Multivariable adaptive control method for chemical reaction kettle

Abstract

The application discloses a multivariable adaptive control method for a chemical reaction kettle, and relates to the technical field of chemical industry control. The method comprises the following steps: S1, collecting the thermal field data of the reaction kettle in real time, and performing data preprocessing; S2, performing spatial interpolation completion on the temperature of the non-distributed point area in the reaction kettle, constructing a continuous thermal field of the reaction kettle, and performing abnormality checking; S3, dividing the thermal field space partition, evaluating the uniformity of the thermal field space partition temperature, identifying abnormal partitions, and implementing partition control measures; S4, quantifying the control amount of each thermal field space partition, generating an adjustment instruction and delivering the adjustment instruction to the corresponding controller; and S5, mapping the continuous thermal field of the reaction kettle, the thermal field space partition and the adjustment instruction to a digital platform, and performing real-time monitoring, abnormal diagnosis and self-recovery optimization on the thermal field state of the reaction kettle. The method solves the problems of uneven thermal field and local overheating risk of the chemical reaction kettle and the difficulty of realizing overall thermal field accurate control by traditional monitoring and PID.

Description

Technical Field

[0001] This invention relates to the field of chemical industry control technology, specifically a multivariable adaptive control method for chemical reactors. Background Technology

[0002] With the rapid development of modern chemical, pharmaceutical, materials, and energy industries, chemical reactors, as key equipment for realizing core processes such as material synthesis, reaction, purification, and mixing, have been widely used in various process industrial production processes. Reactors typically handle complex chemical reaction processes involving high temperature and pressure, strong exothermic or endothermic reactions, complex compositions, and dramatic phase transitions, placing higher demands on the precise control of parameters such as temperature, pressure, and stirring. At the same time, the advancement of national strategies such as green chemistry, energy conservation and emission reduction, and intelligent manufacturing has set stricter standards for the safety, stability, energy efficiency, and automation and intelligence levels of chemical production processes.

[0003] For example, invention patent CN118034220A discloses a process orchestration control system based on chemical experimental reactor equipment. The system includes an industrial IoT subsystem, a low-code subsystem, and a chemical experiment management subsystem. The industrial IoT subsystem includes a data storage module for acquiring and storing standard monitoring data corresponding to the current chemical experimental reactor equipment. The low-code subsystem includes a development tool module for providing low-code development tools and performing process orchestration data analysis on the standard monitoring data to obtain process orchestration data analysis results. The chemical experiment management subsystem includes an experimental data analysis application module for matching target experimental data analysis applications corresponding to the standard monitoring data based on the process orchestration data analysis results, thereby enabling process orchestration control of the current chemical experimental reactor equipment based on the target experimental data analysis applications. This achieves efficient process orchestration control of the chemical experimental reactor equipment.

[0004] For example, invention patent CN117289595A discloses a sliding mode-fuzzy PID composite control method applied to chemical reactors, belonging to the field of industrial control technology. The method described in this invention solves the problem of significant lag in traditional chemical reactor temperature control systems. This invention proposes a fuzzy PID-based control strategy and combines it with sliding mode control to achieve accurate temperature tracking of the reactor. The method is characterized by its insensitivity to external disturbances and fast output response speed. The use of a fuzzy controller to control the sliding mode results in better control performance. Finally, a chemical reactor temperature control system based on fuzzy sliding mode control was developed using MATLAB, and the effectiveness of the proposed control algorithm was verified through simulation and experiments.

[0005] However, chemical reactors suffer from uneven heat distribution, local hot spots that can easily lead to side reactions and material damage, and limited point temperature monitoring, making it difficult for ordinary PID controllers to control the overall thermal field distribution of chemical reactors.

[0006] Therefore, a multivariable adaptive control method for chemical reactors is urgently needed to address the above problems. Summary of the Invention

[0007] Technical problems to be solved

[0008] To address the shortcomings of existing technologies, this invention provides a multivariable adaptive control method for chemical reactors, which solves the problem that traditional monitoring and PID control methods are insufficient to achieve precise overall thermal field control in chemical reactors due to the risk of uneven thermal field and local overheating.

[0009] Technical solution

[0010] To achieve the above objectives, the present invention provides the following technical solution: a multivariable adaptive control method for a chemical reactor, comprising the following steps: S1, real-time acquisition of multi-point temperature values, three-dimensional spatial coordinates of spatial points, and stirring speed of the chemical reactor to construct reactor thermal field data, and preprocessing the reactor thermal field data; S2, based on the multi-point temperature values ​​and three-dimensional spatial coordinates, spatial interpolation is performed to complete the temperature of the uncontrolled areas within the reactor, and the completed temperature is combined to construct a continuous thermal field of the reactor, and anomaly verification is performed on the continuous thermal field of the reactor; S3, the continuous thermal field of the reactor is further processed... Cluster analysis is performed to divide the thermal field space into partitions. The temperature uniformity of the thermal field space partitions is evaluated using multi-point temperature values ​​within each partition. Abnormal partitions are identified based on the temperature uniformity of the thermal field space partitions, and partition control measures are implemented. In step S4, the control amount of each thermal field space partition is quantified using multi-point temperature values ​​within each partition. Temperature and stirring speed adjustment commands are generated based on the adjustment amounts and sent to the corresponding controllers. In step S5, the continuous thermal field of the reactor, the thermal field space partitions, and the adjustment commands are mapped to the digital platform in real time for real-time monitoring, anomaly diagnosis, and self-healing optimization of the reactor's thermal field status.

[0011] Furthermore, the thermal field data of the chemical reactor is constructed by real-time acquisition of multiple temperature values, three-dimensional spatial coordinates of spatial points, and stirring speed. The specific process of data preprocessing for the reactor thermal field data is as follows: a high-density temperature sensor array is uniformly arranged at multiple key physical locations of the reactor to collect the temperature values ​​of each measuring point in real time, and the three-dimensional spatial coordinates of each spatial point are obtained by combining the spatial structural parameters of the reactor. At the same time, the stirring speed is collected in real time through the output of the frequency converter. The three-dimensional spatial coordinates of each spatial point, the real-time temperature value, and the real-time stirring speed are combined to construct the reactor thermal field data. For the original reactor thermal field data, a Gaussian filtering denoising algorithm is used to reduce signal noise, and data synchronization is achieved through global timestamp alignment and linear interpolation. Outliers are detected and removed using the 3σ discriminant method, and missing data is filled in using temporal nearest neighbor interpolation. At the same time, the reactor thermal field data is standardized and normalized to establish a reactor thermal field control database. The original and preprocessed reactor thermal field data are stored in the reactor thermal field control database with timestamps.

[0012] Furthermore, based on multi-point temperature values ​​and three-dimensional spatial coordinates, the specific process of spatial interpolation to complete the temperature of the un-measured area within the reactor is as follows: For target point i without sensor placement, all N measurement points j in the reactor are sequentially traversed; for each pair of target point i and measurement point j, the three-dimensional spatial coordinates of target point i and measurement point j and the temperature value of measurement point j are obtained, and the spatial distance value is obtained by calculating the Euclidean distance between the three-dimensional spatial coordinates; the spatial distance attenuation value is obtained by dividing the square of the spatial distance value by twice the square of the spatial weighting factor and taking the negative value as the exponent; the standard factor of the reactor thermal field distribution is obtained by taking the square root of the product of the constant 2, the constant pi, and the spatial weighting factor and taking the reciprocal; the standard factor of the reactor thermal field distribution, the spatial distance attenuation value, and the temperature value of measurement point j are multiplied to obtain the weighted temperature contribution value, and the weighted temperature contribution values ​​of all measurement points j are accumulated to obtain the interpolated temperature value of target point i.

