Laboratory energy-saving optimization control system and method based on energy efficiency model driving

Abstract

The invention provides a laboratory energy-saving optimization control system and method based on energy efficiency model driving, and relates to the technical field of intelligent control, and the method comprises the steps: recognizing a core coordinate of a temperature overrun region of a hazardous chemical substance storage region based on a fusion data set, and taking the core coordinate as a first point, identifying the oxidation reaction rate peak position coordinate of the hazardous waste treatment area as a second point, and identifying the boundary centroid coordinate of the abnormal heat dissipation area of the high-energy-consumption equipment as a third point; constructing a key point coordinate set based on the first point, the second point and the third point, and connecting the key points to generate a static polygon monitoring domain; performing grid discretization processing by taking the static polygon monitoring domain as a boundary to generate structured grid data containing vertex coordinates and unit adjacency relations; deformation parameters of each unit are calculated, and a non-uniform deformation field scalar operator is generated through fusion; and taking the scalar operator as a physical field coupling weight, and generating an optimization instruction set. According to the invention, intelligent control of equipment is realized.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control technology, and in particular to a laboratory energy-saving optimization control system and method driven by an energy efficiency model. Background Art

[0002] Laboratory environments typically include a variety of high-energy-consuming equipment (such as fume hoods, incubators, ovens, analytical instruments, etc.), environmentally sensitive hazardous chemical storage areas, and hazardous waste treatment areas involving chemical reactions. These areas have strict requirements on environmental parameters such as temperature, humidity, and airflow, and their operating states influence each other and change dynamically, making laboratory energy consumption management face some challenges.

[0003] Some of the current laboratory energy-saving control technologies have the following limitations: For example, some traditional energy-saving control methods treat the laboratory as a whole or simply partition it (e.g., by room or function), failing to accurately identify and focus on core micro-areas with energy consumption anomalies, safety risks, or efficiency bottlenecks. For example, localized temperature exceeding limit points in hazardous chemical storage cabinets, hot spots with the most intense oxidation reactions in hazardous waste treatment equipment, or localized high-temperature areas caused by poor heat dissipation in large instruments—these critical points are often obscured by the average values ​​of overall environmental parameters. This broad monitoring approach results in a lack of targeted control instructions.

[0004] For example, existing energy efficiency models are mostly constructed based on static or quasi-static assumptions, and fail to fully consider the strong dynamic characteristics of the spatiotemporal coupling of multiple physical fields in the laboratory (thermal field, flow field, concentration field, etc.). Factors such as environmental parameter fluctuations, equipment start-up and shutdown, and changes in chemical reaction processes will change the energy flow and distribution pattern. Static models are difficult to capture these rapid changes, resulting in delays or failures in optimization strategies. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a laboratory energy-saving optimization control system and method driven by an energy efficiency model, which realizes the refined division of the laboratory space. The independent analysis of each triangular unit can capture subtle energy consumption anomalies or environmental changes in the space.

[0006] In order to solve the above technical problems, the technical solutions of the present invention are as follows: A laboratory energy-saving optimization control method driven by an energy efficiency model, the method comprising: Collect multi-physics field data of laboratory environment and equipment operating parameters, perform spatiotemporal fusion processing, and generate fused data sets; Based on the fused data set, the core coordinates of the temperature-exceeding area in the hazardous chemical storage area are identified as the first point, the coordinates of the peak position of the oxidation reaction rate in the hazardous waste treatment area are identified as the second point, and the coordinates of the centroid of the boundary of the abnormal heat dissipation area of ​​high-energy-consuming equipment are identified as the third point. A key point coordinate set is constructed based on the first, second, and third points, and the key points are connected to generate a static polygon monitoring domain. Using the static polygon monitoring domain as the boundary, the grid is discretized to generate structured grid data containing vertex coordinates and unit adjacency relationships; Based on the grid cell characteristics of the key point area, the deformation parameters of each cell are calculated and fused to generate a non-uniform deformation field scalar operator; Using scalar operators as physical field coupling weights, a dynamic energy efficiency optimization functional is constructed; Combined with the real-time changes in environmental parameters, the final solution of the dynamic energy efficiency optimization functional is solved and an optimization instruction set is generated; Send the optimized instruction set to the execution terminal.

[0007] A laboratory energy-saving optimization control system driven by an energy efficiency model, including: The acquisition module is used to collect multi-physics field data of the laboratory environment and equipment operating parameters, perform spatiotemporal fusion processing, and generate a fused data set; A processing module is used to identify the core coordinates of the temperature exceeding the limit area in the hazardous chemical storage area as the first point, identify the coordinates of the peak position of the oxidation reaction rate in the hazardous waste treatment area as the second point, and identify the centroid coordinates of the boundary of the abnormal heat dissipation area of ​​the high-energy-consuming equipment as the third point based on the fused data set; construct a key point coordinate set based on the first point, the second point, and the third point, and connect the key points to generate a static polygon monitoring domain; A generation module is used to perform grid discretization processing based on the static polygon monitoring domain as the boundary, and generate structured grid data containing vertex coordinates and unit adjacency relationships; The calculation module is used to calculate the deformation parameters of each unit based on the grid unit characteristics of the key point area, and fuse them to generate a non-uniform deformation field scalar operator; A building block for constructing dynamic energy efficiency optimization functionals using scalar operators as physical field coupling weights; The sending module is used to combine the real-time changes in environmental parameters to solve the final solution of the dynamic energy efficiency optimization functional and generate an optimization instruction set; and send the optimization instruction set to the execution terminal.

[0008] The above solution of the present invention includes at least the following beneficial effects: By collecting multi-physical field data of the laboratory environment (such as temperature, humidity, gas concentration, etc.) and equipment operating parameters, and after spatiotemporal fusion processing, a high-precision data set covering the entire laboratory is formed. This data set can reflect the multi-dimensional status of hazardous chemical storage, hazardous waste treatment and equipment operation in real time, avoiding monitoring blind spots caused by data fragmentation.

[0009] Based on the fused data set, the temperature exceeding the limit points in the hazardous chemical storage area, the oxidation reaction peak points in the hazardous waste treatment area, and the abnormal heat dissipation areas of high-energy consumption equipment are identified, and a key point coordinate set is constructed. This can concentrate monitoring resources on high-risk and high-energy consumption areas. Compared with traditional full-area uniform monitoring, the amount of invalid data collection is reduced, and the monitoring accuracy of key parameters (such as hazardous chemical temperature and reaction rate) is improved.

