TEC semiconductor-based temperature control method, device and equipment

Abstract

The invention relates to the technical field of TEC semiconductors, and discloses a temperature control method, device and equipment based on a TEC semiconductor, and the method comprises the steps: carrying out the distributed temperature collection and preprocessing of a region to be subjected to temperature control, and obtaining a temperature correlation matrix; performing temperature anomaly detection to obtain a temperature anomaly feature matrix; calculating a temperature influence range of the TEC controller, and mapping the control group to a virtual control network; performing global temperature optimization to obtain a global optimal control variable, and dividing the global optimal control variable into local control instructions according to control groups; a temperature control cost function is constructed, and TEC control parameters are obtained through gradient iteration optimization calculation; and TEC control parameters are input into the temperature dynamic prediction model, robust optimization is carried out in a rolling prediction interval to obtain a local optimal control sequence, and temperature closed-loop control is executed. According to the invention, reasonable distribution of control resources is realized, mutual interference among TEC controllers is reduced, and accuracy and reliability of temperature control are ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of TEC semiconductors, and in particular to a temperature control method, device and equipment based on TEC semiconductors. Background Art

[0002] The existing TEC temperature control system mainly adopts a single-point control strategy, and each TEC controller works independently. This control method has many shortcomings when dealing with large-scale temperature control scenarios. As the control area expands, the number of temperature collection points increases sharply. The traditional full-area temperature collection strategy leads to a significant increase in data processing overhead and computing load, affecting the real-time performance and control accuracy of temperature regulation.

[0003] In addition, the existing TEC control algorithm faces convergence problems in low temperature fluctuation environments. Due to the spatial coupling characteristics of the temperature field, the control behaviors between adjacent TEC controllers will interfere with each other, resulting in a decrease in the overall temperature stability of the system. At the same time, the temperature anomaly detection method is relatively simple and cannot effectively identify and process temperature anomalies in complex environments, which increases the uncertainty of the temperature control system. There is a lack of effective multi-TEC collaborative control framework in the existing technology. There is a lack of information sharing and collaborative decision-making mechanism between the various TEC controllers, and it is impossible to achieve optimal control of the global temperature field. At the same time, the control algorithm is not robust enough, and it is difficult to maintain stable control performance in the face of external interference and model uncertainty. Summary of the invention

[0004] The present invention provides a temperature control method, device and equipment based on TEC semiconductor, which realizes the reasonable allocation of control resources, reduces the mutual interference between TEC controllers, and ensures the accuracy and reliability of temperature control.

[0005] In a first aspect, the present invention provides a temperature control method based on a TEC semiconductor, the temperature control method based on a TEC semiconductor comprising: Perform distributed temperature collection and preprocessing on the temperature-controlled area to obtain preprocessed temperature data and calculate the corresponding temperature correlation matrix; Inputting the preprocessed temperature data into a Gaussian mixture model to perform temperature anomaly detection to obtain a temperature anomaly feature matrix; Calculating the temperature influence range of the TEC controller according to the temperature correlation matrix and the temperature anomaly characteristic matrix, dividing a plurality of TEC controllers into a plurality of control groups based on the temperature influence range, and mapping the control groups to a virtual control network; Inputting the temperature distribution data of the virtual control network into a mixed integer linear programming model for global temperature optimization to obtain a global optimal control variable, and dividing the global optimal control variable into local control instructions according to the control group; Based on the local control instructions, a temperature control cost function is constructed, and a TEC control parameter is obtained by gradient iterative optimization calculation; The TEC control parameters are input into the temperature dynamic prediction model, robust optimization is performed within the rolling prediction interval to obtain a local optimal control sequence, and temperature closed-loop control is performed.

[0006] In a second aspect, the present invention provides a temperature control device based on a TEC semiconductor, the temperature control device based on a TEC semiconductor comprising: The acquisition module is used to perform distributed temperature acquisition and preprocessing on the temperature-controlled area, obtain preprocessed temperature data, and calculate the corresponding temperature correlation matrix; A detection module, used for inputting the preprocessed temperature data into a Gaussian mixture model to perform temperature anomaly detection and obtain a temperature anomaly feature matrix; a calculation module, configured to calculate a temperature influence range of the TEC controller according to the temperature correlation matrix and the temperature anomaly characteristic matrix, divide a plurality of TEC controllers into a plurality of control groups based on the temperature influence range, and map the control groups to a virtual control network; An optimization module, used for inputting the temperature distribution data of the virtual control network into a mixed integer linear programming model for global temperature optimization, obtaining a global optimal control variable, and dividing the global optimal control variable into local control instructions according to the control group; A construction module is used to construct a temperature control cost function based on the local control instruction, and obtain TEC control parameters through gradient iterative optimization calculation; The execution module is used to input the TEC control parameters into the temperature dynamic prediction model, perform robust optimization within the rolling prediction interval to obtain a local optimal control sequence, and perform temperature closed-loop control.

[0007] A third aspect of the present invention provides a computer device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the computer device executes the above-mentioned TEC semiconductor-based temperature control method.

[0008] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned temperature control method based on TEC semiconductor.

[0009] In the technical solution provided by the present invention, by introducing distributed temperature acquisition and preprocessing technology and combining the calculation method of the temperature correlation matrix, the data processing overhead of the system is significantly reduced, while the accuracy and reliability of the temperature data are guaranteed; the Gaussian mixture model is used for temperature anomaly detection, and the construction of the temperature anomaly feature matrix is ​​coordinated to improve the system's recognition ability and processing efficiency of temperature anomalies; the TEC controller grouping strategy based on the temperature influence range realizes the reasonable allocation of control resources and reduces the mutual interference between controllers; through the construction of a virtual control network and the application of a mixed integer linear programming model, the optimal control of the global temperature field is realized, and the overall control accuracy of the system is improved; the TEC control parameter calculation method based on gradient iterative optimization is adopted to ensure the optimality of the control parameters and the dynamic response performance of the system; the temperature dynamic prediction model and the rolling time domain optimization strategy are introduced to enhance the anti-interference ability and control stability of the system; the closed-loop control of the PWM drive signal is realized to ensure the accuracy and reliability of the temperature control. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0011] Figure 1 A schematic diagram of the steps of a temperature control method based on a TEC semiconductor in an embodiment of the present invention; Figure 2 Schematic diagram of the structure of a temperature control device based on TEC semiconductor in an embodiment of the present invention; Figure 3 It is a schematic block diagram of the structure of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0012] The embodiments of the present invention provide a temperature control method, device and equipment based on TEC semiconductor. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0013] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 , an embodiment of the temperature control method based on TEC semiconductor in the embodiment of the present invention includes: Step S1, performing distributed temperature collection and preprocessing on the temperature-controlled area to obtain preprocessed temperature data, and calculating the corresponding temperature correlation matrix; It is understandable that the execution subject of the present invention may be a temperature control device based on TEC semiconductor, or a terminal or a server, which is not limited here. The embodiment of the present invention is described by taking a server as the execution subject.

[0014] Specifically, the temperature gradient distribution per unit area of ​​the temperature-controlled area is calculated. By analyzing the temperature change trend at different locations in the area, a suitable temperature collection point arrangement scheme is determined so that the collection points can cover the temperature change of the entire area and generate a temperature collection point distribution sequence. For each temperature collection point, the data collection time interval and data sampling frequency are set, and the original temperature data is collected. The original temperature data is normalized and converted into a standardized numerical sequence. A sliding time window is constructed based on the standardized temperature numerical sequence. The temperature data at each moment is smoothed to eliminate the impact of short-term fluctuations. A more stable temperature sequence is obtained by weighted average calculation within the sliding window, reducing data fluctuations caused by external interference or system noise. By weighted average, according to the importance of temperature changes in different time periods, its contribution to the average value is adjusted so that the smoothed data is more in line with the actual situation. The sliding average temperature sequence is smoothed. The data is smoothed by different filtering algorithms (such as low-pass filtering or Gaussian filtering) to obtain preprocessed temperature data. The preprocessed temperature data is constructed into an M×N-dimensional temperature data array according to the spatial coordinates of the temperature collection points, where M represents the number of temperature collection points and N represents the length of the time series corresponding to each collection point. The Pearson correlation coefficient is calculated for the M×N dimensional temperature data array. The Pearson correlation coefficient is a method to measure the linear relationship between two variables. By calculating the correlation between temperature data, the relationship between different temperature collection points can be judged, so as to understand the spatial characteristics of temperature distribution. The Pearson correlation coefficient matrix is ​​weighted superimposed with the preset spatial distance weight. The spatial distance weight considers the spatial position relationship of each temperature collection point, that is, the temperature points with closer distances have higher correlation, while the temperature points with farther distances have lower correlation. Through weighted superposition, the actual temperature correlation between different collection points is reflected, and finally the temperature correlation matrix is ​​obtained.