[0013] Furthermore, the specific process of constructing a continuous thermal field for the reactor based on the completed temperature and performing anomaly verification on the continuous thermal field is as follows: The interpolated temperature values ​​of the target points requiring interpolation in the reactor space are calculated in real time; temperature completion is performed on the unmarked areas in the three-dimensional space of the reactor; and the continuous thermal field of the reactor is constructed by combining the three-dimensional spatial coordinates and temperature values ​​of each spatial point after temperature completion. Simultaneously, for fluctuations in temperature values ​​in different regions of the continuous thermal field, an adaptive algorithm based on local sampling point density is used to adjust the spatial weighting factor. The spatial continuity and physical consistency of the continuous thermal field are checked. If local abnormal jumps or interpolation anomalies are found, multi-point spatial neighborhood interpolation is used for repair. That is, with the abnormal point as the center, the nearest N normal measurement points and interpolation points are selected, the interpolated temperature values ​​are recalculated, and the new values ​​replace the abnormal data. The continuous thermal field of the reactor is written into the reactor thermal field control database and output.

[0014] Furthermore, the specific process of performing cluster analysis on the continuous thermal field of the reactor and dividing the thermal field into spatial partitions is as follows: receiving the continuous thermal field of the reactor, and using a clustering algorithm based on temperature similarity and spatial proximity to perform cluster analysis on all spatial points according to the temperature values ​​and three-dimensional spatial coordinates of each spatial point in the continuous thermal field of the reactor, identifying regions with different temperature distributions of spatial points in the continuous thermal field of the reactor, and dynamically dividing the thermal field into N spatial partitions according to the cluster analysis results, and recording the number of spatial points and corresponding temperature values ​​in each thermal field spatial partition in real time.

[0015] Furthermore, the specific process of evaluating the temperature uniformity of a thermal field spatial partition using multi-point temperature values ​​within the partition is as follows: For each spatial point i within the thermal field spatial partition, acquire the temperature value and calculate the average value to obtain the partition temperature mean. Calculate the absolute difference between the temperature value of spatial point i and the partition temperature mean. Take the negative of the absolute difference as the exponent and perform natural exponentiation to obtain the partition temperature deviation value of spatial point i. Sum the partition temperature deviation values ​​of all spatial points within the thermal field spatial partition to obtain the total partition temperature deviation. Divide the partition temperature deviation value of spatial point i by the total partition temperature deviation to obtain the temperature probability weight term of spatial point i. Perform natural logarithmic operation on the temperature probability weight term of spatial point i to obtain the temperature weight logarithm term of spatial point i. Multiply the temperature probability weight term of spatial point i by the temperature weight logarithm term of spatial point i to obtain the temperature information entropy weighted value of spatial point i. Sum the temperature information entropy weighted values ​​of all spatial points within the thermal field spatial partition and take the negative value to obtain the partition temperature uniformity evaluation value.

[0016] Furthermore, the specific process of identifying abnormal zones based on the temperature uniformity of the thermal field space partitions and implementing partition control measures is as follows: The temperature uniformity assessment value of each thermal field space partition is calculated in real time. This assessment value is compared with a temperature uniformity threshold. When the assessment value is greater than or equal to the threshold, it indicates that the temperature distribution of the thermal field space partition is uniform and in good operating condition; routine monitoring and control of the chemical reactor are maintained. When the assessment value is less than the threshold, it indicates that the temperature distribution of the thermal field space partition is abnormal. For abnormal thermal field space partitions, partition control measures are implemented: abnormal thermal field space partitions are marked, the boundaries of the thermal field space partitions are dynamically adjusted, and the stirring speed of the chemical reactor is increased. The temperature uniformity assessment value and control measures are updated in real time. The temperature uniformity assessment value, abnormal response process, and corresponding thermal field space partition control of each thermal field space partition are written into the reactor thermal field control database in real time.

[0017] Furthermore, the specific process of quantifying the control quantity of each thermal space partition using multi-point temperature values ​​within the thermal space partition is as follows: For each thermal space partition, an independent partition adaptive controller is configured for each thermal space partition to perform independent closed-loop control of the target temperature of each thermal space partition; for each thermal space partition, based on a sliding time window, the average temperature of the partition is obtained and the median is selected as the partition target temperature value; the partition temperature difference is obtained by subtracting the current partition average temperature value from the target temperature value; the absolute value of the partition temperature difference is taken as the square root and multiplied by the temperature difference adjustment weighting factor to obtain the partition temperature difference adjustment term; the partition temperature difference is then subjected to a sign function... The numerical operation yields the partition temperature difference direction term. When the partition temperature difference is greater than zero, the partition temperature difference direction term is assigned a value of 1; when the partition temperature difference is less than zero, the partition temperature difference direction term is assigned a value of -1; and when the partition temperature difference is equal to zero, the partition temperature difference direction term is assigned a value of zero. The mean temperature of all current thermal field spatial partitions is obtained, and the standard deviation is calculated to obtain the global temperature field standard deviation. The natural logarithm of the sum of the global temperature field standard deviation and constant 1 is performed, and multiplied by the temperature fluctuation compensation weight factor to obtain the partition temperature fluctuation compensation term. The partition temperature difference adjustment term is multiplied by the partition temperature difference direction term, and added to the partition temperature fluctuation compensation term to obtain the partition dynamic control value.

[0018] Furthermore, the specific process of generating temperature and stirring speed adjustment commands based on the control values ​​and sending them to the corresponding controllers is as follows: The dynamic control value of each thermal zone is calculated in real time and written into the reactor thermal field control database. When the dynamic control value of a zone is greater than the control threshold, an adjustment command to increase the zone temperature is generated based on the dynamic control value, including increasing the heating power of the thermal zone, reducing the cooling flow rate, and increasing the stirring speed of the chemical reactor. This command is then sent to the corresponding zone adaptive controller via the industrial bus to adjust the heater, cooling unit, and stirring device in real time. When the dynamic control value of a zone is less than the control threshold, an adjustment command to decrease the zone temperature is generated based on the dynamic control value, including reducing the heating power of the thermal zone, increasing the cooling flow rate, and increasing the stirring speed of the chemical reactor. This command is then sent to the corresponding zone adaptive controller to perform a cooling operation. When the dynamic control value of a zone is equal to the control threshold, no additional adjustment is performed; the adaptive controller maintains its original state and only performs real-time monitoring of the dynamic control value of the zone.

[0019] Furthermore, the specific process of real-time monitoring, anomaly diagnosis, and self-healing optimization of the reactor's thermal field status by mapping the continuous thermal field, thermal field spatial partitions, and adjustment commands to the digital platform is as follows: The continuous thermal field, each thermal field spatial partition, and adjustment command information of the reactor are mapped to the digital twin platform to visualize the three-dimensional spatial thermal field distribution of the reactor. Simultaneously, maintenance prompts are pushed in real-time based on the visualized information, and specific control suggestions, equipment self-inspection procedures, and maintenance plans for each thermal field spatial partition are generated. For abnormal thermal field spatial partitions, the cause of the anomaly is traced using an anomaly tracking algorithm, constructing partition self-healing suggestions and self-healing operation links, and dynamically tracking and evaluating the self-healing effect to continuously optimize the self-healing algorithm. At the same time, the partition temperature uniformity evaluation values, partition dynamic control values, and adjustment effects of historical thermal field spatial partitions are analyzed. Sliding window statistics and K-means clustering algorithms are used for feature extraction and pattern recognition to dynamically adjust and optimize the temperature uniformity threshold and control threshold.

[0020] Beneficial effects

[0021] The present invention has the following beneficial effects:

[0022] (1) This invention, through the acquisition of multi-point temperature and spatial coordinate data, combined with spatial interpolation and thermal field reconstruction algorithms, realizes high-resolution completion of temperature in unmarked areas inside the reactor and continuous modeling of three-dimensional thermal field, significantly improving the comprehensiveness and accuracy of thermal distribution perception and overcoming the limitations of traditional point monitoring.