[0010] Structured grid data is generated with the static polygon monitoring domain as the boundary, and the vertex coordinates and unit adjacency relationship are determined to provide a standardized framework for physical field coupling calculations. The grid can adapt to the complex spatial structure of the laboratory (such as fume hoods and equipment array layouts) and meet the refined control needs of precision instrument areas; the deformation parameters are calculated based on the grid unit characteristics of the key point area, and dynamic weight factors are generated. This non-uniform processing method can improve the response speed to key areas while reducing the computational load in non-key areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 The present invention provides a flow chart of a laboratory energy-saving optimization control method driven by an energy efficiency model.

[0012] Figure 2 This is a schematic diagram of a laboratory energy-saving optimization control system driven by an energy efficiency model provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0013] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0014] like Figure 1 As shown, an embodiment of the present invention proposes a laboratory energy-saving optimization control method driven by an energy efficiency model, the method comprising the following steps: Step 1: Collect multi-physics field data of the laboratory environment and equipment operating parameters, perform spatiotemporal fusion processing, and generate a fused data set; Step 2: Based on the fused dataset, identify the core coordinates of the temperature-exceeding zone in the hazardous chemical storage area as the first point, identify the coordinates of the peak position of the oxidation reaction rate in the hazardous waste treatment area as the second point, and identify the centroid coordinates of the boundary of the abnormal heat dissipation zone of the high-energy-consuming equipment as the third point. A key point coordinate set is constructed based on the first, second, and third points, and the key points are connected to generate a static polygon monitoring domain. Step 3: Using the static polygon monitoring domain as the boundary, perform grid discretization processing to generate structured grid data containing vertex coordinates and unit adjacency relationships; Step 4: Based on the grid cell characteristics of the key point area, calculate the deformation parameters of each cell and fuse them to generate a non-uniform deformation field scalar operator; Step 5: Use the scalar operator as the physical field coupling weight to construct a dynamic energy efficiency optimization functional; Step 6: Combined with the real-time changes in environmental parameters, the final solution of the dynamic energy efficiency optimization functional is solved and an optimization instruction set is generated; Step 7: Send the optimized instruction set to the execution terminal.

[0015] In an embodiment of the present invention, by collecting multi-physical field data of the laboratory environment (such as temperature, humidity, gas concentration) and equipment operating parameters (power, speed, etc.) and performing spatiotemporal fusion, fragmented, multi-source heterogeneous data can be integrated into a fused data set in a unified format. On the one hand, the time delay and spatial misalignment problems between data are eliminated, and different types of data can be accurately matched, thereby improving the availability and accuracy of data; on the other hand, a complete and accurate data basis is provided for subsequent key area identification, energy efficiency analysis, etc., avoiding misjudgment caused by incomplete or inaccurate data, and ensuring the effectiveness of subsequent control strategies. For example, the fused data set can clearly show the impact of high-energy-consuming equipment on the surrounding environment temperature and humidity during operation, providing a basis for precise regulation, and improving data utilization compared to traditional single data collection methods.

[0016] Based on the fused data set, the temperature exceeding the limit area in the hazardous chemical storage area, the peak position of the oxidation reaction rate in the hazardous waste treatment area, and the abnormal heat dissipation area of ​​high-energy consumption equipment are accurately identified. A key point coordinate set is constructed and a static polygon monitoring domain is generated. This step focuses on the high-risk and high-energy consumption key areas of the laboratory to achieve precise allocation of monitoring resources. Compared with traditional full-area uniform monitoring, it can reduce the workload of ineffective monitoring and concentrate monitoring resources on areas that really need attention. At the same time, the monitoring accuracy of core parameters in key areas (such as hazardous chemical temperature and oxidation reaction rate) is significantly improved. In addition, the static polygon monitoring domain provides a definite boundary range for subsequent grid discretization and energy efficiency analysis, ensuring the targeted calculation and control and improving monitoring and control efficiency.

[0017] Using a static polygonal monitoring domain as its boundary, the grid is discretized to generate structured grid data containing vertex coordinates and cell adjacency relationships, providing a standardized, structured framework for complex physical field calculations. Structured grid data can adapt to the complex spatial structures of laboratories, such as fume hoods and lab bench layouts, enabling subsequent energy efficiency optimization calculations to have a higher spatial resolution (up to 0.5m×0.5m), meeting the need for refined control of laboratory precision instrument areas. Furthermore, the defined vertex coordinates and cell adjacency relationships facilitate physical field coupling calculations, significantly reducing computational complexity, improving efficiency, and reducing computational time compared to unstructured grids.

[0018] The deformation parameters are calculated based on the grid unit characteristics of the area where the key points are located, and the scalar operators of the non-uniform deformation field are generated by fusion to achieve differentiated treatment of different key areas. For example, in the hazardous chemical storage area, the scalar operator can focus on the coupling relationship between temperature and energy consumption; in the high-energy consumption equipment area, it focuses on the matching relationship between heat dissipation and power. This non-uniform processing method enables the model to dynamically adjust the calculation weights according to the characteristics of different areas, and the response speed to changes in key areas is improved by about 50%. At the same time, the computational load of non-key areas is reduced, which reduces the computing power consumption. Under the premise of ensuring control accuracy, the model operation efficiency and resource utilization are effectively improved.

[0019] By using scalar operators as physical field coupling weights to construct a dynamic energy efficiency optimization functional, they achieved comprehensive consideration and coordinated optimization of multiple laboratory physical fields (such as temperature, airflow, and energy consumption). By dynamically adjusting the physical field coupling weights, they were able to balance safety and energy conservation goals based on different scenarios and needs. For example, when oxidation reactions intensified in the hazardous waste treatment area, the system could automatically increase ventilation to ensure safety while optimizing air conditioning operation strategies to reduce energy consumption.

[0020] In another preferred embodiment of the present invention, a variety of sensors are deployed in key areas of the laboratory for data collection, and high-precision temperature sensors and gas concentration sensors are installed in the hazardous chemical storage area to monitor the temperature changes of the storage environment and the gas concentration generated by the volatilization of hazardous chemicals in real time; oxidation reaction rate sensors and humidity sensors are arranged in the hazardous waste treatment area to accurately capture the rate fluctuations and environmental humidity conditions during the reaction process; for high-energy consumption equipment, temperature sensors and power sensors are installed at their heat dissipation ports and power supplies to obtain the equipment heat dissipation temperature and operating power data. These sensors continuously collect environmental multi-physical field data and equipment operating parameters according to the set sampling frequency.