[0015] Step S2, inputting the preprocessed temperature data into a Gaussian mixture model for temperature anomaly detection to obtain a temperature anomaly feature matrix; Specifically, the preprocessed temperature data is divided into multiple temperature intervals according to a preset temperature interval. The temperature data in each temperature interval represents the frequency distribution of temperature within a certain range. By calculating the frequency of occurrence in each temperature interval, a temperature distribution histogram is obtained. The temperature distribution histogram is calculated to generate a probability density distribution curve of the temperature data. The curve shows the probability of occurrence of different temperature values, thereby revealing the overall distribution characteristics of the temperature data. The probability density distribution curve is input into the Gaussian mixture model, and the parameters are iteratively calculated through the expectation maximization (EM) algorithm to obtain the various parameters of the mixed Gaussian model. The Gaussian mixture model is a tool that can capture the complex distribution of data. It describes the multimodal characteristics of temperature data by weighted combination of multiple Gaussian distributions. Through the iteration of the EM algorithm, the parameters of the model are continuously optimized to obtain the optimal Gaussian distribution model. Based on the parameters of the mixed Gaussian model, the conditional probability of each temperature point is calculated. The conditional probability reflects the possibility of a certain temperature value appearing in the model. In order to determine which temperature data are abnormal data, a target conditional probability value is set. When the conditional probability of a certain temperature point is lower than the target value, the point is considered to be an abnormal point. This process helps identify abnormal values ​​in temperature data and obtain preliminary abnormality detection results. For each abnormal point in the preliminary detection results, calculate its gradient value with adjacent temperature points. The gradient value represents the rate of change of temperature in space. A larger gradient value means abnormal temperature change, which indicates a more significant temperature anomaly. According to the calculated gradient value, each abnormal point is classified into an abnormality level. The temperature point with a higher abnormality level indicates that its abnormality is more serious and requires more control intervention. Through this step, a level is assigned to the abnormal temperature data, and a basis is provided for the subsequent temperature control strategy to obtain an abnormal temperature level sequence. The abnormal temperature level sequence is subjected to spatial correlation analysis with the detection data of adjacent temperature sensors to reveal the characteristics of the spatial distribution of the abnormal points, so as to understand whether the temperature anomaly is a local phenomenon or whether it has a larger range of propagation trends. If the anomaly of a temperature point has a strong spatial correlation with the temperature data of multiple surrounding sensors, it means that the anomaly is caused by changes in environmental factors in the region, rather than a single device failure. Through analysis, the target anomaly detection result is obtained. The spatial distribution characteristics, temperature deviation characteristics, and time duration characteristics of the abnormal points in the target anomaly detection result are extracted to construct a temperature anomaly feature matrix. The spatial distribution characteristics reflect the distribution pattern of abnormal points in space, the temperature deviation characteristics show the degree of deviation between the abnormal points and the normal temperature range, and the time duration characteristics describe how long the abnormal phenomenon lasts.

[0016] Step S3, calculating the temperature influence range of the TEC controller according to the temperature correlation matrix and the temperature anomaly characteristic matrix, dividing the multiple TEC controllers into multiple control groups based on the temperature influence range, and mapping the control groups to the virtual control network;Specifically, the temperature correlation matrix is ​​subjected to singular value decomposition according to the main diagonal. The temperature data matrix is ​​converted into a set of basis vectors that can reveal the characteristics of temperature spatial distribution. The singular value decomposition method can effectively extract the main features in the temperature data, and obtain the temperature space mapping matrix, which reflects the temperature change pattern between different spatial regions. The temperature space mapping matrix is ​​convolved with the temperature anomaly feature matrix, and the characteristics of temperature correlation and temperature anomaly are combined to generate the temperature response characteristic map of the TEC controller, which shows the response ability of each TEC controller under different temperature conditions and its temperature control effect in different regions. Based on the temperature response characteristic map, the temperature influence coefficient of each TEC controller is calculated. The temperature influence coefficient reflects the degree of influence of the controller on the temperature change of the surrounding area. By calculating these coefficients, it is determined which areas are significantly affected by a specific TEC controller. For areas with a temperature influence coefficient greater than a preset threshold, it is determined to be the temperature influence range of the controller. In this way, the temperature control capability range of each TEC controller in space is delineated. According to the overlap between the temperature influence ranges, the correlation matrix of the TEC controller is constructed. The correlation matrix can quantify the interaction and correlation between different controllers, thereby revealing their synergistic effect in the temperature control process. By performing K-means clustering analysis on the TEC controller association matrix, the controllers are divided into multiple control groups according to the similarity between them. The controllers in each control group work together through the overlap of the temperature influence range to achieve the optimal effect of overall temperature control. A hierarchical control structure is constructed based on the TEC control grouping results, and the entire control system is organized and managed according to different hierarchical relationships, in which each control group operates as an independent unit and is assigned different control weights according to its importance and role in the system. A unique identifier is assigned to each control group, and a master-slave relationship is established, so that some controllers act as master controllers and others act as slave controllers within the control group to achieve more effective collaboration and management. The construction of a hierarchical control structure makes the control of the entire system more flexible and scalable, while improving the system's fault tolerance and reliability. Based on the control group management model, virtual nodes are constructed. Virtual nodes are abstractions of each control group and their relationships. Through virtual nodes, the temperature control characteristics of each control group and the relationships between groups are mapped to virtual space to form an initial virtual control network. In order to improve the efficiency and adaptability of the control network, dynamic link optimization is performed on the initial virtual network. By dynamically adjusting the connection mode between nodes in the network, the efficiency of data transmission and the speed of control command transmission are maximized. During the link optimization process, the data transmission path between nodes is set according to the direction of the temperature gradient to ensure the fastest path for information transmission and the optimal control strategy.The direction of the temperature gradient reflects the trend of temperature change. Therefore, when setting the transmission path between nodes, the consideration of the temperature gradient can help avoid redundancy and lag in information transmission, ensuring that the entire temperature control system can respond to temperature changes in real time and make corresponding adjustments. The virtual control network after link optimization has higher operating efficiency and adaptability, allowing the TEC controller to control the temperature more accurately under complex environmental conditions, ensuring the temperature stability and energy efficiency of the system.

[0017] Step S4, inputting the temperature distribution data of the virtual control network into a mixed integer linear programming model for global temperature optimization, obtaining a global optimal control variable, and dividing the global optimal control variable into local control instructions according to the control group; Specifically, an A×B dimensional state matrix is ​​constructed for the temperature data of each node in the virtual control network, where A represents the number of control groups and B represents the length of the sampling time. The temperature data of each control group is arranged in time series to form a temperature distribution feature matrix, which reflects the temperature state of each node in the entire virtual control network at different time points. The temperature distribution feature matrix is ​​input into the linear prediction model, and the temperature change trend is calculated by the least squares method. The least squares method is a regression analysis tool that can effectively fit temperature data and predict future temperature changes. Based on the prediction model, a temperature prediction sequence is generated to help the system predict future temperature fluctuations. Based on the temperature prediction sequence, the objective function of global temperature optimization is constructed. The construction of this objective function considers two main factors: temperature control error and control quantity change. In order to ensure the accuracy of temperature control and the stability of the system, the weight of temperature control error is set to 0.7, and the weight of control quantity change is set to 0.3. By setting the weight, it is ensured that the optimization process will not cause excessive control variable fluctuations while pursuing temperature accuracy, ensuring the smooth operation of the system. After the objective function is determined, a mixed integer linear programming model is constructed, which is used to optimize the global temperature control of the system. The mixed integer linear programming model can obtain the global optimal control variable by mathematical modeling, comprehensively considering the temperature error, the change of the control quantity and the constraints between the controllers. The mixed integer linear programming model is solved by the interior point method. The interior point method is an efficient optimization algorithm that can quickly solve high-dimensional linear programming problems while ensuring the accuracy of the solution. Through this method, the global optimal control variable is obtained, which represents the operation that each TEC controller in the system should take at different time points. In order to facilitate implementation, the global optimal control variable is decomposed by dimensionality reduction. According to the spatial distribution of the control group, the control quantity is mapped to each TEC unit to obtain the initial control instruction to guide the specific operation of each controller in different time periods. The initial control instruction is matched with the topological structure of the virtual control network. Through the matching operation, the operation coordination between the control groups is ensured, so as to achieve the coordinated work of the overall system. The generation of the coordinated control sequence needs to consider the accuracy of temperature control and ensure the consistency of the timing between different control groups. The coordinated control sequence will be sliced ​​according to the time dimension to generate multiple control subsequences. Each subsequence represents the control instruction at a certain moment, and each subsequence contains the synchronization mark between groups. These synchronization marks ensure the time coordination and instruction synchronization between groups when multiple control groups are operating, thereby avoiding temperature fluctuations or control failures caused by inconsistent control instructions. Distributed control instructions are obtained by processing the time slices and inter-group synchronization marks of the coordinated control sequence. These instructions will be grouped according to the identifier of the control group, provide specific guidance for the operation of each control group, and add execution timestamps to each group of instructions.The addition of timestamps can accurately control the execution timing of each control group, ensuring that in the distributed control system, each TEC controller can execute its task in a predetermined time sequence, maximizing the control efficiency and temperature control accuracy. Through the above steps, the local control instructions are finally obtained.