[0023] (2) This invention realizes adaptive spatial partitioning of the reactor thermal field through cluster analysis based on temperature similarity and spatial proximity, and evaluates the temperature uniformity of the partitions through multivariate index of temperature information entropy. It can dynamically and accurately identify local abnormal partitions and potential hot and cold spots, and improve the agility of anomaly detection and response.

[0024] (3) In this invention, by configuring an independent adaptive controller for each thermal space partition, and using a sliding window and weight self-learning algorithm, the partition temperature and stirring speed adjustment commands are generated as needed, supporting dynamic closed-loop control with multi-loop and partition coordination, which significantly improves the precision and adaptability of thermal field regulation.

[0025] (4) The present invention can automatically trace the source and perform self-healing operations on abnormal partitions based on anomaly tracking and self-healing strategies. It can also use historical operating data to continuously optimize the control threshold and control parameters through adaptive algorithms, thereby realizing an intelligent closed loop of reactor thermal field monitoring, abnormal response and self-healing optimization.

[0026] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0027] Figure 1 A flowchart of a multivariable adaptive control method for a chemical reactor;

[0028] Figure 2 A simplified rendering of the continuous thermal field and spatial partitioning distribution across the cross-section of the reactor.

[0029] Figure 3 A bar chart visualizing the target temperature, average temperature, and dynamic control parameters for each zone. Detailed Implementation

[0030] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. As those skilled in the art will understand, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0031] Please see Figures 1-3 This invention provides a technical solution: a multivariable adaptive control method for a chemical reactor, such as... Figure 1As shown, the process includes the following steps: S1, real-time acquisition of multi-point temperature values, three-dimensional spatial coordinates of spatial points, and stirring speed of the chemical reactor to construct reactor thermal field data, and preprocessing the reactor thermal field data; S2, based on multi-point temperature values ​​and three-dimensional spatial coordinates, spatial interpolation is performed to complete the temperature of the unmarked areas within the reactor, and the completed temperature is combined to construct a continuous thermal field of the reactor, and anomaly verification is performed on the continuous thermal field; S3, cluster analysis is performed on the continuous thermal field of the reactor to divide the thermal field into spatial partitions, and the temperature uniformity of the thermal field spatial partitions is evaluated using multi-point temperature values ​​within the thermal field spatial partitions. Based on the temperature uniformity of the thermal field spatial partitions, abnormal partitions are identified, and partition control measures are implemented; S4, using multi-point temperature values ​​within the thermal field spatial partitions, the control amount of each thermal field spatial partition is quantified, and temperature and stirring speed adjustment commands are generated based on the control amounts and sent to the corresponding controllers; S5, the continuous thermal field of the reactor, thermal field spatial partitions, and adjustment commands are mapped to a digital platform in real time for real-time monitoring, anomaly diagnosis, and self-healing optimization of the reactor thermal field status.

[0032] Specifically, the process involves real-time acquisition of multiple temperature points, three-dimensional spatial coordinates of spatial points, and stirring speed from a chemical reactor to construct reactor thermal field data. The preprocessing of this data is as follows: a high-density temperature sensor array is uniformly arranged at multiple key physical locations within the reactor to acquire temperature values ​​at each point in real time. The three-dimensional spatial coordinates of each point are obtained in conjunction with the reactor's spatial structure parameters. Simultaneously, the stirring speed is acquired in real-time via the output of a frequency converter. The three-dimensional spatial coordinates, real-time temperature values, and real-time stirring speed of each point are combined to construct the reactor thermal field data. For the raw reactor thermal field data, a Gaussian filtering denoising algorithm is used for signal reduction. This algorithm effectively eliminates high-frequency noise and transient interference by applying a weighted average filter to the acquired raw temperature signal sequence, improving data stability. Furthermore, data synchronization is achieved through global timestamp alignment and linear interpolation. Global timestamp alignment refers to uniformly recording the acquisition time of data from each measuring point, avoiding time drift caused by acquisition delays from different sensors, and ensuring the comparability of multi-point data under the same time reference. Linear interpolation is used to synchronize and align data streams with different sampling frequencies, forming a structured time-series matrix from all measurement points. The 3σ discriminant method is used to detect and remove outliers. Centered on the data mean, temperature sampling points exceeding ±3 standard deviations of the mean are identified as outliers and automatically marked and removed to prevent them from interfering with subsequent algorithm modeling and control optimization. Nearest neighbor interpolation is used to fill in missing data. For data gaps caused by sensor failure or signal loss, valid observations from before and after the missing data are found and filled in according to the nearest neighbor principle, improving the continuity and completeness of the thermal field data. Simultaneously, the reactor thermal field data is standardized and normalized to eliminate dimensional influences. A reactor thermal field control database is established, storing the raw and preprocessed reactor thermal field data with timestamps.

[0033] In this implementation scheme, a high-density temperature sensor array is uniformly arranged at multiple key physical locations within the reactor. Combined with a CAD model to obtain three-dimensional spatial coordinates and real-time acquisition of stirring speed, multi-dimensional reactor thermal field data containing three-dimensional coordinates of spatial points, temperature values, and stirring speed is constructed, achieving global high-precision perception of the reactor's thermal field status. Multiple data preprocessing algorithms, including Gaussian filtering, global timestamp alignment, linear interpolation, 3σ discriminant analysis, and temporal nearest neighbor interpolation, are employed to effectively improve the synchronization, continuity, and accuracy of temperature data, spatial coordinates, and stirring speed. Standardization and normalization processes unify the numerical scale of various data types, laying a data foundation for subsequent reactor thermal field modeling, zonal analysis, and adaptive control. Finally, the raw and preprocessed reactor thermal field data are stored in the reactor thermal field control database according to timestamps, comprehensively improving the accuracy of thermal field status monitoring and the standardization of data analysis, providing strong data support and assurance for achieving dynamic control and self-healing of the reactor thermal field.

[0034] Specifically, based on multi-point temperature values ​​and three-dimensional spatial coordinates, the process of spatial interpolation to complete the temperature of the un-sensored area within the reactor is as follows: For a target point i without a sensor, all N sensor points j in the reactor are sequentially traversed; for each pair of target point i and sensor point j, the three-dimensional spatial coordinates of target point i and sensor point j and the temperature value of sensor point j are obtained, and the spatial distance value is obtained by calculating the Euclidean distance between the three-dimensional spatial coordinates; the spatial distance attenuation value is obtained by dividing the square of the spatial distance value by twice the square of the spatial weighting factor and taking the negative value as the exponent; the standard factor of the reactor thermal field distribution is obtained by taking the square root of the product of the constant 2, the constant pi, and the spatial weighting factor and taking the reciprocal; the standard factor of the reactor thermal field distribution, the spatial distance attenuation value, and the temperature value of sensor point j are multiplied to obtain the weighted temperature contribution value, and the weighted temperature contribution values ​​of all sensor points j are accumulated to obtain the interpolated temperature value of target point i, effectively realizing the continuity of spatial temperature and high-resolution thermal field reconstruction. This fully utilizes the physical information of spatial point distribution, improves interpolation accuracy, and avoids the physical distortion that may be caused by simple linear interpolation.

[0035] The specific formula for the interpolated temperature value of target point i is as follows:

[0036] ;

[0037] In the formula, The interpolated temperature value of target point i is used to complete the temperature and reconstruct the continuous thermal field of target point i in the reactor space where no temperature sensor is deployed. By using the spatial coordinates and temperature values ​​collected from all deployed points, the most likely temperature in physical location is estimated for any target point in space by means of spatial distance weight, so as to achieve continuous estimation of high-resolution thermal field distribution. This represents the spatial distance value, reflecting the Euclidean distance between target point i and measuring point j. This represents the temperature value at measuring point j, the actual observed temperature value, which forms the basis for the interpolated temperature value. The spatial weighting factor is used to control the influence range and decay rate of spatial distance on the weighted temperature contribution value. It is constructed by statistically analyzing the pairwise spatial distance values ​​between all deployed temperature measuring points, and the median of the spatial distance set is selected as the optimal spatial weighting factor, with a value range between 0.5 and 2.0. This represents the standard factor for the thermal field distribution in the reactor, ensuring that the influence weight of each measuring point on the target point has a standard distribution characteristic, which facilitates theoretical analysis. This represents the spatial distance attenuation value, ensuring that spatial distance plays a dominant role in the interpolated temperature value. It reflects the engineering logic that the closer the distance, the greater the temperature influence, and embodies the physical law of heat diffusion.

