed, verifying that the invention can reduce temperature deviations between regions and improve the overall temperature uniformity in the graphene production process.

[0212] To further verify the control effect of this invention, a traditional independent temperature control method was used as a comparison. Under the same production conditions, the temperature changes, temperature control deviations, maximum temperature differences between regions, and energy consumption changes in multiple temperature control zones were recorded. During the experiment, multiple temperature control zones were set up for continuous operation, and the control results were statistically analyzed over multiple control cycles. Test results show that this invention, through multi-zone time-delay aligned Hankel matrices combined with an improved DeePC algorithm, achieves the fusion of regional heat propagation information, making the temperature prediction results closer to the actual response process.

[0213] Experimental data show that, after adopting the method of this invention, the root mean square error of the average temperature in each temperature control zone decreased from 6.82℃ in the traditional method to 2.14℃, significantly reducing temperature prediction deviation; the maximum temperature difference between zones decreased from 21.6℃ in the traditional method to 8.5℃, improving temperature consistency between different temperature control zones; the number of times the temperature exceeded the target range decreased from 34 times to 6 times, reducing the impact of temperature fluctuations on the production process; and the unit energy consumption index decreased from 1.00 to 0.87, reducing energy consumption caused by repeated heating and overcompensation. Specific control effect data are shown in Table 1.

[0214] Table 1: Comparison of Temperature Stability and Energy Consumption Optimization Effects Based on Multi-Temperature Control Zone Interlocking Control

[0215] Evaluation indicators Traditional independent temperature control method The present invention provides a linkage temperature control method. Optimization effect Root mean square error of average temperature (°C) 6.82 2.14 Reduced by 68.6% Maximum temperature difference in the region (°C) 21.6 8.5 Reduced by 60.6% Number of times the temperature exceeded the target range (times / 1000 control cycles) 34 6 Reduced by 82.4% Average lag period (period) of temperature response 18 7 Reduced by 61.1% Thermal response prediction bias (°C) 8.3 2.7 Reduced by 67.5% Unit energy consumption index 1.00 0.87 Decrease by 13.0% Multi-region temperature consistency evaluation value 0.82 0.96 An increase of 17.1%

[0216] As shown in Table 1, the multi-zone linkage temperature control method for graphene production based on edge computing proposed in this invention has significant improvements in temperature control accuracy, regional temperature consistency, thermal response prediction capability, and energy utilization efficiency.

[0217] From the perspective of temperature control accuracy, the traditional independent temperature control method fails to effectively consider the heat transfer effects between different temperature control zones, and historical control data cannot reflect the thermal response time lag between zones. This results in a significant deviation between the predicted temperature and the actual temperature response, with an average root mean square error (RMSE) of 6.82℃. This invention establishes a directed thermal coupling relationship and aligns the historical heating input trajectory with the regional thermal response time lag to construct a multi-zone time-lag aligned Hankel matrix. This allows the predictive control process to integrate heat propagation information between different temperature control zones. Combined with the improved DeePC algorithm for temperature trajectory prediction, the average RMSE is reduced to 2.14℃, a reduction of 68.6%, demonstrating that this invention can improve the accuracy of temperature prediction and control across multiple temperature control zones.

[0218] From the perspective of temperature uniformity, traditional independent temperature control methods adjust the heating power of each region separately, without considering the continuous influence of the heat source region on the heated region. This results in significant temperature differences between different temperature control regions, with the maximum regional temperature difference reaching 21.6℃. This invention utilizes the regional thermal coupling relationship to collaboratively solve the heating power of multiple temperature control regions. Based on the predicted temperature trajectory and the temperature difference between associated regions, the heating power is dynamically adjusted, enabling coordinated control of the heat transfer process. The maximum regional temperature difference is reduced to 8.5℃, a reduction of 60.6%, and the multi-region temperature consistency evaluation value is improved from 0.82 to 0.96, thus enhancing the overall temperature distribution uniformity during graphene production.