[0021] Collected data is transmitted to the data processing center via wired or wireless networks. There, the raw data undergoes preliminary cleaning to remove outliers and duplicates caused by sensor failures and network fluctuations. Next, each piece of data is timestamped and spatially tagged. The timestamp precisely records the moment of data collection, while the spatial tag identifies the corresponding collection location (e.g., the third floor of rack A in the hazardous chemicals storage area or the heat dissipation outlet of high-energy-consuming equipment X).

[0022] After data labeling is complete, spatiotemporal fusion processing is performed. Data of different types and locations at the same time are associated and integrated based on timestamps. Data from adjacent locations or physically related areas are spatially aligned based on spatial tags. For example, the operating power and heat dissipation temperature of high-energy-consuming equipment can be integrated with the temperature and airflow velocity data of the surrounding environment at the same time. Ultimately, a fused dataset containing multi-dimensional information and unified time and space is generated.

[0023] In a preferred embodiment of the present invention, the first point is determined by the following process: Extract the temperature monitoring data of hazardous chemical storage units from the fused dataset, identify the spatial area where the temperature continuously exceeds the preset safety threshold, calculate the minimum bounding rectangle of the spatial area, and use its center coordinates as the first point; The second point is determined by the following process: Analyze the gas concentration time series data of the hazardous waste treatment area in the fused dataset; calculate the gas concentration change rate of adjacent time slices, locate the spatial position with the maximum change rate, and take the weighted centroid coordinates within a preset radius as the second point with this position as the center; The third point is determined by the following process: Acquire infrared thermal imaging data of high-energy consumption equipment, segment the abnormal heat dissipation area based on the temperature gradient threshold, extract the closed boundary of the area, and calculate the coordinates of the centroid of the closed boundary polygon as the third point.

[0024] In the embodiment of the present invention, the process of determining the first point (core coordinates of the temperature-exceeding-limit area in the hazardous chemicals storage area) is as follows: In the fused data set, the temperature monitoring data of all hazardous chemical storage units is first extracted through the data filtering function. This data covers the temperature values ​​of different storage units at various time nodes. Then, based on the hazardous chemical safety storage standards, a preset safety threshold is set as the basis for judging whether the temperature exceeds the limit. The system continuously monitors these temperature data. When it is found that the temperature in a certain spatial area exceeds the preset safety threshold for multiple consecutive times, the area can be determined to be a temperature-exceeding area.

[0025] After determining the out-of-limit area, in order to more accurately mark its location, the spatial coordinate point set of the temperature out-of-limit area is extracted from the fused data set. These points represent the specific locations where the temperature continuously exceeds the standard (such as the three-dimensional coordinates of hazardous chemical storage racks); the coordinate point set is denoised to eliminate outliers (such as isolated points caused by sensor errors), and the valid points in the continuous out-of-limit area are retained through the neighborhood density screening method to ensure that subsequent calculations are based on the actual temperature anomaly area.

[0026] The denoised coordinate point set is converted into a two-dimensional plane projection (e.g., the storage area floor is used as the XY plane). A rasterization method is used to fill the discrete points into a continuous region, generating a binary image (where the temperature exceeding the limit is white and the background is black). An edge detection algorithm (e.g., the Canny operator) is used to extract the closed contour of the exceeding temperature range, forming a continuous boundary curve to ensure that the contour completely encloses all temperature exceeding the limit points. For initial rectangle fitting, the minimum axial enclosing rectangle (AABB) of the contour point set is first calculated, i.e., the minimum rectangle parallel to the coordinate axis, which serves as the iteration benchmark. For rotation search optimization, the contour is rotated around its geometric center within the range of 0°-90° with a preset step size (e.g., 1°). The AABB is recalculated after each rotation, and the area of ​​the rectangle at the current rotation angle is recorded. For minimum area determination, after traversing all rotation angles, the rectangle with the smallest area is selected as the minimum bounding rectangle (OBB). The edges of this rectangle are aligned with the main direction of the exceeding temperature range, which can more compactly enclose the target area.

[0027] After determining the coordinates of the four vertices of the minimum bounding rectangle, calculate the coordinates of the intersection of the two diagonals of the rectangle. This intersection is the geometric center of the rectangle, and its coordinate value is obtained by taking the arithmetic average of the vertex coordinates (for example, the X coordinate is the average of the X values ​​of the two diagonal vertices, and the same is true for the Y coordinate). If the rectangle is a three-dimensional area in three-dimensional space, the midpoint between the minimum and maximum Z coordinates (in the height direction) can be used as the center Z coordinate, ultimately forming the core coordinate point in three-dimensional space.

[0028] The process of determining the second point above (the coordinates of the peak position of the oxidation reaction rate in the hazardous waste treatment area) is as follows: The time series data of gas concentrations in the hazardous waste treatment area were screened out from the fused dataset. The data covered the measurement values ​​of all gas concentration monitoring points in the area at different timestamps, accurately recording the dynamic changes of concentrations at each point over time. The time slice division rules were set based on the process characteristics of laboratory hazardous waste treatment and the data sampling frequency. For example, if the sampling frequency is once per minute, the data can be divided into a time slice every 15 minutes. If the reaction process changes dramatically, the time slice length is shortened to 5 minutes to ensure that the data in each time slice has relatively stable statistical characteristics. After the division is completed, each time slice contains a gas concentration data matrix at different monitoring points within a continuous time period.

[0029] Process the data in each time slice point by point: To calculate the difference between adjacent time points, starting with the first monitoring point within a time slice, the gas concentration difference between adjacent time points is calculated sequentially. For example, the concentration difference between the nth and n+1th time points within a time slice is the concentration value at the nth time point minus the concentration value at the nth time point. To normalize time intervals, since the data sampling interval is fixed (e.g., 1 minute), the concentration difference is directly divided by the fixed time interval to obtain the concentration change rate between adjacent time points. To integrate the change rates within a time slice, the above steps are repeated to calculate the change rates for all adjacent time points within the time slice, ultimately forming a sequence containing multiple change rate values. By comparing the values ​​in the sequence, the maximum change rate within the time slice and its corresponding time point are found. All time slices are traversed, and the maximum change rate values ​​within each time slice are compared. The time slice with the maximum change rate is identified. Within this time slice, the location of the monitoring point with the largest change rate is determined based on the spatial coordinates of the gas concentration monitoring point. If multiple monitoring points have the same maximum change rate, the monitoring point with the coordinates closest to the center of the region is selected as the core location, which is used as the central coordinate for subsequent weighted calculations.