[0038] This implementation scheme fully utilizes the temperature values ​​and three-dimensional spatial coordinates of multiple points in the reactor. By constructing a spatial weighting factor and a spatial distance attenuation mechanism, temperature interpolation is performed on target points without deployed sensors, achieving continuous reconstruction and high-resolution representation of the reactor's spatial temperature field. This effectively reduces the interference of distant measuring points on target point interpolation, improving the physical rationality of local temperature distribution and the accuracy of interpolation results. Normalization processing using a standard factor for the reactor's thermal field distribution ensures the numerical stability and physical interpretability of the interpolation results. Overall, this improves the spatial continuity and integrity of the reactor's thermal field data, providing a solid data foundation for subsequent anomaly detection, zonal control, and adaptive control.

[0039] Specifically, the process of constructing a continuous thermal field for the reactor based on the completed temperature and performing anomaly verification on the continuous thermal field is as follows: The interpolated temperature values ​​of the target points requiring interpolation in the reactor space are calculated in real time. Temperature completion is performed on the unmarked areas in the three-dimensional space of the reactor. The continuous thermal field of the reactor is constructed by combining the three-dimensional spatial coordinates and temperature values ​​of each spatial point after temperature completion. This involves forming a spatial temperature distribution matrix by combining the three-dimensional spatial coordinates of all spatial points, including actual measuring points and interpolation points, with their corresponding temperature values. This achieves a full-space, continuous model of the internal heat distribution of the reactor, facilitating subsequent thermal field analysis and control. Simultaneously, to address the fluctuations in temperature values ​​in different regions of the reactor's continuous thermal field, an adaptive algorithm based on local sampling point density is used to adjust the spatial weighting factor. Local sampling point density refers to the density of sensor distribution within a certain area. The adaptive algorithm dynamically adjusts the value of the spatial weighting factor according to the number of local measuring points, making the interpolation in dense areas focus more on local measuring points, while appropriately widening the interpolation range in sparse areas, thereby improving the rationality and physical consistency of the interpolated temperature values. The continuous thermal field of the reactor is tested for spatial continuity and physical consistency. Spatial continuity testing mainly involves comparing the temperature gradients of adjacent spatial points with historical distribution patterns to determine if there are any unreasonable temperature abrupt changes. Physical consistency testing combines heat diffusion patterns with equipment operating conditions to determine whether the thermal field distribution violates the actual process logic. If local abnormal jumps or interpolation anomalies are found, multi-point spatial neighborhood interpolation is used for repair. That is, taking the abnormal point as the center, the nearest N normal measuring points and interpolation points are selected, the interpolated temperature value is recalculated, and the new value replaces the abnormal data, realizing local self-repair of thermal field anomalies. The continuous thermal field of the reactor is written into the reactor thermal field control database and output.

[0040] In this implementation scheme, interpolated temperature values ​​are calculated in real time for unsampling areas within the reactor space. By combining the three-dimensional spatial coordinates and temperature values ​​of all spatial points, a continuous and complete thermal field distribution of the reactor is constructed, achieving high-resolution full-space modeling of the internal temperature state. An adaptive algorithm based on local sampling point density is constructed to dynamically adjust the spatial weighting factor, improving the flexibility and accuracy of the interpolation algorithm in different regions and effectively balancing the physical rationality of sensor-dense and sparse areas. Through spatial continuity and physical consistency checks, local anomalous jumps and interpolation anomalies are promptly detected and corrected, enhancing the stability and reliability of the thermal field data. This provides high-quality, standardized data support for subsequent partitioned clustering, anomaly monitoring, and intelligent control.

[0041] Specifically, the process of clustering analysis to divide the continuous thermal field of the reactor into spatial partitions is as follows: The continuous thermal field of the reactor is received, and based on the temperature values ​​and three-dimensional spatial coordinates of each spatial point within the continuous thermal field, a clustering algorithm based on temperature similarity and spatial proximity is used. Specifically, K-means clustering is selected to perform clustering analysis on all spatial points. During the clustering analysis, temperature similarity is calculated using the Euclidean distance between spatial points' temperature values, and spatial proximity is evaluated using the distance between three-dimensional spatial coordinates. A weighted distance metric is constructed by combining both methods to more scientifically group spatial points with similar temperatures and proximity into the same partition, thus identifying regions with different temperature distributions within the reactor's continuous thermal field. Based on the clustering analysis results, N spatial partitions of the thermal field are dynamically divided. The clustering algorithm can adaptively adjust the number of clusters (N) according to the thermal field distribution characteristics under the current operating state of the reactor, achieving flexible partitioning for dynamic changes in the thermal field. When significant changes occur in the thermal field distribution or new abnormal regions are detected, automatic re-clustering can be performed, dynamically adjusting the partition boundaries and the number of partitions to ensure the timeliness and accuracy of the thermal field spatial partitioning. The system records the number of spatial points and their corresponding temperature values ​​within each thermal field spatial partition in real time. During clustering, to ensure real-time performance and adaptability, a sliding window clustering strategy is employed to rapidly respond to dynamic adjustments in the reactor's thermal field as operating conditions change. After each clustering analysis, the system updates the number of spatial points in each thermal field spatial partition and the temperature values ​​of all spatial points within each partition in real time, forming a three-dimensional correlation mapping between spatial points, partitions, and temperature values. This provides a partition-level thermal field data foundation for subsequent partition uniformity assessment, anomaly detection, and partition control.

[0042] like Figure 2 The image shows a simplified rendering of the continuous thermal field and spatial zoning distribution across the cross-section of a reactor. The cross-section of the chemical reactor is visually represented in a two-dimensional scatter plot, which can be seen as the temperature distribution thermal field and spatial zoning clustering of a certain plane within the reactor: the coordinates of the points represent the spatially normalized sampling point locations, covering the entire reactor cross-section; the color of the points reflects the actual temperature value of each spatial sampling point, with color bars gradually changing from low to high temperatures, and the temperature range corresponding to the color scale on the right side of the legend, clearly indicating the temperature intervals and allowing for intuitive identification of the spatial heterogeneity of the thermal field distribution, local high-temperature areas, and local low-temperature areas; the color of the outer ring of zoning lines indicates the spatial zoning of the sampling points automatically assigned by the K-means spatial clustering algorithm, with different colors representing different thermal field spatial zonings, each zoning clustering spatial points with similar temperatures and locations.

[0043] In this implementation scheme, the temperature values ​​and three-dimensional spatial coordinates of the continuous thermal field spatial points in the reactor are used to dynamically divide the thermal field into N spatial partitions using a clustering algorithm. The partition structure can be automatically adjusted according to the actual changes in the thermal field distribution. The number of spatial points and temperature values ​​of each partition are recorded in real time, realizing adaptive optimization and dynamic management of the thermal field spatial partitions. This provides an efficient and reliable foundation for subsequent partition uniformity assessment, anomaly detection, and adaptive control.