[0219] From the perspective of dynamic response capability, traditional control methods mainly rely on the temperature feedback of the current region for adjustment. When heat is transferred from one temperature-controlled region to an associated region, due to the thermal response delay, the control action cannot reflect the temperature changes between regions in a timely manner, with an average temperature response lag period of 18 control cycles. This invention uses a multi-zone time-delay aligned Hankel matrix to time-match the heating input changes of the heat source region with the temperature response of the heated region. It uses temperature prediction residual feedback to correct the regional thermal response lag, enabling the control model to dynamically adjust the prediction relationship according to actual temperature changes. The average temperature response lag period is reduced to 7 control cycles, a reduction of 61.1%, and the thermal response prediction deviation is reduced from 8.3℃ to 2.7℃, improving the system's adaptability to multi-zone thermal dynamic changes.

[0220] From the perspectives of production stability and energy utilization, traditional control methods, lacking a multi-regional coordination mechanism, require frequent adjustments to heating power when the temperature deviates from the target range, easily leading to overheating or repeated compensation, resulting in the temperature exceeding the target range 34 times. This invention generates historical trajectory weights by matching production conditions and uses these weights to constrain the DeePC trajectory combination coefficients, allowing historical trajectories with a high degree of matching to the current production state to participate in predictive control. Simultaneously, by combining temperature safety boundaries and heating power constraints to generate linked temperature control execution quantities, the number of times the temperature exceeds the target range is reduced to 6 times, a reduction of 82.4%. The unit energy consumption index decreases from 1.00 to 0.87, a reduction of 13.0%, indicating that this invention improves temperature control stability while reducing ineffective heating processes and improving energy utilization efficiency.

[0221] The above data analysis verifies that the present invention achieves coordinated control of multiple temperature control zones in the graphene production process through edge computing, multi-zone thermal coupling analysis, time-delay alignment trajectory construction, operating condition weighted DeePC predictive control, and residual feedback update mechanism. This effectively improves the problems of large temperature prediction deviation, significant regional temperature difference, lag in thermal response, and high energy consumption in traditional multi-zone temperature control processes.

[0222] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A multi-zone coordinated temperature control method for graphene production based on edge computing, characterized in that, Includes the following steps: Temperature control data from each temperature control zone is collected using edge computing nodes and preprocessed to generate standardized multi-zone temperature control data. Standardize multi-zone temperature control data by associating temperature control areas and control cycles to generate historical heating input trajectories, historical temperature response trajectories, and current control trajectories with production condition identifiers; Based on the sequential relationship between historical heating input changes and historical temperature response changes, the directed thermal coupling relationship between temperature control zones and the regional thermal response time delay are determined; Based on the directed thermal coupling relationship and the regional thermal response time delay, the historical heating input trajectory of the heat source region is time-shifted and aligned with the historical temperature response trajectory of the heated region to construct a multi-region time delay aligned Hankel matrix; The weight of the historical trajectory is determined based on the degree of matching between the current control trajectory and the corresponding production conditions of the historical trajectory. The weight of the historical trajectory is used to constrain the DeePC trajectory combination coefficient to generate the weighted trajectory combination coefficient. The multi-zone time-delay aligned Hankel matrix, the current control trajectory, and the weighted trajectory combination coefficients are input into the improved DeePC algorithm to generate the predicted temperature trajectory and candidate heating power trajectory for each temperature control zone. With the goal of reducing the deviation between the predicted temperature trajectory and the target temperature trajectory and the temperature difference between the associated temperature control regions, based on the directed thermal coupling relationship, the linkage temperature control execution quantity of each temperature control region is solved collaboratively under the constraints of temperature safety boundary and heating power, and the heating power of the corresponding temperature control region is controlled synchronously. The actual temperature response after collection and execution is used to generate temperature prediction residuals. Based on the temperature prediction residuals, the regional thermal response time delay and historical trajectory weights are corrected, and the multi-region time delay aligned Hankel matrix is ​​updated. The updated results are used to generate the linkage temperature control execution quantity for the next control cycle.

2. The multi-zone coordinated temperature control method for graphene production based on edge computing according to claim 1, characterized in that, The preprocessing process includes: Collect original records of temperature, heating power and production conditions for each temperature control zone, and establish a corresponding relationship according to the zone identifier and the time of collection. Perform integrity and uniqueness checks on the original records, and remove missing, duplicate, and abnormal data; Divide the time window according to the preset control cycle and align all types of records to the corresponding control time. Standardize data fields, units, and dimensions, and arrange them by region identifier, control time, temperature, power, and operating condition to generate standardized multi-zone temperature control data.