[0018] Step S5, constructing a temperature control cost function based on the local control instruction, and obtaining TEC control parameters through gradient iterative optimization calculation; Specifically, a temperature control constraint matrix is ​​created based on local control instructions. The matrix contains the operation constraints of each control group, reflecting the limitations and requirements of temperature regulation during the operation of the system. When constructing the temperature control constraint matrix, it is necessary to comprehensively consider various factors of temperature control, including temperature error, control response capability, and abnormal state. The temperature control constraint matrix is ​​input into the weighted cost function generator, and the temperature control error term is given a weight of 0.6, and the abnormal state term is given a weight of 0.4. This cost function will guide the subsequent temperature regulation and control process to ensure that the system can respond quickly in temperature control and effectively avoid abnormal fluctuations. Based on the temperature control cost function, the TEC power supply current value is piecewise linearized to obtain a current control quantization table. Based on the current control quantization table, a response function between the current value and the temperature change rate is established to obtain a TEC temperature response model, which reflects the actual impact of current adjustment on temperature change. Gradient descent optimization is performed based on the TEC temperature response model. Gradient descent is an optimization algorithm that finds the optimal current parameter through iterative calculation. During the gradient descent optimization process, the current parameter is gradually adjusted until the system reaches an optimal state of temperature control. Through this process, the current parameter optimization sequence obtained shows the most appropriate current value in the TEC system, which can achieve the most accurate regulation of the target temperature. Based on the current parameter optimization sequence, the temperature regulation step is calculated. The temperature regulation step refers to the amplitude of temperature change in each control cycle. By dividing the temperature change by the control cycle, a standardized temperature regulation parameter is obtained, which can maintain consistency in different control cycles to ensure the continuity and stability of temperature regulation. Stability analysis is performed on the standardized temperature regulation parameters. An upper limit of the temperature regulation step is set, and the range of the control cycle is defined to ensure that the temperature regulation process is neither too drastic nor too drastic. The parameter constraints obtained from the stability analysis are matched and screened with the current parameter optimization sequence to select the optimal parameter combination that meets the constraints.

[0019] Step S6: Input the TEC control parameters into the temperature dynamic prediction model, perform robust optimization within the rolling prediction interval to obtain the local optimal control sequence, and execute temperature closed-loop control.

[0020] Specifically, the TEC control parameters are input into the system to construct an A×3-dimensional control matrix, where A is the number of control groups, and the three dimensions correspond to the TEC power supply current value, control cycle, and temperature adjustment step size. The control matrix is ​​the core parameter matrix of the temperature control system, which contains the basic control information of each control group in different time periods. The control matrix is ​​parameter-cascaded with the temperature anomaly feature matrix to construct a state vector. By inputting the state vector into the temperature dynamic prediction model, the system can predict the temperature state in the rolling time domain. In the rolling prediction interval, the temperature dynamic prediction model predicts the temperature change trend in the future period based on the historical temperature data and the current state vector. The temperature prediction result is output in the form of a prediction matrix and used as the input of the robust optimization controller. The robust optimization controller generates a temperature optimization objective function based on the prediction matrix to further optimize the temperature control process. In order to make the optimization objective function meet the actual needs of the system, a temperature control deviation threshold is set to ensure that the optimization process does not deviate from the set control range. Based on this objective function, a quadratic programming problem is constructed and solved using a sequential quadratic programming algorithm to obtain a control quantity sequence, which represents the optimal operation plan of the temperature control system in the future control cycle, ensuring that the temperature can be adjusted according to the set target. The control quantity sequence is updated online to generate the segmented optimal control sequence. The rolling update process ensures that the control sequence can reflect the changes in temperature status in real time and adjust according to the new temperature data, so as to continuously optimize the control effect. Through online optimization, the system can continuously adjust and improve the control strategy in each time period to ensure that the temperature can be stabilized near the target value. According to the segmented optimal control sequence, the TEC devices are grouped and controlled, and the corresponding control instructions are assigned in order of priority. The control commands executed in parallel are converted into PWM (pulse width modulation) drive signals through the current modulation circuit. The PWM signal can accurately control the current input of the TEC device, thereby achieving fine temperature regulation. By adjusting the duty cycle of the PWM signal, the control system adjusts the heat output of the TEC unit to achieve the desired temperature regulation effect. The temperature closed-loop control is performed through the preset switching frequency to ensure the stability and consistency of the PWM signal, making the temperature control process smoother and more efficient.

[0021] In the embodiment of the present invention, by introducing distributed temperature acquisition and preprocessing technology and combining the calculation method of the temperature correlation matrix, the data processing overhead of the system is significantly reduced, while the accuracy and reliability of the temperature data are guaranteed; the Gaussian mixture model is used for temperature anomaly detection, and the construction of the temperature anomaly feature matrix is ​​coordinated to improve the system's recognition ability and processing efficiency of temperature anomalies; the TEC controller grouping strategy based on the temperature influence range realizes the reasonable allocation of control resources and reduces the mutual interference between controllers; through the construction of a virtual control network and the application of a mixed integer linear programming model, the global temperature field is optimized and the overall control accuracy of the system is improved; the TEC control parameter calculation method based on gradient iterative optimization is adopted to ensure the optimality of the control parameters and the dynamic response performance of the system; the temperature dynamic prediction model and the rolling time domain optimization strategy are introduced to enhance the anti-interference ability and control stability of the system; the closed-loop control of the PWM drive signal is realized to ensure the accuracy and reliability of the temperature control.

[0022] In a specific embodiment, the process of executing step S1 may specifically include the following steps: Calculate the temperature gradient distribution per unit area of ​​the temperature-controlled area, arrange the temperature collection points, and obtain the temperature collection point distribution sequence; Setting a data collection time interval and a data sampling frequency for each temperature collection point in the temperature collection point distribution sequence, and collecting original temperature data; Perform normalization calculation on the original temperature data to generate a standardized temperature value sequence; A sliding time window is constructed based on the standardized temperature value sequence, and the temperature values ​​in each sliding time window are weighted averaged to obtain a sliding average temperature sequence; Perform data smoothing on the sliding mean temperature series to obtain preprocessed temperature data; The preprocessed temperature data is constructed into an M×N dimensional temperature data array according to the spatial coordinates of the temperature collection points, where M is the number of temperature collection points and N is the length of the sampling time series; The Pearson correlation coefficient is calculated for the M×N dimensional temperature data array to obtain a temperature correlation coefficient matrix, and the temperature correlation coefficient matrix is ​​weighted superimposed with a preset spatial distance weight to obtain a temperature correlation matrix.