[0044] Specifically, the process of evaluating the temperature uniformity of a thermal spatial partition using multi-point temperature values ​​within the partition is as follows: For each spatial point i within the thermal spatial partition, the temperature value is obtained, and the average value is calculated to obtain the partition temperature mean, reflecting the temperature center trend of the thermal spatial partition; the absolute difference between the temperature value of spatial point i and the partition temperature mean is calculated to characterize the deviation of the single-point temperature from the overall temperature of the partition, reflecting the dispersion of the temperature distribution; the negative of the absolute difference is used as the exponent for natural exponentiation to obtain the partition temperature deviation value of spatial point i. Natural exponentiation compresses larger deviations into smaller weights, highlighting the clustering of the temperature distribution; the partition temperature deviation values ​​of all spatial points within the thermal spatial partition are summed to obtain the total partition temperature deviation; the partition temperature deviation value of spatial point i is divided by the total partition temperature deviation to obtain the temperature probability weight term of spatial point i; the temperature probability weight term of spatial point i is subjected to natural logarithmic operation to obtain the temperature weight logarithmic term of spatial point i; the temperature probability weight term is used to normalize the contribution of each point to the temperature uniformity within the partition, so that the sum of the probability weights of all spatial points equals 1. The natural logarithm operation is used to highlight the influence of extreme weights, which is helpful in revealing the extremes and dispersion of temperature distribution within a zone. Multiplying the temperature probability weight term of spatial point i by the logarithmic term of its temperature weight yields the temperature information entropy weighted value for spatial point i. Combining the probabilistic and extreme distributions, the temperature uniformity is comprehensively quantified. The temperature information entropy weighted values ​​of all spatial points within the thermal field zone are summed, and the negative value is taken to obtain the zone's temperature uniformity assessment value. A larger value indicates a more uniform temperature distribution and fewer anomalies and extreme points within the thermal field zone.

[0045] The specific formula for evaluating the temperature uniformity of the zones is as follows:

[0046] ;

[0047] In the formula, This represents the evaluation value of temperature uniformity in a thermal field space. It is used to quantify the uniformity of temperature distribution within a thermal field space partition and to determine whether the temperature distribution within a thermal field space partition is uniform, whether there are extreme points, abnormal areas, hot spots, and cold spots. It is an important basis for anomaly detection and adaptive temperature control. A negative value is used to ensure that the calculation result is non-negative. This represents the temperature value at point i in space, and indicates the spatial partitioning of the thermal field. Temperature data of each spatial point i at time t; This represents the average temperature of the zone, indicating the spatial partition of the thermal field. The average temperature of all spatial points within the zone at time t is used as the central reference for the temperature distribution of the zone. The temperature deviation value of spatial point i represents the partition temperature deviation value. The temperature deviation is converted into a weight through an exponential decay function. The larger the deviation, the lower the weight, highlighting the clustering and extreme nature of the temperature distribution. This represents the sum of temperature deviations within a partition, used to normalize the temperature deviation values ​​of all points within the partition. The temperature probability weighting term represents the probability contribution of point i in the spatial partition of the thermal field to the uniformity of temperature distribution. The logarithmic term representing the temperature weights at spatial point i is used for weighting, highlighting the contributions of extreme and dispersed weight distributions; The weighted value of the temperature information entropy at spatial point i represents the contribution of each point to the temperature information entropy of the partition. The sum of the contributions of all points within the partition reflects the overall uniformity.

[0048] In this implementation scheme, by applying exponential weighting to the absolute difference between the temperature value of each spatial point within the thermal field partition and the mean temperature of the partition, the clustering and dispersion of temperature distribution are scientifically quantified, significantly improving the accuracy of the assessment of partition temperature uniformity. By combining temperature probability weights with temperature logarithmic weights, the temperature distribution characteristics of each spatial point within the partition are comprehensively reflected as a partition temperature uniformity assessment value, achieving highly sensitive monitoring of temperature distribution within the partition and effective identification of anomalies. It can objectively reflect the temperature uniformity within the thermal field partition in real time, providing reliable quantitative basis and data support for subsequent anomaly identification, fine-tuning, and partition adaptive control, greatly improving the scientific rigor and accuracy of intelligent management of the reactor thermal field.

[0049] Specifically, the process of identifying abnormal zones based on the temperature uniformity of the thermal space partitions and implementing partition control measures is as follows: The temperature uniformity assessment value of each thermal space partition is calculated in real time. This assessment value is then compared with a temperature uniformity threshold. When the temperature uniformity assessment value is greater than or equal to the temperature uniformity threshold, it indicates that the temperature distribution of the thermal space partition is uniform and in good operating condition. In this case, routine monitoring and control of the chemical reactor are maintained. At this time, only routine data collection and recording of temperature changes within the partition are performed, maintaining the current heating, cooling, and stirring control strategies to ensure stable process operation. When the temperature uniformity assessment value of a zone is less than the temperature uniformity threshold, it indicates that the temperature distribution of the thermal field zone is abnormal. For abnormal thermal field zones, zone control measures are implemented: Abnormal thermal field zones are marked, and the boundaries of the thermal field zones are dynamically adjusted to accurately pinpoint the abnormal range, facilitating subsequent control and maintenance; the stirring speed of the chemical reactor is increased to enhance the mixing uniformity of the local temperature distribution. By increasing the stirring speed, heat transfer and diffusion within the zone are effectively accelerated, suppressing the continued development of temperature extremes and abnormal areas; the zone temperature uniformity assessment value and zone control measures are updated in real time, and the latest reactor thermal field data is continuously collected and analyzed, dynamically feeding back changes in the zone temperature uniformity assessment value. Based on the control effectiveness, corresponding control strategies are adjusted in a timely manner, forming a closed-loop adaptive control mechanism; the zone temperature uniformity assessment value, abnormal response process, and corresponding thermal field zone control of each thermal field zone are written into the reactor thermal field control database in real time, providing accurate data traceability and analysis basis for maintenance personnel and subsequent system optimization.

[0050] In this implementation scheme, by calculating the temperature uniformity assessment value of each thermal field spatial zone in real time and dynamically comparing it with the temperature uniformity threshold, timely identification and accurate location of abnormal temperature distribution in each zone are achieved. For detected abnormal thermal field spatial zones, the system can automatically mark the abnormal zones, dynamically adjust the zone boundaries, and effectively improve the mixing uniformity of local temperatures through zone control measures such as increasing the stirring speed, promoting the rapid diffusion of heat within the zones and suppressing the development of extreme temperatures and abnormal zones. Simultaneously, the system continuously collects and analyzes reactor thermal field data, dynamically updates the zone temperature uniformity assessment value and control strategies, forming a closed-loop adaptive control mechanism. This significantly improves the intelligence level of reactor thermal field monitoring and the scientific nature of control.

[0051] Specifically, the process of quantifying the control parameters of each thermal space partition using multi-point temperature values ​​within the partition is as follows: For each thermal space partition, an independent partition adaptive controller is configured. This refers to an intelligent adjustment unit deployed separately for each thermal space partition, capable of dynamically adjusting the partition's heating, cooling, and stirring actuators based on the partition's real-time temperature and target value, achieving precise temperature control and enabling independent closed-loop control of the target temperature for each thermal space partition. For each thermal space partition, based on a sliding time window, the partition's average temperature is obtained, and the median is selected as the partition's target temperature value. The sliding time window refers to statistically analyzing the real-time trend of the average temperature of each partition within a recent continuous time interval, effectively smoothing out the impact of instantaneous fluctuations and improving the robustness of the partition's target temperature value. Using the median of the partition's average temperature value avoids interference from extreme points in setting the target temperature, making it more scientific and reasonable. The partition temperature difference is obtained by subtracting the average current partition temperature from the target temperature value. The absolute value of the partition temperature difference is square-rooted and multiplied by the temperature difference adjustment weighting factor to obtain the partition temperature difference adjustment term. The partition temperature difference adjustment term is used to measure the deviation of the current temperature from the target value. Taking the square root of the absolute value improves the sensitivity to small deviations. The temperature difference adjustment weighting factor can be adaptively set according to the historical temperature fluctuation data of the partition, enhancing the flexibility of the adjustment strategy. The partition temperature difference direction term is obtained by performing a sign function operation on the partition temperature difference. That is, when the partition temperature difference is greater than zero, the partition temperature difference direction term is assigned a value of 1; when the partition temperature difference is less than zero, the partition temperature difference direction term is assigned a value of -1; and when the partition temperature difference is equal to zero, the partition temperature difference direction term is assigned a value of zero. The partition temperature difference direction term is used to determine whether the partition temperature should be increased or decreased, ensuring that the adjustment command direction is correct and preventing the control from failing in reverse. The average temperature of all current thermal field spatial partitions is obtained, and the standard deviation is calculated to obtain the global temperature field standard deviation. The sum of the global temperature field standard deviation and a constant is then subjected to natural logarithm calculation, and multiplied by the temperature fluctuation compensation weighting factor to obtain the partition temperature fluctuation compensation term. The global temperature field standard deviation reflects the overall temperature fluctuation across all partitions; the natural logarithm calculation compresses large fluctuations; and the temperature fluctuation compensation weighting factor is adaptively adjusted based on the historical global temperature field standard, thereby providing real-time compensation for the partition dynamic control values ​​and improving the overall stability and anti-interference capability of the thermal field. The partition temperature difference adjustment term is multiplied by the partition temperature difference direction term and added to the partition temperature fluctuation compensation term to obtain the partition dynamic control value. This is the core decision quantity for subsequently generating partition temperature control and stirring speed control commands, comprehensively considering the internal deviation amplitude, control direction, and global fluctuations.