[0030] With the located core location as the center, a circular range with a preset radius is delineated (the radius can be set according to the area of ​​the hazardous waste treatment area and the density of monitoring points, such as 1 meter or 2 meters). Each monitoring point within the range participates in the weighted calculation, and the weight assignment rules are as follows: Data accuracy weights are used to grade the reliability of monitoring point data based on the sensor's historical calibration history and failure frequency. For example, a sensor with no failures in the past three months and regularly calibrated will have a data accuracy weight of 0.8; a sensor with one or two failure records will have a weight of 0.6; and a sensor with frequent failures will have a weight of 0.3.

[0031] Distance weight: the closer the distance between the monitoring point and the core location, the greater the impact on the weighted result. The levels are divided according to the distance: for monitoring points with a distance less than 0.5 times the radius, the distance weight is set to 0.7; for monitoring points with a distance of 0.5-1 times the radius, the weight is set to 0.5; for monitoring points with a distance greater than 1 times the radius, the weight is set to 0.3; for comprehensive weight calculation, the data accuracy weight of each monitoring point is multiplied by the distance weight to obtain the final comprehensive weight. For example, if the data accuracy weight of a monitoring point is 0.8 and the distance weight is 0.7, then its comprehensive weight is 0.8×0.7=0.56; the weighted centroid solution is based on the comprehensive weight of each monitoring point and its spatial coordinates, and the coordinates are weighted averaged, that is, the X, Y, and Z coordinates are weighted and summed respectively, and then divided by the sum of all weights. The final coordinates are the weighted centroid coordinates, which are used as the second point to characterize the peak position of the oxidation reaction rate; the calculated weighted centroid coordinates are compared with the historical reaction peak data. If the coordinates are in the high concentration change area during multiple periods of intense oxidation reactions, their validity is confirmed; if there is a significant deviation, the time slice division, weight setting and other links are rechecked, and the parameters are adjusted and recalculated until accurate coordinates that conform to the actual reaction law are obtained.

[0032] The process of determining the third point above (the coordinates of the centroid of the boundary of the abnormal heat dissipation area of ​​high-energy-consuming equipment) is as follows: Infrared thermal imagers deployed around high-energy-consuming equipment collect real-time temperature distribution image data on the equipment surface. Since the original image may contain noise (such as pixel anomalies caused by environmental heat source interference and instrument errors), it is necessary to preprocess it first. Using the median filtering algorithm, the pixel value of each pixel in the image is replaced with the median of the pixel values ​​in its neighborhood, effectively eliminating salt and pepper noise. If Gaussian noise is present, the image is smoothed using Gaussian filtering. Histogram equalization technology is used to expand the grayscale dynamic range of the image, making areas with smaller temperature differences easier to distinguish in the image and improving the recognition of abnormal areas. Referring to the normal operating temperature range provided by the equipment manufacturer and combining it with the equipment's historical operating data, the heat dissipation temperature distribution of the equipment under different operating conditions is analyzed. For example, if the surface temperature of the equipment is between 30°C and 50°C during normal operation, the threshold can be set to 55°C to take into account environmental fluctuations and equipment aging factors. If the temperature of the equipment is low at the beginning of operation and gradually rises later, a dynamic threshold is used and adaptively adjusted according to the mean and standard deviation of recent temperature data; for the preprocessed infrared thermal imaging image, its temperature value is compared with the set threshold pixel by pixel, and pixels with temperatures above the threshold are marked white (representing abnormal heat dissipation areas), and pixels with temperatures below the threshold are marked black (representing normal heat dissipation areas), thereby segmenting the image into a black and white binary image, clearly demarcating abnormal and normal areas.

[0033] The binary image is processed using an image edge detection algorithm (such as the Sobel operator or the Canny operator). These algorithms calculate the gradient amplitude and direction of the pixel points in the image to identify the edges where the pixel values ​​change significantly, thereby extracting the contour lines of the abnormal heat dissipation area and checking whether the extracted contour is closed. If there are breakpoints or gaps, morphological processing methods (such as dilation and erosion operations) are used to repair the boundary. The dilation operation can connect adjacent edge pixels to fill the gap; the erosion operation removes small burrs to make the boundary smoother. The above operations are repeated until a complete and continuous closed boundary is obtained; the coordinates of the pixel points on the closed boundary are connected in sequence to form a polygon; for the constructed polygon, the polygon is divided into multiple triangular sub-areas (this can be achieved by selecting any vertex and connecting it to other non-adjacent vertices), and the centroid coordinates and area of ​​each triangle are calculated respectively. The centroid coordinates are then weighted averaged according to the triangle area. The final coordinates are the centroid coordinates of the polygon, which is the third point and represents the center position of the abnormal heat dissipation area of ​​the high-energy-consuming equipment.

[0034] Through the above method, the present invention can quickly and accurately locate key areas in the laboratory where hazardous chemicals exceed the temperature limit, hazardous waste reactions are abnormal, and equipment heat dissipation failures. Compared with traditional manual inspections or extensive monitoring, the positioning accuracy can be improved, and safety hazards can be effectively avoided from expanding due to failure to discover them in time; focusing on key areas for precise positioning can concentrate monitoring resources on areas that really need attention, reduce ineffective monitoring work, improve monitoring efficiency and reduce monitoring costs, and make more efficient use of limited resources; after accurately obtaining the coordinates of key areas, more targeted management strategies can be formulated for these areas. For example, in hazardous chemical storage areas, temperature-exceeding areas can be discovered in time and cooling measures can be taken to reduce the risk of explosion and leakage of hazardous chemicals; in areas with abnormal heat dissipation of high-energy-consuming equipment, fault hazards can be discovered in advance, equipment operation can be optimized, energy consumption can be reduced, and the average operating efficiency of the equipment can be improved.

[0035] In a preferred embodiment of the present invention, a key point coordinate set is constructed based on the first point, the second point, and the third point, and a static polygon monitoring domain is generated by connecting the key points, including: Taking the key point coordinate set consisting of the first point, the second point and the third point as input, performing homogeneous transformation processing on the coordinates of each key point in three-dimensional space to obtain a transformed key node set in a unified coordinate system; For the transformed key node set, calculate the center of mass O of all nodes projected on the XOY plane, establish a polar coordinate system with O as the pole, calculate the azimuth of each node relative to O, and sort all nodes in increasing order of azimuth to obtain an ordered node sequence; Based on the ordered node sequence, a stack structure is initialized and a starting node with the smallest polar angle is pushed into the stack structure; the remaining nodes in the sequence are traversed in sequence, a vector cross product sign of the two nodes at the top of the stack and the currently traversed node is calculated, and the top element of the stack is iteratively adjusted according to the cross product sign until the ordered node sequence is traversed, and the node sequence retained in the stack is the vertex set constituting the convex hull; The vertex set of the convex hull is spatially sorted according to the right-hand spiral rule, and the connecting edges between adjacent vertices are smoothed to generate the static polygon monitoring domain for mesh discretization.