[0023] Specifically, the entire area to be controlled is gridded. Through gridding, the temperature distribution within each unit area is determined, and the temperature gradient of each grid unit is obtained. The temperature gradient describes the rate of temperature change within a unit area, and its formula is: ; in, represents the temperature gradient, It represents the temperature difference per unit area. Represents the distance. By dividing the entire area into grids, calculating the temperature gradient within each grid, and obtaining the temperature gradient distribution of the entire temperature-controlled area. Arrange the temperature collection points according to the temperature gradient distribution. More temperature collection points are required in areas with larger temperature gradients to ensure that rapid temperature changes can be captured, while the number of collection points should be appropriately reduced in areas with smaller temperature gradients. In this way, the arrangement of temperature collection points can more effectively reflect the temperature characteristics of the area to be controlled, and obtain the distribution sequence of temperature collection points. Set the data collection time interval and data sampling frequency for each temperature collection point in the temperature collection point distribution sequence. Assume that the time interval is , the sampling frequency is The number of samples in each time period is . Comprehensively consider the rate of temperature change and the system's response time to temperature regulation. For rapidly changing temperature areas, the sampling frequency should be higher to capture finer temperature change details, while for slower changing areas, the sampling frequency should be appropriately reduced. Collect the original temperature data and perform normalization calculations on the original data to standardize the values ​​of different temperature collection points into a unified dimension range. The formula for normalization calculation is expressed as: ; in, represents the normalized temperature value, represents the original temperature value, and are the minimum and maximum values ​​of the temperature sequence of the collection point, respectively. Through normalization, the temperature data is standardized to the range of [0,1], making the data of different collection points more comparable. A sliding time window is constructed based on the standardized temperature value sequence. The length of the sliding time window is Indicates that the role of the window is to segment the time series data so that the data in each time period can be aggregated. For the data in each sliding time window, a weighted average calculation is performed. The weighted average formula is as follows: ; in, represents the weighted average temperature within the sliding window, Indicates The temperature value at a time point, Represents the weight corresponding to the time point. Smooth the temperature data to reduce the impact of noise on the overall temperature trend. The weight is set according to the importance of the time point. For example, the closer the data is to the current moment, the greater the weight is, so that it can better reflect the current temperature changes. After the weighted average calculation within the sliding time window is completed, the sliding average temperature sequence is obtained. Smooth the data of the sliding average temperature sequence. Use low-pass filtering or Gaussian filtering. The formula for low-pass filtering is: ; in, For the The smoothed temperature value at time, is the smoothing coefficient, and its value range is between [0,1]. By selecting an appropriate smoothing coefficient, high-frequency noise is effectively filtered out, making the temperature series smoother and obtaining preprocessed temperature data. The preprocessed temperature data is constructed according to the spatial coordinates of the temperature collection point. dimensional array of temperature data, where is the number of temperature collection points, is the length of the sampling time series. The matrix contains the temperature information of each acquisition point at each sampling time, which can reflect the temperature distribution characteristics of the entire controlled area. Based on the constructed temperature data array, the temperature relationship between different acquisition points is analyzed. By calculating The Pearson correlation coefficient of the dimensional temperature data array is obtained to obtain the temperature correlation coefficient matrix. The calculation formula of the Pearson correlation coefficient is: ; in, Indicates the temperature collection point and collection points The correlation coefficient between and Respectively Collection point at a time point and collection points The temperature value, and The collection points and The average temperature value of the Pearson correlation coefficient ranges from -1 to 1, reflecting the linear correlation between the two acquisition points. is close to 1, indicating that there is a strong positive correlation between the two collection points; if If it is close to -1, it means that there is a strong negative correlation between them; if If it is close to 0, it means that there is no obvious linear relationship between them. After obtaining the temperature correlation coefficient matrix, consider the spatial distribution of the temperature collection points. The distance between different collection points will affect their temperature correlation. The temperature correlation coefficient matrix is ​​weighted and superimposed with the preset spatial distance weight to obtain the final temperature correlation matrix. The formula for weighted superposition is: ; in, is the final temperature correlation matrix element, For collection point and The Pearson correlation coefficient, According to the collection point and The distance between The calculated weight. The farther the collection point is, the smaller its weight is, to reflect the weaker correlation between collection points at longer distances.

[0024] In a specific embodiment, the process of executing step S2 may specifically include the following steps: The preprocessed temperature data is divided into temperature intervals according to a preset temperature interval, and the occurrence frequency of the data in each temperature interval is calculated to obtain a temperature distribution histogram; Perform probability density distribution calculation on the temperature distribution histogram to generate a probability density distribution curve of the temperature data; The probability density distribution curve is input into the Gaussian mixture model, and the parameters are iteratively calculated through the expectation maximization algorithm to obtain the parameters of the mixed Gaussian model; The conditional probability of each temperature point is calculated based on the parameters of the mixed Gaussian model, and the temperature data with a conditional probability value lower than the target value is marked as an abnormal point to obtain the initial abnormality detection result; For each abnormal point in the initial abnormal detection result, the gradient value between it and the adjacent temperature point is calculated, and the abnormal level is divided according to the gradient value to obtain the abnormal temperature level sequence; Perform spatial correlation analysis on the abnormal temperature level sequence and the detection data of adjacent temperature sensors to obtain the target abnormality detection result; The spatial distribution characteristics, temperature deviation characteristics and time duration characteristics of the abnormal points in the target anomaly detection results are extracted, and the temperature anomaly feature matrix is ​​constructed.

[0025] Specifically, the preprocessed temperature data is divided into temperature intervals according to the preset temperature intervals. Assume that the preset temperature interval is , all temperature data are divided into several intervals, and the range of each interval is ,in is the lower limit of a certain interval. The frequency of occurrence of data in each temperature interval is calculated to obtain a temperature distribution histogram, which displays the frequency of temperature data in each temperature interval and intuitively reflects the overall distribution of temperature data. For example, the temperature data of a certain area to be controlled is between [20,30] degrees. , the temperature is divided into ten intervals such as [20,21], [21,22], and then the frequency of data in each interval is counted. Based on these frequencies, a temperature distribution histogram is constructed to describe the distribution of temperature values ​​in different ranges. The probability density distribution of the temperature distribution histogram is calculated to generate a probability density distribution curve for the temperature data. The probability density function is used to represent the relative frequency of a temperature value in the entire data set, and its definition is as follows: ; in, is the temperature value The probability density of Indicates the frequency of occurrence of temperature data in the interval, Represents the total number of temperature data samples, is the width of the temperature interval. Through probability density distribution calculation, the temperature histogram is converted into a smooth probability density distribution curve, making the distribution of temperature data more continuous and statistically significant. The probability density distribution curve of temperature is input into the Gaussian mixture model, and the parameters are iteratively calculated through the expectation maximization (EM) algorithm to obtain the parameters of the mixed Gaussian model. The Gaussian mixture model is a probability-based statistical model that can fit multimodal data distribution and is suitable for data sets such as temperature that are affected by multiple factors. The Gaussian mixture model is expressed as follows: ; in, Indicates temperature value The mixing probability, Indicates The weights of a Gaussian distribution, For the The mean of a Gaussian distribution, For the The variance of a Gaussian distribution, is the number of Gaussian distributions. The EM algorithm is used to iteratively solve the parameters of the model. In the expectation step, the probability of each data point belonging to each Gaussian component is calculated based on the current model parameters. Then, in the maximization step, the parameters of each component are updated according to the calculated probability. The iteration is repeated until convergence to obtain the parameters of the mixed Gaussian model. The conditional probability of each temperature point is calculated based on the parameters of the Gaussian mixture model. The calculation of conditional probability is used to determine whether a temperature value is within the normal range. The formula is as follows: ; in, GMM) represents the temperature under the Gaussian mixture model The conditional probability of the abnormal point is calculated. In order to detect abnormal points, the conditional probability value is compared with the preset target threshold. If the conditional probability of a temperature point is lower than the threshold, the temperature data is marked as an abnormal point to obtain the initial abnormal detection result. For each abnormal point in the initial abnormal detection result, its relationship with the adjacent temperature points is analyzed. The gradient value between the abnormal point and its adjacent temperature points is calculated. The gradient calculation formula is: ; in, Indicates temperature point and The temperature gradient between and Temperature points and The temperature value, Represents the distance between two points. By calculating these gradient values, the temperature change rate of each abnormal point is evaluated. According to the size of the gradient value, the abnormal points are divided into different abnormal levels. The larger the gradient value, the more drastic the temperature change and the higher the abnormal level of the point. Thus, an abnormal temperature level sequence is obtained to analyze the severity of the abnormal situation. The abnormal temperature level sequence is subjected to spatial correlation analysis with the detection data of adjacent temperature sensors to determine whether the abnormal point is associated with the temperature sensor data around it. If there is a strong spatial correlation between the abnormal point and the temperature data of multiple adjacent sensors, it indicates that the anomaly is caused by the overall temperature fluctuation in the area, rather than the error of a single sensor. Spatial correlation analysis is achieved by calculating the Pearson correlation coefficient, and its formula is: ; in, Indicates sensor and The correlation coefficient between and Sensors and In time The temperature value, and Sensors and The average temperature of the abnormal point. By analyzing the correlation between the abnormal point and the adjacent temperature sensor, a more accurate target abnormality detection result is obtained. The spatial distribution characteristics, temperature deviation characteristics and time duration characteristics of the abnormal point in the target abnormality detection result are extracted. The spatial distribution characteristics describe the location of the abnormal point in the entire controlled area, for example, whether it is concentrated in a specific area. The temperature deviation characteristic represents the difference between the temperature value of the abnormal point and the normal value, and its calculation formula is: ; in, is the temperature deviation, is the temperature value of the abnormal point, is the mean value of the normal temperature range. The time duration feature describes the duration of an abnormal state, which is used to determine whether the abnormality is a short-term fluctuation or a long-term existence. Based on these features, a temperature anomaly feature matrix is ​​constructed to describe the characteristics of each abnormal point.