[0052] The specific formula for the dynamic control value of the zone is as follows:

[0053] ;

[0054] In the formula, This represents the dynamic control value for each zone, used to determine the direction and magnitude of adjustment based on the deviation between the target temperature of each thermal zone and the current average temperature. It also incorporates adaptive compensation based on the fluctuation characteristics of the global thermal field, thereby achieving precise, dynamic, and coordinated control of the zone temperature. Indicates the target temperature value for the zone; This represents the average temperature of the zone. This represents the standard deviation of the global temperature field; This represents the temperature difference adjustment weight factor, which is based on historical zone temperature difference data and the root mean square value is calculated. It is obtained by fitting the data through a genetic algorithm and the value ranges from 0.5 to 5.0. The temperature fluctuation compensation weight factor is based on historical global temperature field standard deviation data. The global temperature field standard deviation is compared with the fluctuation threshold to obtain the number of fluctuations exceeding the standard. The global temperature field standard deviation data and the number of fluctuations exceeding the standard are used as inputs and adaptively fitted using a Bayesian optimization algorithm. The value range is between 0.1 and 2.0. This indicates the zone temperature difference adjustment term, reflecting the magnitude of the difference between the target temperature and the current zone temperature average. The absolute value is used to ensure that the response intensity increases with the increase of the deviation magnitude, and the square root controls the response sensitivity, determining the basic amplitude of the adjustment output. The larger the deviation, the larger the adjustment, and the smaller the deviation, the smaller the adjustment. This indicates the direction of the temperature difference between zones, determining whether to increase, decrease, or maintain the temperature, ensuring the positive or negative direction of the adjustment output, and preventing over-adjustment and reverse adjustment. This represents the temperature fluctuation compensation term for the zone, which reflects that the larger the thermal field fluctuation, the greater the adjustment compensation amount is, thus achieving global adaptive steady-state optimization.

[0055] In this embodiment, Table 1 is a data table of dynamic control values ​​for each zone. The temperature difference adjustment weight factor is set to 2.0, the temperature fluctuation compensation weight factor is set to 0.5, and the global temperature field standard deviation is 2.5. The table details the target temperature value, average temperature value, and dynamic control value of five different zones at the same time. Specifically, the target temperature value for zone 1 is 65.2, the average temperature value is 62.8, and the dynamic control value is 3.725; the target temperature value for zone 2 is 61.7, the average temperature value is 63.3, and the dynamic control value is -1.903; the target temperature value for zone 3 is 70.4, the average temperature value is 69.7, and the dynamic control value is 2.300; the target temperature value for zone 4 is 67.9, the average temperature value is 68.2, and the dynamic control value is -0.469; and the target temperature value for zone 5 is 66.1, the average temperature value is 65.3, and the dynamic control value is 2.415.

[0056] Table 1. Data Table of Dynamic Adjustment Values ​​for Each Zoning Area

[0057]

[0058] like Figure 3 The image shows a bar chart visualizing the target temperature, average temperature, and dynamic control values ​​for each zone. A dual Y-axis is used to display the comparison between the target temperature, average temperature, and dynamic control values ​​for each of the five zones. The left Y-axis represents temperature, and the right Y-axis represents the dynamic control values. Different zone numbers correspond to three sets of bars of different colors. (Based on Table 1 and...) Figure 3 It can be seen that the difference between the target temperature value and the average temperature value of most zones is small, reflecting that the overall temperature control is relatively accurate. However, the average temperature value of zone 2 is slightly higher than the target temperature value, and the target temperature value of zone 3 is slightly higher than the average temperature value, indicating that there is a certain temperature control deviation in some zones. When the dynamic control value of a zone is positive, it means that the target temperature value of the zone is higher than the current average temperature value of the zone, and heating and enhanced temperature control are needed; when it is negative, it means that the target temperature value of the zone is lower than the average temperature value of the zone, and cooling and weakening of heating are needed. The negative values ​​of zones 2 and 4 are consistent with the trend that the actual average temperature value of the zones is higher than the target temperature value of the zones. The dynamic control values ​​of zones 1, 3, and 5 are positive and have a large amplitude, indicating that these zones need to continue to heat up and strengthen control, while zone 4 hardly needs any adjustment, and the temperature control is close to the ideal state.

[0059] In this implementation scheme, an independent adaptive controller is configured for each thermal space partition. A sliding time window is used to acquire the average temperature of each partition in real time, and the median is selected as the target temperature value for that partition. Combined with partition temperature difference adjustment, partition temperature difference direction, and partition temperature fluctuation compensation terms, the dynamic control value for each partition is dynamically calculated, achieving closed-loop adaptive control of each thermal space partition. This comprehensively reflects the deviation of the current temperature from the target temperature, the adjustment direction, and the global temperature fluctuation status, improving the scientific nature and sensitivity of partition temperature control. It significantly enhances the uniformity and stability of the reactor's thermal field distribution, providing a quantitative and reliable decision-making basis for the accurate generation and execution of subsequent partition control commands.

[0060] Specifically, the process of generating and issuing temperature and stirring speed control commands to the corresponding controllers based on the control parameters is as follows: The dynamic control value for each thermal zone is calculated in real time and written into the reactor thermal field control database. When the dynamic control value exceeds the control threshold, a control command to increase the zone temperature is generated based on the dynamic control value. This includes increasing the heating power of the thermal zone (including increasing the output power of the electric heater and steam jacket heating device), reducing the cooling flow rate, and adjusting the speed of the stirring motor controlled by the frequency converter to increase the stirring speed of the chemical reactor. This increases energy input, improves the mixing efficiency of the zoned fluids, and accelerates the uniformity of heat distribution. The command is then sent to the corresponding zone adaptive controller via the industrial bus to adjust the heaters, cooling units, and stirring devices in real time, thus enabling communication between the distributed zone controllers and the central monitoring system. Efficient and reliable command transmission ensures timely and accurate adjustment actions. When the dynamic control value of a zone is less than the control threshold, an adjustment command to reduce the zone temperature is generated based on the dynamic control value. This includes reducing the heating power of the thermal zone, increasing the cooling flow rate (including increasing the flow rate of cooling coils and chilled water circulation), and increasing the stirring speed of the chemical reactor to achieve rapid heat balance and prevent the spread of abnormal local temperatures. The command is then sent to the corresponding zone adaptive controller to execute the cooling operation. When the dynamic control value of a zone is equal to the control threshold, no additional adjustment is performed. The adaptive controller maintains its original state and only performs real-time monitoring of the dynamic control value of the zone. It continuously collects and records the real-time average temperature of each thermal zone, the stirring speed of the reactor, and the feedback status of the actuator. This facilitates trend analysis and effect evaluation of subsequent control strategies, ensuring long-term process stability.