[0036] In the embodiment of the present invention, the specific implementation process of constructing a static polygon monitoring domain is as follows: Coordinate system 1: Since the first, second, and third points may come from different monitoring devices (such as temperature sensors in hazardous chemical storage areas, gas concentration sensors in hazardous waste treatment areas, and infrared thermal imagers in equipment), their original coordinates may be based on different local coordinate systems (such as the equipment's own coordinate system and the laboratory partition coordinate system). The global reference coordinate system must be determined first (such as the laboratory entrance as the origin, due east as the X-axis, due north as the Y-axis, and vertically upward as the Z-axis).

[0037] The homogeneous transformation operation converts the coordinates of each key point into homogeneous coordinates (i.e., adding a homogeneous component of 1 to the three-dimensional coordinates) through a translation transformation (moving the local origin to the global origin) and a rotation transformation (adjusting the coordinate axis direction to be consistent with the global coordinate system). For example, if the coordinates of a key point in the local coordinate system are (x', y', z'), the coordinates of its local origin in the global coordinate system are (x0, y0, z0), and the angle between the local X axis and the global X axis is θ, then the coordinates are first translated by (x'-x0, y'-y0, z'-z0), and then rotated about the Z axis by θ, to finally obtain the coordinates (x, y, z) in the global coordinate system.

[0038] Polar coordinate system establishment and node sorting: XOY plane projection centroid calculation: project all transformed key point coordinates onto the XOY plane (i.e., ignore the Z coordinate) and calculate the centroid O of the projection point set; polar angle calculation and sorting: establish a polar coordinate system with O as the pole and the positive X-axis of the XOY plane as the polar axis. For each key point projection point, calculate its azimuth relative to O (i.e., the angle between the line connecting the point and O and the positive X-axis); sort all nodes by azimuth angle from small to large. If the azimuth angles are the same, sort them by distance to O from near to far to form an ordered node sequence.

[0039] Convex hull vertex set generation (based on Graham scan algorithm): To initialize the stack, find the node with the smallest polar angle (i.e., the point closest to the positive X-axis) in the ordered node sequence and push it onto the stack. If multiple nodes have the same polar angle, choose the node farthest from O as the starting point. To determine the cross product and adjust the stack, start traversing the sequence from the second node, taking the first two nodes A and B at the top of the stack each time. The current traversed node is C. The direction of the cross product between vector AB and vector AC is calculated (by determining whether the rotation formed by the three points is clockwise or counterclockwise). If the cross product sign indicates a concave point (i.e., a clockwise rotation), pop the top node B from the stack. Repeat this determination until the cross product sign indicates a counterclockwise rotation, and then push C onto the stack. For example, if the three points A→B→C rotate clockwise, B is a concave point and must be removed from the convex hull. To generate the convex hull, after traversing all nodes, the remaining nodes in the stack are ordered to form the vertex set of the convex hull. This convex hull is the smallest convex polygon that can enclose all keypoints.

[0040] Spatial sorting and boundary smoothing: Sorting is done using the right-hand spiral rule, expanding the convex hull vertex set from the XOY plane to three-dimensional space. With the positive direction of the Z axis as the normal vector direction, the spatial order of the vertices is determined according to the right-hand spiral rule (that is, when the four fingers of the right hand rotate from the first vertex to the second vertex, the thumb points to the positive direction of the Z axis) to ensure that the orientation of the polygon faces is consistent, which facilitates subsequent mesh discretization. Boundary smoothing is performed, and Bezier curves or spline curves are used to smoothly fit the connecting edges between adjacent vertices. For example, 1-2 intermediate points are inserted between adjacent vertices A and B, and the position of the intermediate points is calculated based on the vertex coordinates and normal vectors, so that the polygon boundary is closer to the shape of the actual monitoring area and the impact of sharp corners on subsequent mesh division is reduced.

[0041] The polygons generated by the convex hull algorithm of the present invention can minimize the coverage of all key points (hazardous chemical over-limit areas, hazardous waste reaction peak areas, equipment heat dissipation abnormal areas), and can avoid the waste of monitoring resources in invalid areas compared to rectangular or circular monitoring domains; the unified coordinate system and ordered vertex set provide standardized input for grid discretization, which can improve the speed of structured grid generation, and the grid units have a higher degree of fit with the actual monitoring area; it adapts to complex three-dimensional space scenes, and through homogeneous transformation and spatial sorting, it can handle the high and low equipment layout in the laboratory (such as fume hoods, wall cabinets, etc.), and the generated polygonal monitoring domain can accurately reflect the spatial distribution of key areas in three-dimensional space; dynamic monitoring strategy optimization, the smoothed polygonal boundary can reduce singular units (such as narrow and long triangles) during grid discretization, which can improve the calculation accuracy of non-uniform deformation fields, and thereby improve the accuracy of solving energy efficiency optimization functionals.

[0042] In a preferred embodiment of the present invention, step 3, using the static polygon monitoring area as the boundary, performs mesh discretization processing to generate structured mesh data containing vertex coordinates and unit adjacency relationships, including: Step 31: Using the static polygon monitoring domain as the input domain, uniformly inserting discrete point sets along the boundary of the static polygon monitoring domain, and forming an initial discrete node set with the vertices of the static polygon monitoring domain and the inserted boundary discrete points; Step 32, based on the initial discrete node set, using the Delaunay triangulation algorithm for processing, including: generating a convex hull boundary containing all the initial discrete nodes, iteratively inserting internal diagonals based on the criterion of maximizing the minimum internal angle of the unit, and decomposing the static polygon monitoring domain into a set of non-overlapping triangular units; Step 33, assigning homogeneous coordinate attributes to each triangle unit in the triangle unit set, that is, expanding the three-dimensional space coordinates of the three vertices of each triangle unit into a four-dimensional homogeneous coordinate form; Step 34: Based on the geometric relationship between the units in the triangular unit set, identify the triangular unit pairs that share edges, and construct a sparse adjacency matrix that describes the adjacency topological relationship between all the triangular units. Step 35 : Integrate the four-dimensional homogeneous coordinate data of vertices of all units in the triangular unit set with the sparse adjacency matrix data to generate structured mesh data.