[0026] In a specific embodiment, the process of executing step S3 may specifically include the following steps: Perform singular value decomposition on the temperature correlation matrix according to the main diagonal to obtain the temperature space mapping matrix; The temperature space mapping matrix is ​​convolved with the temperature anomaly characteristic matrix to generate a temperature response characteristic diagram of the TEC controller; Calculating a temperature influence coefficient for each TEC controller based on the temperature response characteristic diagram, and determining an area where the temperature influence coefficient is greater than a preset coefficient threshold as a temperature influence range; According to the overlap between the temperature influence ranges, a TEC controller association matrix is ​​constructed, and K-means clustering is performed on the TEC controller association matrix to obtain the TEC control grouping results; According to the TEC control grouping results, a hierarchical control structure is constructed, and a unique identifier is assigned to each control group in the hierarchical control structure, and the master-slave relationship of the controllers in the group is established to generate a control group management model; Virtual nodes are constructed based on the control group management model. The temperature control characteristics of each control group and the relationship between groups are mapped to the virtual space to form an initial virtual network. The initial virtual network is dynamically optimized, and the data transmission path between nodes is set according to the direction of the temperature gradient to obtain a virtual control network.

[0027] Specifically, singular value decomposition is performed on the temperature correlation matrix to obtain the temperature space mapping matrix. Its expression is: ; in, and are the left singular matrix and the right singular matrix respectively, is a diagonal singular value matrix. Singular value decomposition decomposes the original matrix into multiple lower-rank matrices, and obtains the temperature space mapping matrix by selecting the main singular value , which is used to describe the main variation pattern of temperature in space. Temperature space mapping matrix It is the part extracted from the singular value decomposition, which contains the main temperature distribution information in the entire controlled area. The temperature space mapping matrix is ​​convolved with the temperature anomaly feature matrix to generate the temperature response feature map of the TEC controller. The information of the temperature space is combined with the anomaly feature to obtain the response capability of each TEC controller in space. The expression of convolution is: ; in, is the temperature response characteristic map after convolution, * represents the convolution operation. The convolution operation is understood as using the temperature space mapping matrix As a filter, the abnormal feature matrix Filter and get a new feature map , which describes the response capability of each TEC controller when facing abnormal temperature distribution. The temperature influence coefficient is calculated for each TEC controller based on the temperature response characteristic graph. The temperature influence coefficient is used to measure the temperature regulation effect of a certain TEC controller on a specific spatial area, and its calculation method is: ; in, For the The temperature influence coefficient of a TEC controller is is the coefficient of the spatial weight, which is determined according to the distance between the controller and the controlled area. The temperature response characteristic diagram is The controller is in The response value of each position. By weighted summing the responses of all positions, the total influence of the controller in the entire controlled area is obtained. The temperature influence coefficient is compared with the preset coefficient threshold. If the temperature influence coefficient of a controller is greater than this threshold, the corresponding area is determined as the temperature influence range, thereby ensuring that each TEC controller only acts on the area it can effectively control, thereby improving the overall efficiency of the temperature control system. According to the overlap between the temperature influence ranges, the TEC controller association matrix is ​​constructed. . Each element of the correlation matrix Representative Controller and controller The degree of overlap of the temperature influence range is calculated as follows: ; in, and Controller and controller The temperature influence range is represents the intersection area of ​​the influence ranges of the two controllers, Represents the union area of ​​the influence range of the two controllers. The correlation matrix is ​​used to measure the synergy between different TEC controllers in temperature control. , K-means clustering is performed on the TEC controllers to determine the grouping results of TEC control. Controllers with similar temperature influence ranges are grouped together to facilitate collaborative work. The clustering objective function is as follows: ; in, is the clustering objective function, is the number of clusters, Indicates Clusters, For controller The correlation vector of For the The center vector of each cluster. By minimizing the objective function , and the best clustering grouping result is obtained. A hierarchical control structure is constructed based on the grouping results of TEC controllers. The controllers in each control group are assigned a unique identifier, and a master-slave relationship is established among the controllers in the group. The master controller is responsible for coordinating the operations of the slave controllers within the group to achieve precise temperature control of the entire area. Through the hierarchical control structure, the mutual interference between the controllers can be effectively reduced, and the control efficiency and stability of the system can be improved. Based on the hierarchical control structure, virtual nodes are constructed, and the temperature control characteristics of each control group and the relationship between the groups are mapped to the virtual space to form an initial virtual network. In the virtual network, each control group is abstracted as a virtual node, and the connection between the nodes represents the collaborative relationship between different control groups. The initial virtual network reflects the spatial and functional associations between the control groups. In order to improve the operating efficiency of the virtual network, dynamic link optimization is performed on it. The data transmission path between the nodes is set according to the direction of the temperature gradient to ensure that the information can be transmitted along the direction of the fastest temperature change. The direction of the temperature gradient is calculated by the following formula: ; in, is the temperature gradient vector, , , Respectively indicate the temperature , , The rate of change in the direction. According to the calculated temperature gradient, the optimal data transmission path is set between virtual nodes to ensure that the control information can be transmitted in the system in a timely manner to cope with the dynamic changes in temperature. By dynamically optimizing the links of the initial virtual network, an efficient virtual control network is obtained. The virtual control network can respond to temperature changes in real time and dynamically adjust the operation of the TEC controller according to the temperature requirements of the system to achieve accurate temperature management of the entire area to be controlled.

[0028] In a specific embodiment, the process of executing step S4 may specifically include the following steps: An A×B dimensional state matrix is ​​constructed for the temperature data of each node in the virtual control network, where A is the number of control groups and B is the sampling time length, and the temperature distribution characteristic matrix is ​​obtained; The temperature distribution feature matrix is ​​input into the linear prediction model, and the temperature change trend is calculated based on the least squares method to obtain the temperature prediction sequence; The global temperature optimization objective function is constructed based on the temperature prediction sequence, the temperature control error weight is set to 0.7, the control amount change weight is set to 0.3, and the mixed integer linear programming model is obtained; The mixed integer linear programming model is solved by the interior point method to obtain the global optimal control variables, and the global optimal control variables are decomposed by dimensionality reduction. The control quantities are mapped to each TEC unit according to the spatial distribution of the control group to obtain the initial control instructions. The initial control instruction is matched with the topological structure of the virtual control network to obtain a coordinated control sequence, and the coordinated control sequence is sliced ​​in the time dimension to generate a control subsequence, and an inter-group synchronization mark is added to obtain a distributed control instruction; The distributed control instructions are grouped according to the identifier of the control group, and an execution timestamp is added to each group of instructions to obtain local control instructions.