3. The multi-zone coordinated temperature control method for graphene production based on edge computing according to claim 2, characterized in that, The generation process of the historical heating input trajectory, historical temperature response trajectory, and current control trajectory includes: Group the temperature control data from multiple zones by region identifier and sort them by control time. Select data that are continuous in control time and consistent in production stage, and divide them into several historical trajectory windows according to a preset length; Within each historical trajectory window, heating power and its changes are combined in chronological order to form a historical heating input trajectory. Within each historical trajectory window, the temperature and its changes are combined in chronological order to form a historical temperature response trajectory. The production stage and production condition of the corresponding window are used as identifiers and associated with the historical heating input trajectory and historical temperature response trajectory. Select continuous control cycle data before the current control moment, and combine heating power, temperature and production conditions in chronological order to form the current control trajectory.

4. The multi-zone coordinated temperature control method for graphene production based on edge computing according to claim 3, characterized in that, The process of determining the directed thermal coupling relationship and the regional thermal response time delay includes: Read the historical heating input trajectory of each temperature control zone, divide the part in which the power change reaches the preset threshold and the direction is consistent within the continuous control cycle into power change segments, and record the area, start and end time and change direction. Using the region as the heat source region, the temperature response segment located after its start time and within the preset response window is retrieved from the historical temperature response trajectory of other regions. The segment whose temperature change reaches the preset threshold and is in the same direction is selected to determine the corresponding heated region. The power change segment and the temperature response segment are matched according to the heat source area, the heated area and the direction of change. The number of control cycles between the two at the start time is calculated to obtain the candidate thermal response time delay. Candidate thermal response time delays are summarized by combination of heat source region and heated region, abnormal matching results are eliminated, and the number of valid responses, the proportion of consistent directions and the degree of dispersion are statistically analyzed. When the indicators meet the preset conditions, a directed thermal coupling relationship is established from the heat source region to the heated region, and the median value of the candidate thermal response time delay is used as the corresponding region's thermal response time delay.

5. The multi-zone coordinated temperature control method for graphene production based on edge computing according to claim 4, characterized in that, The process of constructing the multi-region time-delay aligned Hankel matrix includes: Based on the regional thermal response time lag, the historical heating input trajectory of the heat source area is shifted backward to determine the arrival time of the thermal impact of each heating input record; Extract the temperature response record of the heated area at the time of arrival of the heat effect, bind it with the translated heating input record, arrange it according to the time of arrival of the heat effect, and generate a cross-region time delay alignment trajectory. For each heated area, the historical heating input trajectory, historical temperature response trajectory, and all cross-regional time-delay alignment trajectories pointing to this area are combined according to the same control time and fixed heat source area sequence to generate a multi-regional joint trajectory. A sliding window is set according to the sum of the current control trajectory length and the preset prediction time domain length. Trajectory segments are extracted from the multi-zone joint trajectory for each control cycle. The first and last segments of each trajectory segment are arranged as historical intervals and prediction intervals, respectively. Each trajectory segment is used as a matrix column. The matrix columns are spliced ​​according to the heated area identifier to construct a multi-zone time-delay aligned Hankel matrix.

6. The multi-zone coordinated temperature control method for graphene production based on edge computing according to claim 5, characterized in that, The process of using historical trajectory weights to constrain the DeePC trajectory combination coefficients includes: Read the historical production stage and historical operating status carried by the historical trajectory corresponding to each matrix column in the multi-zone time-delay aligned Hankel matrix, and perform consistency matching between the current production stage and the historical production stage. For historical trajectories that are consistent with the production stage, the normalized difference between the corresponding fields in the current working condition and the historical working condition is calculated, the difference between each field is summarized to obtain the working condition difference, and the historical trajectories are arranged in order of working condition difference from smallest to largest. According to the preset decreasing weight sequence, the historical trajectories are assigned weights from high to low in order of arrangement, and the historical trajectories with inconsistent production stages are assigned weights lower than those with consistent production stages. Each historical trajectory weight is bound to the corresponding matrix column in the multi-zone time-delay aligned Hankel matrix and the DeePC trajectory combination coefficient; The allowable range of values ​​for the corresponding DeePC trajectory combination coefficients is set based on the historical trajectory weights; Under the condition that the current control trajectory is consistent with the historical interval of the multi-zone time delay alignment Hankel matrix, the DeePC trajectory combination coefficient constrained by the historical trajectory weight is solved to generate the weighted trajectory combination coefficient.