[0043] In the embodiment of the present invention, the specific implementation process of the above-mentioned grid discretization processing is as follows: In step 31 above, the density of inserted discrete points is set based on the complexity of the static polygon monitoring domain's boundaries and the accuracy requirements for subsequent calculations. For example, for boundaries with sides longer than 2 meters, a discrete point is inserted every 0.5 meters. If the boundary contains sharp corners (e.g., angles less than 60°), the density is increased to one point every 0.2 meters to ensure accurate representation of the boundary contour. Linear interpolation is performed along each edge of the polygon boundary to calculate the coordinates of the inserted points. The original vertices of the static polygon are merged with the newly inserted discrete points, and after deduplication, an initial discrete node set is formed. For example, if the polygon has 6 vertices and 3 points are inserted on each edge, the initial node set contains approximately 6 + 3 × the number of edges (after deduplication).

[0044] The above step 32, Delaunay triangulation and triangle element generation, is as follows: Convex hull boundary generation: Based on the initial discrete node set, the GiftWrap algorithm or the Andrew algorithm is used to generate a convex hull containing all nodes. For example, the lower leftmost node is first found as the starting point, and nodes are added in sequence by polar angle to form a minimum convex polygon boundary that wraps all nodes.

[0045] The internal diagonals are inserted iteratively, starting from the convex hull boundary, and the internal region is divided according to the following rules: Select an edge AB on the convex hull and find a node C that is not on AB, so that the interior angle of the triangle formed by ABC is minimized; check whether the triangle meets the Delaunay condition (that is, the triangle circumcircle does not contain other nodes). If so, insert the diagonal AB to split the convex hull into ABC and the remaining area; repeat the above process for the newly generated edges BC and AC until all the internal areas are divided into triangular units; during the division process, if the minimum interior angle of a triangle is less than 30°, adjust it by flipping the diagonal (for example, flipping the diagonal AD of the quadrilateral ABDC to BC) to ensure that the minimum interior angle of all triangles is maximized to prevent narrow and long units from affecting subsequent calculations.

[0046] The above step 33, the homogeneous coordinate expansion of the triangle unit, is as follows: The conversion from three-dimensional coordinates to four-dimensional homogeneous coordinates expands the three vertices (x, y, z) of each triangular unit into the four-dimensional homogeneous coordinate form of (x, y, z, 1). For example, the vertex coordinates (2, 3, 5) are converted to (2, 3, 5, 1), which facilitates the subsequent unified matrix operations of geometric transformations such as translation, rotation, and scaling. The batch processing of coordinate attributes traverses all triangular units through a script, extracts the three-dimensional coordinates of the vertices, and generates four-dimensional homogeneous coordinates in batches according to the above rules. These coordinates are stored as a unit attribute table, with each unit corresponding to three four-dimensional coordinate values.

[0047] The above step 34, adjacency topology relationship construction and sparse matrix generation, is specifically as follows: Shared edge identification: For each triangular unit's three edges, record the coordinates of the two vertices of the edge (stored in an ordered pair, such as (v1, v2) and v1 < v2), traverse the edges of all units, and if two units have the same ordered edge pair, they are determined to be adjacent units with shared edges.

[0048] A sparse adjacency matrix is ​​constructed with cell indices as rows and columns. If cells i and j share an edge, positions (i, j) and (j, i) in the matrix are marked with 1, and all other positions are marked with 0. Since each cell is adjacent to at most three cells, most elements in the matrix are 0. Therefore, a compressed storage format (such as CSR or CSC format) is used to reduce memory usage. For example, an adjacency matrix with 1000 cells has approximately 3000 non-zero elements, improving storage efficiency by over 90%.

[0049] The above step 35, structured grid data integration, is specifically as follows: Vertex coordinate data is organized by sorting the four-dimensional homogeneous coordinates of all triangular cells by vertex index and storing them as a two-dimensional array. For example, the first column is the vertex index, and the next four columns are the x, y, z, and w coordinate values, making it easy to quickly query the coordinates of a vertex. The adjacency matrix is ​​associated with the cell data, mapping the sparse adjacency matrix to the indices of the triangular cells to form a topological structure table containing the cell adjacency relationships. For example, each cell records its three vertex indices and three adjacent cell indices, forming a bidirectional "cell-vertex-adjacency" mapping. The integrated vertex coordinate data and adjacency matrix data are stored in a standardized format (such as VTK, STL, or a custom binary format).

[0050] Through boundary discrete point encryption and Delaunay subdivision, the present invention achieves a grid-boundary fitting error of ≤2% for the static polygonal monitoring domain, accurately capturing the geometric features of complex areas such as hazardous chemical storage areas and equipment heat dissipation areas, and improving boundary representation accuracy by 30% compared to uniform grids. The subdivision criterion of maximizing the minimum internal angle improves the quality of triangular units (the average minimum internal angle is ≥45°), reduces error diffusion in numerical calculations, and reduces the number of iterations by 25% when solving physical field coupling problems, while ensuring that the calculation error of parameters such as temperature and flow rate is ≤5%. The four-dimensional homogeneous coordinates support unified calculation of transformations such as rotation and scaling, facilitating the subsequent dynamic deformation of grid units in non-uniform deformation fields (for example, the grid in the equipment heat dissipation area can be adaptively stretched with the temperature gradient), and the deformation calculation speed is increased by 40%. The sparse adjacency matrix supports fast query of unit adjacency relationships. When solving dynamic energy efficiency optimization functionals, the physical field gradients (such as temperature gradient and pressure gradient) of adjacent units can be efficiently calculated, shortening the optimization instruction generation time to ≤50ms, meeting real-time control requirements.