[0029] Specifically, the temperature data of each node in the virtual control network is processed to construct dimensional state matrix, where represents the number of control groups, is the sampling time length. This matrix describes the temperature change of each control group at different time points and can effectively reflect the temporal and spatial distribution characteristics of the system temperature. Suppose the temperature data matrix is , each element of which Representative The control group The temperature data at each time point. The temperature distribution feature matrix is ​​input into the linear prediction model, the temperature change trend is calculated using the least squares method, and a temperature prediction sequence is generated. The linear prediction model assumes that the temperature change is described by a series of linear relationships, and the model is expressed as: ; in, For the The control group The temperature at a point in time, is the coefficient to be estimated, is the error term, is the regression order. The optimal coefficient of the model is obtained by minimizing the sum of squares of temperature prediction errors through the least squares method. . Using these coefficients, we predict the future temperature values ​​and get the temperature prediction sequence. Based on the temperature prediction sequence, we build the objective function of global temperature optimization. This objective function aims to balance the temperature control error and the change of the control quantity to ensure that the temperature of the system fluctuates stably within the set range. The mathematical expression of the objective function is: ; in, is the target temperature value for each control group, and are the weights of the temperature error term and the control quantity change term, respectively. and . The first term is the sum of squares of temperature control errors, and the second term is the sum of squares of changes in the control quantity, which is used to penalize sharp fluctuations in temperature. By minimizing this objective function, smooth temperature control can be achieved while avoiding unnecessary fluctuations in the control quantity. The above optimization objective function is input into the mixed integer linear programming model for solution. Mixed integer linear programming is a method for optimization problems that is suitable for mixed problems involving discrete variables and continuous variables. The constraints of the objective function include the current limit of the TEC controller, the time step limit, and the temperature stability constraint of the system. Through the interior point method or other numerical optimization algorithms, the mixed integer linear programming model can efficiently solve the global optimal control variables. Assume that the control variable is , then the optimization problem is expressed as: ; in, is the control variable vector, which contains the operation values ​​of all controllers at each time point. By solving this optimization problem, the global optimal control variable is obtained . The global optimal control variables are decomposed by dimensionality reduction, and the control quantities are mapped to each TEC unit according to the spatial distribution of the control group. Assuming that the optimization result is a vector containing the control quantities at all time points, the purpose of dimensionality reduction is to map the control quantities to specific TEC devices. The mapped control quantity sequence will be used as the initial control instruction to guide the temperature adjustment of the TEC unit. The initial control instruction is matched with the topological structure of the virtual control network. The topological structure of the virtual control network describes the connection relationship between the control groups. Based on this topology, it is determined how to coordinate the control instructions between different control groups. The matching operation process is implemented through matrix operations or graph theory algorithms to ensure that the control instructions meet the topological constraints of the physical network when they are transmitted between the control groups. By coordinating the generation of the control sequence, it is sliced ​​in the time dimension and the control instructions are divided into multiple subsequences. Each control subsequence corresponds to a time period and contains the control instructions of all control groups in the time period. In order to ensure the coordination of control, an inter-group synchronization mark is added to the control subsequence to indicate the coordination execution order between different control groups. Therefore, the control tasks are executed between multiple control groups in a predetermined time sequence to ensure that the temperature change of each control group is carried out according to the global optimal strategy. All control instructions are grouped according to the identifier of the control group. The control instructions in each group are arranged in sequence according to the operation of the controller in the group, and the execution timestamp is added to obtain the local control instructions. The timestamp marks the execution time of the control instruction to ensure that the operation of each control group can be carried out at the right time. Through the distributed control method, the work of each TEC controller can be efficiently managed and coordinated to achieve the optimal control of the global temperature.

[0030] In a specific embodiment, the process of executing step S5 may specifically include the following steps: A temperature control constraint matrix is ​​created based on the local control instructions, and the temperature control constraint matrix is ​​input into a weighted cost function generator, a weight of 0.6 is assigned to the temperature control error term, and a weight of 0.4 is assigned to the abnormal state term, to obtain a temperature control cost function; According to the temperature control cost function, the TEC power supply current value is processed piecewise linearly to obtain a current control quantization table, and according to the current control quantization table, a response function between the current value and the temperature change rate is established to obtain a TEC temperature response model; Based on the TEC temperature response model, gradient descent optimization is performed to obtain the current parameter optimization sequence, and the temperature adjustment step is calculated based on the current parameter optimization sequence. The temperature change is divided by the control period to generate a standardized temperature adjustment parameter. The stability analysis of the standardized temperature adjustment parameters is carried out, the upper limit of the temperature adjustment step and the range of the control cycle are set, the parameter constraints are obtained, and the parameter constraints are matched and screened with the current parameter optimization sequence, and the optimal parameter combination that meets the constraints is selected as the TEC control parameters.

[0031] Specifically, a temperature control constraint matrix is ​​created based on local control instructions. The local control instructions contain the specific operation requirements of each control unit in different time periods. These instructions are expressed as temperature control constraints in the form of a matrix. Suppose the temperature control constraint matrix is , where each element Indicates The control unit is The temperature control constraint matrix is ​​input into the weighted cost function generator to generate a temperature control cost function for system optimization. When constructing the cost function, two main factors are considered: temperature control error and abnormal state. The temperature control error refers to the deviation between the actual temperature and the target temperature, while the abnormal state refers to the abnormal temperature fluctuation during the operation of the system. The weight of the temperature control error term is set to 0.6, and the weight of the abnormal state term is set to 0.4. The mathematical expression of the cost function is: ; in, is the temperature control cost function, Indicates The actual temperature at a given moment, is the target temperature, For the Abnormal status indicators at a time point, is the total number of time points. The first term represents the sum of squares of temperature control errors, and the second term represents the penalty for abnormal conditions. By assigning different weights, the cost function places more emphasis on the accuracy of temperature control, while taking into account the stability of the system to avoid the frequent occurrence of abnormal conditions. Based on the temperature control cost function, the power supply current value of TEC is piecewise linearized to better control and optimize the current input. The relationship between current and temperature is usually nonlinear. Through the piecewise linearization method, the nonlinear relationship is approximated as multiple linear segments. Assume is the supply current, and the temperature change rate is , the change of current is divided into multiple segments, each segment corresponds to a linear relationship. The piecewise linearization is expressed as: ; in, and They are The slope and intercept of the segment, is the temperature boundary value of each segment. Through this step, the nonlinear relationship between current and temperature is decomposed into multiple linear segments to improve the accuracy of system control. Based on the segmented linear relationship, a current control quantization table is established to correspond different temperature intervals to the corresponding power supply current, thereby helping the control system to select the appropriate current value under different temperature conditions. The response function between the current value and the temperature change rate is established using the current control quantization table to obtain the TEC temperature response model. The response function describes the direct effect of current input on temperature change, and its mathematical expression is: ; in, represents the rate of temperature change, is the response function of current to the temperature change rate, and are the slope and intercept of the linear model respectively. This model can help predict the rate of change of temperature under a given current input and provide data support for the control optimization of the system. Based on the TEC temperature response model, gradient descent optimization is performed to determine the optimal current parameter sequence. Gradient descent is an optimization algorithm that finds the control variable that minimizes the cost function through continuous iteration. Let the cost function be , then the update rule for gradient descent is: ; in, For the The current value of the iteration, is the learning rate, is the gradient of the cost function to the current. By continuously adjusting the current value , find the cost function The minimum optimal current parameter sequence. Based on the current parameter sequence, calculate the temperature adjustment step. The temperature adjustment step is defined as the change in temperature within a control cycle, and the calculation formula is: ; in, is the temperature adjustment step, is the temperature change, is the control period. By dividing the temperature change by the control period, the standardized temperature adjustment parameters are obtained to facilitate the consistency of temperature adjustment in different control periods. Stability analysis is performed on the standardized temperature adjustment parameters. The temperature adjustment process must be kept within a reasonable range to avoid excessive temperature fluctuations or too fast adjustment leading to system instability. An upper limit for the temperature adjustment step is set to ensure that the system will not cause drastic temperature fluctuations during the control process due to excessive step length. Set the upper limit of the step length to , the range of the control cycle is , then the constraints for temperature regulation are: ; The parameter constraints are matched and screened with the current parameter optimization sequence, and the optimal parameter combination that meets all the constraints is selected as the final TEC control parameters.

[0032] In a specific embodiment, the process of executing step S6 may specifically include the following steps: An A×3-dimensional control matrix is ​​constructed for the TEC control parameters, where A is the number of control groups and the three dimensions correspond to the TEC power supply current value, control cycle and temperature adjustment step, respectively, to obtain the initial matrix of control parameters; The control parameter initial matrix and the temperature anomaly characteristic matrix are parameter-concatenated to construct a state vector; The state vector is input into the temperature dynamic prediction model, a rolling time domain is set for the state vector through the temperature dynamic prediction model, and the temperature state is predicted to obtain a prediction matrix; The prediction matrix is ​​input into the robust optimization controller, the temperature control deviation threshold is set, the temperature optimization objective function is generated, and a quadratic programming problem is constructed based on the temperature optimization objective function. The sequential quadratic programming algorithm is used to solve the problem and the control quantity sequence is obtained. The control quantity sequence is updated online to generate a segmented optimal control sequence, and the TEC devices are grouped and controlled according to the segmented optimal control sequence. The control instructions are assigned in order of priority to obtain control commands for parallel execution. The parallel executed control commands are converted into PWM drive signals through the current modulation circuit, and the temperature closed-loop control is performed using a preset switching frequency.