7. The multi-zone coordinated temperature control method for graphene production based on edge computing according to claim 6, characterized in that, The process of predicting the temperature trajectory and candidate heating power trajectory includes: The multi-zone time-delay aligned Hankel matrix is ​​divided into historical intervals and prediction intervals; The current control trajectory is matched with the historical intervals according to the temperature control area identifier and control time. The matrix columns of the historical intervals are combined using weighted trajectory combination coefficients to obtain the reconstruction result of the current control trajectory. When the deviation between the reconstruction result and the current control trajectory does not exceed the preset trajectory deviation limit, the matrix columns of the prediction interval are combined using the weighted trajectory combination coefficient to obtain the temperature prediction sequence and heating power prediction sequence of each temperature control area in the preset prediction time domain. The temperature prediction sequence and heating power prediction sequence are arranged according to the prediction control time and bound according to the temperature control area identifier to generate the predicted temperature trajectory and candidate heating power trajectory.

8. The multi-zone coordinated temperature control method for graphene production based on edge computing according to claim 7, characterized in that, The process of collaboratively solving the linkage temperature control execution quantities of each temperature control zone under the constraints of temperature safety boundary and heating power includes: Calculate the temperature deviation between the predicted temperature trajectory and the target temperature trajectory of each temperature control zone at each predicted control time. Based on the directional thermal coupling relationship, the heat source region and the heated region are determined, the predicted temperature difference at the corresponding predictive control time is calculated, and the temperature deviation, the predicted temperature difference and the candidate heating power trajectory are associated. When the increase in candidate heating power in the heat source region causes the predicted temperature of the heated region to rise, the candidate heating power of the heated region at the corresponding time is reduced; when the decrease in candidate heating power in the heat source region causes the predicted temperature of the heated region to fall, the candidate heating power of the heated region at the corresponding time is increased, thus obtaining the multi-zone coordinated heating power trajectory. Temperature safety boundary verification is performed on the power trajectory of multi-zone coordinated heating, and the heating power that causes the predicted temperature of any temperature control zone to exceed the limit is corrected, thereby limiting the heating power of the relevant heat source zone. The power range and the variation amplitude of adjacent control cycles are constrained and verified on the verified heating power trajectory to obtain the multi-zone linkage heating power trajectory that meets the constraints. Extract the heating power adjustment amount of the current control cycle and generate the corresponding linkage temperature control execution amount according to the temperature control zone.

9. A multi-zone coordinated temperature control method for graphene production based on edge computing according to claim 8, characterized in that, The process of synchronously controlling the heating power of the corresponding temperature control zone includes: Bind each linked temperature control execution quantity to the corresponding temperature control area identifier and the current control cycle; At the control moment of the current control cycle, the execution quantities of each linkage temperature control are simultaneously sent to the corresponding temperature control area; The heating power of the corresponding temperature control zone is adjusted according to the execution amount of each linkage temperature control, and the heating power of each temperature control zone is adjusted synchronously according to the multi-zone linkage heating power trajectory.

10. A multi-zone coordinated temperature control method for graphene production based on edge computing according to claim 9, characterized in that, The process of using the updated results for the next control cycle includes: According to the temperature control zone identification and control time, the actual temperature response is matched with the predicted temperature trajectory, and the temperature prediction residual between the actual temperature and the predicted temperature is calculated. When the occurrence time of the temperature prediction residual is inconsistent with the predicted temperature change time, the regional thermal response time delay corresponding to the directed thermal coupling relationship is corrected according to the control cycle difference between the two. When the temperature prediction residual under the same production condition continuously exceeds the preset residual limit, the weight of the corresponding historical trajectory is reduced, and the weight of the historical trajectory that matches the actual temperature response is increased. Based on the corrected regional thermal response time delay, the historical heating input trajectory is re-translated, and the heating input and actual temperature response of the current control cycle are written into the corresponding trajectory window, and the multi-region time delay aligned Hankel matrix is ​​updated; The updated multi-zone time-delay aligned Hankel matrix and historical trajectory weights are used to generate the predicted temperature trajectory, candidate heating power trajectory, and linkage temperature control execution quantity for the next control cycle.

Images