[0051] In a preferred embodiment of the present invention, step 4, based on the grid cell characteristics of the area where the key point is located, calculates the deformation parameters of each cell and fuses them to generate a non-uniform deformation field scalar operator, including: Step 41: extracting the four-dimensional homogeneous coordinates of the vertices of each triangular unit based on the structured mesh data; Step 42, based on the four-dimensional homogeneous coordinates of the vertices and in combination with the historical grid data set stored under steady-state working conditions in the laboratory environment, calculate the area tensor of each triangular unit in the structured grid data under steady-state working conditions, and store it as a reference area tensor set corresponding to each unit; Step 43, for the current sampling period, based on the four-dimensional homogeneous coordinates of the vertices obtained in real time from the structured mesh data, calculate the instantaneous area tensor of each triangle unit at the current moment; Step 44: For each triangular element in the structured grid data, a difference tensor is calculated between its instantaneous area tensor in the current sampling period and its corresponding reference area tensor; the Frobenius norm of the difference tensor is calculated, and the Frobenius norm is divided by the Frobenius norm of the element reference area tensor to obtain the corresponding normalized instantaneous deformation change rate; Step 45: The normalized instantaneous deformation change rates of all triangular elements in the structured grid data are used as element deformation parameters to fuse and generate a non-uniform deformation field scalar operator covering the entire static polygon monitoring domain, including: Step 451: constructing a topological weight matrix describing the spatial correlation between grid cells based on the sparse adjacency matrix included in the structured grid data generated in step 34; and calculating a discrete Laplace-Beltrami operator using the topological weight matrix. Step 452: Calculate the covariance matrix of deformation gradients between adjacent cells based on the normalized instantaneous deformation change rate of each cell and the discrete Laplace-Beltrami operator calculated in step 44; Step 453: Perform eigenvalue decomposition on the covariance matrix to extract the principal eigenvector corresponding to the maximum eigenvalue; and based on the principal eigenvector, calculate the eigenvector direction angle on the two-dimensional monitoring plane. Step 454 , converting the eigenvector direction angle into a directional weight factor, and performing weighted projection on the normalized instantaneous deformation change rate using the directional weight factor, mapping it to the local coordinate system defined by the main eigenvector, to obtain the fused unit deformation eigenvalue; Step 455 : Integrate the unit deformation eigenvalues ​​obtained by nonlinear fusion processing of all units to generate a non-uniform deformation field scalar operator with spatial anisotropy characteristics.

[0052] In the embodiment of the present invention, the specific implementation process of generating the above-mentioned non-uniform deformation field scalar operator is as follows: In step 41, the two-dimensional array storing vertex coordinates is located from the structured mesh data generated in step 3, and the four-dimensional homogeneous coordinates (x, y, z, w) of the three vertices corresponding to each triangle unit are extracted in sequence according to the unit index order. For example, for the i-th triangle unit, its three associated vertex indices are found from the array, and then the four-dimensional coordinate data of the corresponding vertex is obtained. These coordinates are organized into a unit coordinate set containing three four-dimensional coordinates for subsequent calculations.

[0053] In the above-mentioned step 42, the historical data of steady-state working conditions is called, and a historical grid data set stored under steady-state working conditions in a laboratory environment is accessed. The data set contains grid unit coordinate information at multiple historical moments. For each triangular unit in the structured grid data, the vertex coordinates of the corresponding unit in the historical grid data are used to calculate the area tensor under steady-state working conditions through a geometric calculation method (such as based on the vector cross multiplication principle). Each triangular unit obtains an area tensor value. The area tensors of all units are aggregated to form a reference area tensor set, and are stored in a one-to-one correspondence with the unit index as reference data for subsequent comparison.

[0054] In step 43, based on the four-dimensional homogeneous coordinates of the triangle unit vertices acquired in real time from the structured mesh data, the instantaneous area tensor of each triangle unit at the current moment is calculated using the same geometric calculation method as step 42 during the current sampling period. For example, a vector is constructed using the vertex coordinates, and the vector cross product is calculated to obtain information related to the triangle area. This instantaneous area tensor is then obtained, and the calculation result for each unit is recorded.

[0055] In the above step 44, for each triangular unit, the instantaneous area tensor of the current sampling period is subtracted from the corresponding element of the reference area tensor stored in step 42 to obtain a difference tensor; the Frobenius norms of the difference tensor and the reference area tensor are calculated respectively. The Frobenius norm calculation process can be understood as taking the square root of the sum of the squares of all elements in the tensor; the Frobenius norm of the difference tensor is divided by the Frobenius norm of the reference area tensor to obtain the normalized instantaneous deformation change rate corresponding to each unit. This value reflects the degree of deformation of the unit at the current moment relative to the steady-state working condition. The normalized instantaneous deformation change rates of all units are compiled into a list as the unit deformation parameter.

[0056] In step 451 above, the topological weight matrix and the discrete Laplace-Beltrami operator are calculated: The topological weight matrix is ​​constructed based on the sparse adjacency matrix generated in step 34. This matrix records the adjacency relationship between grid cells. Based on the adjacency relationship, each cell pair is assigned a weight to construct the topological weight matrix. For example, if two cells share an edge, the weight between them is set to 1; the weight of non-adjacent cells is set to 0. The weight can also be adjusted according to the distance between cells.

[0057] From the sparse adjacency matrix generated in step 34, extract all triangle unit pairs that share edges and record the coordinates of the two endpoints of each edge; for example, if unit i and unit j share edge AB, record the coordinates of vertices A and B of edge AB; split each edge into two directed half-edges (such as AB and BA), and associate each half-edge with its own triangle unit, opposite vertex, and adjacent half-edge. For example, if half-edge AB belongs to unit i, record its opposite vertex as C (the third vertex of unit i), and the adjacent half-edges as BC and CA; collect vertex neighborhoods, and for each vertex, collect all half-edges with the vertex as endpoint and their associated triangle units. For example, the neighborhood of vertex A includes all half-edges with A as the starting point (such as AB, AC, AD, etc.) and their own units; generate an adjacent vertex list, Based on the vertex neighborhood information, a list of adjacent vertices is generated for each vertex. For example, the adjacent vertices of vertex A are B, C, D, etc., which are directly connected to A through shared edges. For cotangent weight calculation, for each pair of adjacent vertices (such as A and B), the sum of the cotangent values ​​of the interior angles of the triangles on both sides of the shared edge AB is calculated. For example, if AB belongs to unit i and unit j, the cotangent values ​​of the interior angles of vertices A and B in unit i and the cotangent values ​​of the corresponding interior angles in unit j are calculated respectively, and the two cotangent values ​​are added as the cotangent weight of edge AB. For each vertex, the sum of the areas of all triangular units in its 1-ring neighborhood is calculated, and the reciprocal is taken as the area normalization factor. For example, the 1-ring neighborhood of vertex A contains triangles ABC, ABD, etc. The sum of the areas of these triangles is calculated and the reciprocal is taken.

[0058] Discrete Laplace matrix construction: Matrix initialization: Create an N×N sparse matrix L (N is the number of vertices) with all initial values ​​set to zero. Diagonal elements are filled in. For each vertex i, the negative value of its area normalization factor is added to the (i, i) position of the matrix. For example, if the area normalization factor of vertex A is k, then L(A, A) = -k. Off-diagonal elements are filled in. For each pair of adjacent vertices i and j, the cotangent weight between them is multiplied by the area normalization factor and then added to the (i, j) and (j, i) positions of the matrix. For example, if the cotangent weight of vertices A and B is w, and the area normalization factor of vertex A is k, then L(A, B) = L(B, A) = w×k.