[0033] Specifically, according to the number of control groups and three key control variables - TEC supply current value , control cycle , and temperature adjustment step , construct a dimensional control matrix. Each row in the matrix corresponds to the control parameters of a control group. For example, assuming there are three control groups, the control matrix is ​​expressed as: ; in, are the supply current values ​​of the first, second and third control groups respectively, are the control cycle, The temperature adjustment step size for each control group is respectively constructed. A control matrix is ​​constructed to provide preliminary control parameter settings for each control group. The control parameter initial matrix is ​​cascaded with the temperature anomaly characteristic matrix to construct a state vector. Assume that the temperature anomaly characteristic matrix is (one dimensional matrix, where is the dimension of the temperature anomaly data), then the state vector The control matrix and the temperature anomaly characteristic matrix are merged column by column, and the mathematical expression is: ; in, is a The state vector of the dimension contains the control parameters of each control group and the corresponding temperature anomaly characteristics. The state vector is input into the temperature dynamic prediction model. The model uses historical temperature data and current control parameters to predict the future temperature change trend. In order to improve the accuracy of the prediction, the temperature dynamic prediction model adopts the rolling time domain method, that is, in each calculation, the state vector at the current moment is used to predict the temperature change in the future. Suppose the temperature prediction sequence is , then the rolling prediction process is described as: ; in, It is a prediction function trained based on historical data. is the current time step, is the prediction time interval, is the state variable at the current moment. Through rolling prediction, the model can dynamically update the prediction results. Input to the robust optimization controller for optimization. The goal of robust optimization is to minimize the temperature control deviation and handle uncertainty. Set a temperature control deviation threshold , that is, when the predicted temperature is equal to the target temperature When the difference between is greater than the threshold, it is considered that the temperature control has deviated. Optimization objective function It is expressed as: ; in, and are the weights of the temperature control deviation term and the temperature adjustment step term, is the predicted temperature, is the target temperature, is the current temperature adjustment step. Through this optimization function, while ensuring the temperature control accuracy, avoid excessive temperature adjustment step, so as to maintain the stability of the system. Using the temperature optimization objective function, the quadratic programming problem is solved by the sequential quadratic programming algorithm to obtain the control quantity sequence. The quadratic programming algorithm is an algorithm used to optimize continuous problems. It solves the optimal solution by continuously iteratively adjusting the control quantity. The optimization process is: ; in, is the control quantity sequence to be solved, is the quadratic matrix of the control variable, is a linear term, is a constraint condition to ensure that the control quantity is within a reasonable range. After optimization, we get Provide appropriate control signals for each control group. Perform online rolling updates on the control quantity sequence to generate a segmented optimal control sequence. Whenever new data arrives, the control system will update the control quantity and re-optimize to generate a new control sequence. The online rolling update process ensures that the control system responds to temperature changes and environmental changes in real time, thereby maintaining the accuracy of temperature control. The generated segmented optimal control sequence is divided according to the time dimension to obtain control instructions in different time periods. For example, assuming a control sequence , through rolling update, the segmented optimal control sequence is obtained , each Control instructions corresponding to a time period. According to the segmented optimal control sequence, the TEC devices are grouped and controlled. Control instructions are assigned between control groups in order of priority, thereby ensuring that each control group is adjusted in time according to the urgency of its temperature change. For example, if the temperature change of the first control group exceeds the set threshold, the control instructions are assigned to this group first to ensure that it reaches the target temperature as soon as possible. The control commands executed in parallel are converted into PWM (pulse width modulation) drive signals through the current modulation circuit to drive the TEC device for precise temperature regulation. The PWM signal adjusts the average value of the current by controlling the switching frequency (usually tens of kilohertz), thereby achieving precise control of the TEC power supply current. Assume that the duty cycle of the PWM signal is , then the output current of the control signal It is expressed as: ; in, is the maximum supply current. By adjusting the duty cycle , accurately control the current, thereby achieving closed-loop regulation of temperature.

[0034] The above describes the temperature control method based on TEC semiconductor in the embodiment of the present invention. The following describes the temperature control device based on TEC semiconductor in the embodiment of the present invention. Figure 2 In one embodiment of the present invention, a temperature control device based on TEC semiconductor includes: The acquisition module is used to perform distributed temperature acquisition and preprocessing on the temperature-controlled area, obtain preprocessed temperature data, and calculate the corresponding temperature correlation matrix; A detection module, used for inputting the pre-processed temperature data into a Gaussian mixture model for temperature anomaly detection to obtain a temperature anomaly feature matrix; A calculation module, used for calculating the temperature influence range of the TEC controller according to the temperature correlation matrix and the temperature anomaly characteristic matrix, dividing the multiple TEC controllers into multiple control groups based on the temperature influence range, and mapping the control groups to a virtual control network; An optimization module is used to input the temperature distribution data of the virtual control network into a mixed integer linear programming model for global temperature optimization, obtain the global optimal control variables, and divide the global optimal control variables into local control instructions according to the control groups; A building module is used to build a temperature control cost function based on local control instructions and obtain TEC control parameters through gradient iterative optimization calculation; The execution module is used to input the TEC control parameters into the temperature dynamic prediction model, perform robust optimization within the rolling prediction interval to obtain the local optimal control sequence, and execute temperature closed-loop control.

[0035] Through the collaborative cooperation of the above-mentioned components, by introducing distributed temperature acquisition and preprocessing technology, combined with the calculation method of the temperature correlation matrix, the data processing overhead of the system is significantly reduced, while the accuracy and reliability of the temperature data are guaranteed; the Gaussian mixture model is used for temperature anomaly detection, and the construction of the temperature anomaly feature matrix is ​​combined to improve the system's recognition ability and processing efficiency of temperature anomalies; the TEC controller grouping strategy based on the temperature influence range realizes the reasonable allocation of control resources and reduces the mutual interference between controllers; through the construction of a virtual control network and the application of a mixed integer linear programming model, the optimal control of the global temperature field is realized, and the overall control accuracy of the system is improved; the TEC control parameter calculation method based on gradient iterative optimization is adopted to ensure the optimality of the control parameters and the dynamic response performance of the system; the introduction of a temperature dynamic prediction model and a rolling time domain optimization strategy enhances the system's anti-interference ability and control stability; through the closed-loop control of the PWM drive signal, the accuracy and reliability of temperature control are ensured.

[0036] Reference Figure 3 In an embodiment of the present invention, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 3As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0037] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0038] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0039] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.

[0040] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0041] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0042] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A temperature control method based on TEC semiconductor, characterized in that: The method comprises: Perform distributed temperature collection and preprocessing on the temperature-controlled area to obtain preprocessed temperature data and calculate the corresponding temperature correlation matrix; Inputting the preprocessed temperature data into a Gaussian mixture model to perform temperature anomaly detection to obtain a temperature anomaly feature matrix; Calculating the temperature influence range of the TEC controller according to the temperature correlation matrix and the temperature anomaly characteristic matrix, dividing a plurality of TEC controllers into a plurality of control groups based on the temperature influence range, and mapping the control groups to a virtual control network; Inputting the temperature distribution data of the virtual control network into a mixed integer linear programming model for global temperature optimization to obtain a global optimal control variable, and dividing the global optimal control variable into local control instructions according to the control group; Based on the local control instructions, a temperature control cost function is constructed, and a TEC control parameter is obtained by gradient iterative optimization calculation; The TEC control parameters are input into the temperature dynamic prediction model, robust optimization is performed within the rolling prediction interval to obtain a local optimal control sequence, and temperature closed-loop control is performed.

2. The temperature control method based on TEC semiconductor according to claim 1, characterized in that: The distributed temperature collection and preprocessing of the temperature-controlled area to obtain preprocessed temperature data and calculate the corresponding temperature correlation matrix include: Calculating the temperature gradient distribution per unit area of ​​the temperature-controlled area, and arranging the temperature collection points to obtain a temperature collection point distribution sequence; Setting a data collection time interval and a data sampling frequency for each temperature collection point in the temperature collection point distribution sequence, and collecting original temperature data; Performing normalization calculation on the original temperature data to generate a standardized temperature value sequence; Building a sliding time window based on the standardized temperature value sequence, and performing weighted average calculation on the temperature values ​​in each sliding time window to obtain a sliding average temperature sequence; Performing data smoothing processing on the sliding average temperature series to obtain preprocessed temperature data; Construct an M×N dimensional temperature data array based on the spatial coordinates of the temperature collection points, wherein M is the number of temperature collection points and N is the length of the sampling time series; The Pearson correlation coefficient is calculated for the M×N dimensional temperature data array to obtain a temperature correlation coefficient matrix, and the temperature correlation coefficient matrix is ​​weighted superimposed with a preset spatial distance weight to obtain a temperature correlation matrix.