[0061] In this implementation scheme, by calculating the dynamic control values ​​of each thermal zone in real time and combining them with control thresholds to generate adjustment commands for temperature and stirring speed, precise linkage control of zone heating power, cooling flow rate, and stirring speed is achieved. All control data and execution processes are written to the reactor thermal field control database in real time, ensuring both the timeliness and accuracy of the adjustment actions and providing a complete data link for subsequent process optimization and fault tracing. This effectively improves the intelligence and response speed of temperature control in each thermal zone, ensuring spatial uniformity and dynamic stability of the internal thermal field of the reactor.

[0062] Specifically, the process of real-time monitoring, anomaly diagnosis, and self-healing optimization of the reactor's thermal field status by mapping the continuous thermal field, thermal field spatial partitions, and adjustment commands to a digital platform is as follows: The continuous thermal field, each thermal field spatial partition, and adjustment command information are mapped to a digital twin platform. This digital twin platform, relying on 3D modeling and real-time data synchronization technology, constructs a virtual space corresponding one-to-one with the physical reactor, visualizing the 3D spatial thermal field distribution of the reactor, thereby achieving dynamic visualization and remote interactive management of the 3D spatial thermal field distribution. Simultaneously, maintenance prompts are pushed in real-time based on the visualized information, and control suggestions, equipment self-inspection procedures, and maintenance plans for specific thermal field spatial partitions are generated. Based on the platform's real-time monitoring results, maintenance prompts for thermal field anomalies and equipment status changes are pushed. Combined with data-driven analysis, personalized control suggestions are generated for each thermal field spatial partition. Equipment self-inspection procedures can be automatically generated and pushed, including sensor verification, actuator self-testing, and communication link detection, assisting maintenance personnel in timely detection and handling of potential risks. For abnormal thermal field spatial partitions, an anomaly tracking algorithm is used to trace the causes of anomalies. Based on the partition temperature uniformity assessment value, partition dynamic control value, and historical reactor thermal field data, the algorithm automatically identifies the main causes of anomalies, such as local heating failure, poor cooling, and insufficient mixing. The anomaly tracking algorithm can then generate partition self-healing suggestions, including combined operations of heating, cooling, and mixing, and construct partition self-healing suggestions and self-healing operation links. After self-healing is executed, the self-healing effect is dynamically tracked and evaluated to continuously optimize the self-healing algorithm. Simultaneously, the partition temperature uniformity assessment value, partition dynamic control value, and adjustment effect of historical thermal field spatial partitions are analyzed. Sliding window statistics and K-means clustering algorithms are used for feature extraction and pattern recognition to dynamically adjust and optimize the temperature uniformity threshold and control threshold. Among them, sliding window statistics are used to analyze the historical trend of partition temperature uniformity and partition dynamic control value in real time, and K-means clustering algorithm automatically identifies the thermal field characteristic patterns of partitions under different operating conditions, thereby dynamically optimizing the temperature uniformity threshold and control threshold, enhancing the adaptability and accuracy of threshold setting.

[0063] In this implementation scheme, by mapping the continuous thermal field, thermal field spatial partitioning, and adjustment command information of the reactor to a digital twin platform in real time, visualization, intelligent monitoring, and self-healing optimization of the three-dimensional thermal field distribution are achieved. Based on the partitioned temperature uniformity assessment value and partitioned dynamic control value, it can intelligently identify anomalies and push control suggestions. Furthermore, by utilizing sliding window statistics and K-means clustering algorithms, it adaptively optimizes the temperature uniformity threshold and control threshold, significantly improving the intelligence level and operational safety of the reactor's thermal field management.

[0064] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0065] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. As those skilled in the art will understand, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A multivariable adaptive control method for a chemical reactor, characterized in that, Includes the following steps: S1, real-time acquisition of multi-point temperature values, three-dimensional spatial coordinates of spatial points and stirring speed of chemical reactor, construction of reactor thermal field data, and data preprocessing of reactor thermal field data; S2, based on multi-point temperature values ​​and three-dimensional spatial coordinates, performs spatial interpolation to complete the temperature of the unmarked area in the reactor, constructs a continuous thermal field of the reactor by combining the completed temperature, and performs anomaly verification on the continuous thermal field of the reactor. S3. Cluster analysis is performed on the continuous thermal field of the reactor to divide the thermal field into spatial partitions. The temperature uniformity of the thermal field spatial partitions is evaluated by using the temperature values ​​of multiple points within the thermal field spatial partitions. Abnormal partitions are identified based on the temperature uniformity of the thermal field spatial partitions, and partition control measures are implemented. S4: Utilize the temperature values ​​at multiple points within the thermal space partition to quantify the control amount of each thermal space partition, generate temperature and stirring speed adjustment commands based on the control amount, and send them to the corresponding controllers. The specific process of quantifying the control parameters of each thermal space partition by utilizing the temperature values ​​at multiple points within the thermal space partition is as follows: For each thermal space partition, an independent partition adaptive controller is configured for each thermal space partition to perform independent closed-loop control of the target temperature of each thermal space partition; For each thermal field space partition, the mean temperature of the partition is obtained based on the sliding time window, and the median is selected as the target temperature value of the partition. The partition temperature difference is obtained by subtracting the average temperature of the current partition from the target temperature value. The absolute value of the partition temperature difference is then taken as the square root and multiplied by the temperature difference adjustment weight factor to obtain the partition temperature difference adjustment term. The sign function operation is performed on the zone temperature difference to obtain the zone temperature difference direction term. That is, when the zone temperature difference is greater than zero, the zone temperature difference direction term is assigned a value of 1; when the zone temperature difference is less than zero, the zone temperature difference direction term is assigned a value of -1; and when the zone temperature difference is equal to zero, the zone temperature difference direction term is assigned a value of zero. Obtain the mean temperature of all current thermal field spatial partitions, calculate the standard deviation to obtain the global temperature field standard deviation, perform natural logarithmic operation on the sum of the global temperature field standard deviation and constant 1, and multiply it with the temperature fluctuation compensation weight factor to obtain the partition temperature fluctuation compensation term. Multiply the zone temperature difference adjustment term by the zone temperature difference direction term, and add it to the zone temperature fluctuation compensation term to obtain the zone dynamic control value; S5 maps the continuous thermal field, thermal field spatial partitioning, and adjustment commands of the reactor to the digital platform in real time, enabling real-time monitoring, anomaly diagnosis, and self-healing optimization of the reactor's thermal field status.

2. The multivariable adaptive control method for a chemical reactor according to claim 1, characterized in that, The specific process for real-time acquisition of multi-point temperature values, three-dimensional spatial coordinates of spatial points, and stirring speed of the chemical reactor to construct reactor thermal field data, and the data preprocessing of the reactor thermal field data is as follows: A high-density temperature sensor array is uniformly arranged at multiple key physical locations in the reactor to collect the temperature values ​​of each measuring point in real time. The three-dimensional spatial coordinates of each spatial point are obtained by combining the spatial structure parameters of the reactor. At the same time, the stirring speed is collected in real time through the output of the frequency converter. The three-dimensional spatial coordinates of each spatial point, real-time temperature value, and real-time stirring speed are combined to construct the reactor thermal field data. For the original reactor thermal field data, a Gaussian filtering denoising algorithm is used to reduce signal noise, and data synchronization is achieved through global timestamp alignment and linear interpolation. Outliers are detected and removed using the 3σ discriminant method, and missing data is filled in using temporal nearest neighbor interpolation. At the same time, the reactor thermal field data is standardized and normalized to establish a reactor thermal field control database. The original and preprocessed reactor thermal field data with timestamps are stored in the reactor thermal field control database.