3. The temperature control method based on TEC semiconductor according to claim 2, characterized in that: The pre-processed temperature data is input into a Gaussian mixture model for temperature anomaly detection to obtain a temperature anomaly feature matrix, including: Dividing the preprocessed temperature data into temperature intervals according to preset temperature intervals, and calculating the occurrence frequency of the data in each temperature interval to obtain a temperature distribution histogram; Performing probability density distribution calculation on the temperature distribution histogram to generate a probability density distribution curve of temperature data; The probability density distribution curve is input into a Gaussian mixture model, and parameter iterative calculation is performed using an expectation maximization algorithm to obtain parameters of the Gaussian mixture model; Calculate the conditional probability of each temperature point based on the mixed Gaussian model parameters, mark the temperature data with a conditional probability value lower than the target value as an abnormal point, and obtain an initial abnormality detection result; For each abnormal point in the initial abnormality detection result, a gradient value between the abnormal point and an adjacent temperature point is calculated, and abnormality levels are divided according to the gradient values ​​to obtain an abnormal temperature level sequence; Performing spatial correlation analysis on the abnormal temperature level sequence and detection data of adjacent temperature sensors to obtain a target abnormality detection result; The spatial distribution characteristics, temperature deviation characteristics and time duration characteristics of the abnormal points in the target abnormality detection results are extracted, and a temperature abnormality feature matrix is ​​constructed.

4. The temperature control method based on TEC semiconductor according to claim 3, characterized in that: The step of calculating the temperature influence range of the TEC controller according to the temperature correlation matrix and the temperature anomaly characteristic matrix, dividing a plurality of TEC controllers into a plurality of control groups based on the temperature influence range, and mapping the control groups to a virtual control network includes: Performing singular value decomposition on the temperature correlation matrix according to the main diagonal to obtain a temperature space mapping matrix; Performing a convolution operation on the temperature space mapping matrix and the temperature anomaly characteristic matrix to generate a temperature response characteristic diagram of the TEC controller; Calculating a temperature influence coefficient for each TEC controller based on the temperature response characteristic diagram, and determining a region where the temperature influence coefficient is greater than a preset coefficient threshold as a temperature influence range; Constructing a TEC controller association matrix according to the overlap between the temperature influence ranges, and performing K-means clustering on the TEC controller association matrix to obtain a TEC control grouping result; Constructing a hierarchical control structure according to the TEC control grouping result, assigning a unique identifier to each control group in the hierarchical control structure, establishing a master-slave relationship between controllers within the group, and generating a control group management model; Based on the control group management model, virtual nodes are constructed, the temperature control characteristics of each control group and the relationship between groups are mapped to the virtual space to form an initial virtual network, and the initial virtual network is dynamically optimized. The data transmission path between nodes is set according to the direction of the temperature gradient to obtain a virtual control network.

5. The temperature control method based on TEC semiconductor according to claim 4, characterized in that: The step of inputting the temperature distribution data of the virtual control network into a mixed integer linear programming model for global temperature optimization to obtain a global optimal control variable, and dividing the global optimal control variable into local control instructions according to the control group includes: Constructing an A×B dimensional state matrix for the temperature data of each node in the virtual control network, where A is the number of control groups and B is the sampling time length, to obtain a temperature distribution characteristic matrix; The temperature distribution characteristic matrix is ​​input into a linear prediction model, and the temperature change trend is calculated based on the least square method to obtain a temperature prediction sequence; Based on the temperature prediction sequence, a global temperature optimization objective function is constructed, the temperature control error weight is set to 0.7, the control amount change weight is set to 0.3, and a mixed integer linear programming model is obtained; Solving the mixed integer linear programming model by interior point method to obtain global optimal control variables, performing dimensionality reduction decomposition on the global optimal control variables, mapping the control quantities to each TEC unit according to the spatial distribution of the control group, and obtaining initial control instructions; Performing a matching operation on the initial control instruction and the topological structure of the virtual control network to obtain a coordinated control sequence, and slicing the coordinated control sequence in a time dimension to generate a control subsequence, adding an inter-group synchronization mark, and obtaining a distributed control instruction; The distributed control instructions are grouped according to the identifier of the control group, and an execution timestamp is added to each group of instructions to obtain local control instructions.

6. The temperature control method based on TEC semiconductor according to claim 5, characterized in that: The temperature control cost function is constructed based on the local control instruction, and the TEC control parameters are obtained by gradient iterative optimization calculation, including: Creating a temperature control constraint matrix based on the local control instruction, and inputting the temperature control constraint matrix into a weighted cost function generator, assigning a weight of 0.6 to the temperature control error term and a weight of 0.4 to the abnormal state term, to obtain a temperature control cost function; According to the temperature control cost function, the TEC power supply current value is subjected to piecewise linearization processing to obtain a current control quantization table, and according to the current control quantization table, a response function between the current value and the temperature change rate is established to obtain a TEC temperature response model; Performing gradient descent optimization based on the TEC temperature response model to obtain a current parameter optimization sequence, calculating a temperature adjustment step based on the current parameter optimization sequence, dividing the temperature change by a control period, and generating a standardized temperature adjustment parameter; The standardized temperature adjustment parameters are subjected to stability analysis, the upper limit of the temperature adjustment step and the range of the control cycle are set, parameter constraints are obtained, and the parameter constraints are matched and screened with the current parameter optimization sequence, and the optimal parameter combination that meets the constraints is selected as the TEC control parameter.

7. The temperature control method based on TEC semiconductor according to claim 6, characterized in that: The TEC control parameters are input into the temperature dynamic prediction model, robust optimization is performed within the rolling prediction interval to obtain a local optimal control sequence, and temperature closed-loop control is performed, including: An A×3-dimensional control matrix is ​​constructed for the TEC control parameters, where A is the number of control groups, and the three dimensions correspond to the TEC power supply current value, the control cycle, and the temperature adjustment step, respectively, to obtain an initial matrix of control parameters; Perform parameter cascading on the control parameter initial matrix and the temperature anomaly characteristic matrix to construct a state vector; Inputting the state vector into a temperature dynamic prediction model, setting a rolling time domain for the state vector through the temperature dynamic prediction model, and performing temperature state prediction to obtain a prediction matrix; The prediction matrix is ​​input into a robust optimization controller, a temperature control deviation threshold is set, a temperature optimization objective function is generated, and a quadratic programming problem is constructed based on the temperature optimization objective function, and a sequential quadratic programming algorithm is used to solve the problem to obtain a control quantity sequence; Performing online rolling update on the control quantity sequence to generate a segmented optimal control sequence, and performing group control on the TEC devices according to the segmented optimal control sequence, distributing control instructions in order of priority, and obtaining control commands for parallel execution; The control commands executed in parallel are converted into PWM drive signals through a current modulation circuit, and temperature closed-loop control is performed using a preset switching frequency.

8. A temperature control device based on TEC semiconductor, characterized in that: For executing the TEC semiconductor-based temperature control method according to any one of claims 1 to 7, the TEC semiconductor-based temperature control device comprises: The acquisition module is used to perform distributed temperature acquisition and preprocessing on the temperature-controlled area, obtain preprocessed temperature data, and calculate the corresponding temperature correlation matrix; A detection module, used for inputting the preprocessed temperature data into a Gaussian mixture model to perform temperature anomaly detection and obtain a temperature anomaly feature matrix; a calculation module, configured to calculate a temperature influence range of the TEC controller according to the temperature correlation matrix and the temperature anomaly characteristic matrix, divide a plurality of TEC controllers into a plurality of control groups based on the temperature influence range, and map the control groups to a virtual control network; An optimization module, used for inputting the temperature distribution data of the virtual control network into a mixed integer linear programming model for global temperature optimization, obtaining a global optimal control variable, and dividing the global optimal control variable into local control instructions according to the control group; A construction module is used to construct a temperature control cost function based on the local control instruction, and obtain TEC control parameters through gradient iterative optimization calculation; The execution module is used to input the TEC control parameters into the temperature dynamic prediction model, perform robust optimization within the rolling prediction interval to obtain a local optimal control sequence, and perform temperature closed-loop control.

9. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and wherein the processor implements the temperature control method based on TEC semiconductor according to any one of claims 1 to 7 when executing the computer program. 10 . A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the processor is caused to execute the TEC semiconductor-based temperature control method according to claim 1 .

Images