3. The multivariable adaptive control method for a chemical reactor according to claim 1, characterized in that, The specific process of spatially interpolating and completing the temperature in the unmarked areas within the reactor based on multi-point temperature values ​​and three-dimensional spatial coordinates is as follows: For a target point i without a sensor, all N measuring points j in the reactor are traversed sequentially. For each pair of target point i and measuring point j, the three-dimensional spatial coordinates of target point i and measuring point j and the temperature value of measuring point j are obtained. The spatial distance value is obtained by calculating the Euclidean distance between the three-dimensional spatial coordinates. The spatial distance attenuation value is obtained by dividing the square of the spatial distance value by twice the square of the spatial weight factor and taking the negative value as the exponent. The standard factor for the thermal field distribution of the reactor is obtained by taking the square root of the product of constant 2, pi constant, and spatial weighting factor, and then taking the reciprocal. The weighted temperature contribution value is obtained by multiplying the standard factor of the thermal field distribution of the reactor, the spatial distance attenuation value, and the temperature value of measuring point j. The weighted temperature contribution values ​​of all measuring points j are then summed to obtain the interpolated temperature value of the target point i.

4. The multivariable adaptive control method for a chemical reactor according to claim 1, characterized in that, The specific process of constructing a continuous thermal field for the reactor by combining the completed temperature and performing anomaly verification on the continuous thermal field is as follows: The interpolation temperature values ​​of the target points that need to be interpolated in the reactor space are calculated in real time, and the temperature of the unpointed areas in the three-dimensional space of the reactor is completed. The three-dimensional spatial coordinates and temperature values ​​of each spatial point after temperature completion are combined to construct the continuous thermal field of the reactor. At the same time, in response to the fluctuation of temperature values ​​in different areas of the continuous thermal field of the reactor, an adaptive algorithm based on the density of local sampling points is used to adjust the spatial weighting factor. The spatial continuity and physical consistency of the continuous thermal field of the reactor are checked. If local abnormal jumps and interpolation abnormalities are found, multi-point spatial neighborhood interpolation is used for repair. That is, with the abnormal point as the center, the nearest N normal measuring points and interpolation points are selected, the interpolation temperature value is recalculated, and the abnormal data is replaced with the new value. The continuous thermal field of the reactor is written into the reactor thermal field control database and output.

5. The multivariable adaptive control method for a chemical reactor according to claim 1, characterized in that, The specific process of performing cluster analysis on the continuous thermal field of the reactor and dividing the thermal field spatial partitions is as follows: The system receives the continuous thermal field of the reactor and, based on the temperature values ​​and three-dimensional spatial coordinates of each spatial point in the continuous thermal field, performs cluster analysis on all spatial points using a clustering algorithm based on temperature similarity and spatial proximity. This identifies regions in the continuous thermal field where the temperature distribution of spatial points differs. Based on the cluster analysis results, the system dynamically divides the thermal field into N spatial partitions and records the number of spatial points and their corresponding temperature values ​​in each spatial partition in real time.

6. The multivariable adaptive control method for a chemical reactor according to claim 1, characterized in that, The specific process of evaluating the temperature uniformity of a thermal space partition using multi-point temperature values ​​within the thermal space partition is as follows: For each spatial point i within the thermal field spatial partition, the temperature value is obtained, and the average value is calculated to obtain the partition temperature mean. The absolute difference between the temperature value of spatial point i and the partition temperature mean is calculated. The negative of the absolute difference is used as the exponent for natural exponentiation to obtain the partition temperature deviation value of spatial point i. The partition temperature deviation values ​​of all spatial points within the thermal field spatial partition are summed to obtain the total partition temperature deviation. Divide the temperature deviation value of spatial point i by the sum of the temperature deviations of the partitions to obtain the temperature probability weight term of spatial point i. Perform natural logarithmic operation on the temperature probability weight term of spatial point i to obtain the temperature weight logarithmic term of spatial point i. Multiply the temperature probability weight term of spatial point i with the logarithmic term of the temperature weight of spatial point i to obtain the temperature information entropy weighted value of spatial point i; sum the temperature information entropy weighted values ​​of all spatial points in the thermal field spatial partition and take the negative value to obtain the partition temperature uniformity evaluation value.

7. The multivariable adaptive control method for a chemical reactor according to claim 1, characterized in that, The specific process of identifying abnormal zones based on the temperature uniformity of the thermal space and implementing zone control measures is as follows: The temperature uniformity assessment value of each thermal field space partition is calculated in real time. The temperature uniformity assessment value of each thermal field space partition is compared with the temperature uniformity threshold. When the temperature uniformity assessment value of each partition is greater than or equal to the temperature uniformity threshold, it indicates that the temperature distribution of the thermal field space partition is uniform and it is in good operating condition. Routine monitoring and control of the chemical reactor are maintained. When the temperature uniformity assessment value of a zone is less than the temperature uniformity threshold, it indicates that the temperature distribution of the thermal field zone is abnormal. For abnormal thermal field zones, zone control measures are implemented: mark the abnormal thermal field zones, dynamically adjust the boundaries of the thermal field zones, and increase the stirring speed of the chemical reactor; update the zone temperature uniformity assessment value and zone control measures in real time. The temperature uniformity assessment values ​​of each thermal field space partition, the abnormal response process, and the corresponding thermal field space partition control are written into the reactor thermal field control database in real time.

8. The multivariable adaptive control method for a chemical reactor according to claim 1, characterized in that, The specific process of generating temperature and stirring speed adjustment commands based on the control amount and sending them to the corresponding controllers is as follows: The dynamic control value of each thermal field space partition is calculated in real time and written into the reactor thermal field control database. When the dynamic control value of the partition is greater than the control threshold, the adjustment command to increase the partition temperature value is generated according to the dynamic control value of the partition, including increasing the heating power of the thermal field space partition, reducing the cooling flow rate and increasing the stirring speed of the chemical reactor. The command is sent to the corresponding partition adaptive controller through the industrial bus to adjust the heater, cooling unit and stirring device in real time. When the dynamic control value of a zone is less than the control threshold, an adjustment command to reduce the temperature of the zone is generated based on the dynamic control value of the zone. This includes reducing the heating power of the thermal field zone, increasing the cooling flow rate, and increasing the stirring speed of the chemical reactor. The command is then sent to the corresponding zone adaptive controller to perform the cooling operation. When the dynamic control value of a partition equals the control threshold, no additional adjustment is made. The adaptive controller maintains its original state and only performs real-time monitoring of the dynamic control value of the partition.

9. The multivariable adaptive control method for a chemical reactor according to claim 1, characterized in that, The specific process of mapping the continuous thermal field, thermal field spatial partitioning, and adjustment commands of the reactor to the digital platform in real time, and performing real-time monitoring, anomaly diagnosis, and self-healing optimization of the reactor's thermal field status, is as follows: The continuous thermal field of the reactor, the spatial partitions of each thermal field, and the adjustment command information are mapped to the digital twin platform to visualize the three-dimensional spatial thermal field distribution of the reactor. At the same time, operation and maintenance prompts are pushed in real time based on the visualized information, and control suggestions, equipment self-inspection procedures, and maintenance plans for specific thermal field spatial partitions are generated. For abnormal thermal field spatial partitions, the cause of the abnormality is traced through the abnormality tracking algorithm, a partition self-healing suggestion and self-healing operation link are constructed, and the self-healing effect is dynamically tracked and evaluated to continuously optimize the self-healing algorithm. Simultaneously, the evaluation values ​​of temperature uniformity, dynamic control values, and control effects of historical thermal field spatial partitions are analyzed. Sliding window statistics and K-means clustering algorithms are used for feature extraction and pattern recognition to dynamically adjust and optimize temperature uniformity thresholds and control thresholds.

Images