[0059] Boundary condition processing: Fixed boundary vertex identification: According to the physical characteristics of the monitoring domain, the boundary vertices (such as the edge vertices of the static polygon monitoring domain) are determined. For fixed boundary vertices, the corresponding matrix row elements are set to 0, and the diagonal elements are set to 1 to ensure that the function values ​​of these vertices remain unchanged in subsequent calculations. For example, if vertex A is a fixed boundary vertex, L(A, i) = 0 (i ≠ A) and L(A, A) = 1.

[0060] The above step 452 analyzes the deformation relationship between adjacent units based on the normalized instantaneous deformation change rate of each unit obtained in step 44 and the discrete Laplace-Beltrami operator calculated in step 451, and calculates the covariance matrix of the deformation gradients between adjacent units through statistical calculation methods. This matrix reflects the degree of mutual correlation and change trend between the deformation change rates of different units.

[0061] In step 453, the covariance matrix is ​​subjected to eigenvalue decomposition to obtain a series of eigenvalues ​​and corresponding eigenvectors. The principal eigenvector corresponding to the maximum eigenvalue is extracted. This vector represents the main direction of the deformation change of the grid unit. Based on the coordinate components of the principal eigenvector on the two-dimensional monitoring plane, its direction angle is calculated through trigonometric functions to determine the dominant direction of the deformation change.

[0062] In step 454, the principal eigenvector is projected from three-dimensional space onto a two-dimensional monitoring plane (e.g., the XOY plane) and its coordinate components (x, y) within the plane are extracted. If the angle between the principal eigenvector and the monitoring plane exceeds 45°, the projection direction with the smallest angle with the plane (e.g., the XOZ or YOZ plane) is selected. The angle θ between the projected principal eigenvector and the positive X-axis of the plane is calculated (range: 0° to 360°). This angle is then normalized to the range of 0° to 180° using a modulo operation (e.g., if θ > 180°, 180° is subtracted) to ensure the uniqueness of the direction angle. For example, 300° is normalized to 120°. For each triangular unit, the direction vector of its normalized instantaneous deformation change rate is calculated. This vector can be determined by the deformation gradient directions of the three vertices of the unit (such as the average direction of the vertex deformation gradient), or directly by taking the direction of the main eigenvector at the unit's location (simplified case). The unit deformation direction vector is projected onto the same two-dimensional monitoring plane, and its angle φ with the positive X-axis is calculated. It is also normalized to the range of 0° to 180° to ensure that it is compared in the same dimension as the main eigenvector direction angle.

[0063] Direction angle calculation and weight mapping: Calculate the absolute value of the angle. Calculate the absolute difference between the unit direction angle φ and the main eigenvector direction angle θ as △θ = |φ-θ|. If △θ>90°, take 180°-△θ (map the angle to the range of 0°~90°, so that the vertical direction weight is 0 and the parallel direction weight is the largest). For example, when △θ=120°, convert it to 60°. Cosine weight mapping. Use the cosine function to convert the angle △θ into a direction weight factor w=cos(△θ×π / 180). For example, when △θ = 0°, w = 1 (completely parallel), when △θ = 90°, w = 0 (completely perpendicular), and when △θ = 60°, w = 0.5. In nonlinear enhancement processing, in order to highlight the weight difference of the main direction, the cosine value can be transformed into a power (such as w = cos²(△θ)), so that the weight of the angle less than 45° is increased by 20%~30%. For example, when △θ = 30°, cos²(30°) = 0.75, which can more significantly distinguish the direction difference than the original cosine value of 0.866.

[0064] Spatial distance weighted overlay: The distance from the unit to the origin of the main eigenvector is calculated by taking the starting point of the main eigenvector on the monitoring plane as the origin and calculating the distance d from the center point of each unit to the origin. The coordinates of the unit center point can be obtained by averaging the coordinates of the three vertices of the triangle ((x1+x2+x3) / 3, (y1+y2+y3) / 3); the distance attenuation function is designed, and the distance weight factor u=1 / (1+k×d) is designed, where k is the attenuation coefficient (such as k=0.5 / m, so that u=0.33 when the distance is 2 meters). This factor ensures that the unit close to the origin of the main eigenvector obtains a higher weight, avoiding the far-field unit from affecting the main eigenvector. Directional interference; multiply the directional weight factor w by the distance weight factor u to obtain the final directional weight factor w_total = w × u; calculate the sum S of w_total of all units, and adjust the directional weight factor of each unit to w_normalized = w_total / S to ensure that the sum of all weight factors is 1. For example, if S = 100 and w_total = 5 for a unit, it is 0.05 after normalization. This step ensures that the weight factor maintains energy conservation in the weighted projection and avoids the deformation eigenvalue offset caused by the deviation of the sum of weights.

[0065] In the above step 455, the geometric center position of each triangular unit is first calculated by averaging the coordinate values ​​of the three vertices of the unit. Then, the entire static polygon monitoring domain is divided into multiple small areas according to a regular grid, and the size of each area is set to twice the average unit size. According to the geometric center position of each unit, they are respectively classified into corresponding small areas, so that in subsequent operations, the adjacent units of each unit can be quickly found. Based on the divided areas and the determined unit position relationship, for each target unit, the distance between it and the center point of the surrounding units is calculated; according to the principle that the closer the distance, the higher the weight, the preliminary distance weight is assigned; then, combined with the direction of the main eigenvector determined in step 453, the weights of the adjacent units are adjusted, and the distance between the adjacent units is calculated. Units with consistent main directions are given additional weights, while units in perpendicular directions are given lower weights to highlight the importance of the main direction. Finally, a minimum weight threshold is set to exclude units that are too far away or have too large a direction difference, streamline the calculation range, reduce unnecessary calculations, and obtain the final neighborhood weight matrix. Using the constructed neighborhood weight matrix, for each unit, the eigenvalues ​​of the neighboring units are used to perform a weighted average to obtain a preliminary fusion result. In order to enhance the difference between high eigenvalue areas and low eigenvalue areas, a nonlinear transformation is performed on the preliminary fusion result. Based on the corrected weights, the weighted average is re-performed to further optimize the fusion effect, so that the deformation eigenvalues ​​of each unit can more accurately reflect the deformation and directional characteristics of its surroundings.

[0066] For the results of the fusion processing, the internal fusion values ​​of the cells located at the boundary of the monitoring domain are mixed with the preset boundary values ​​to ensure the physical rationality at the boundary and avoid abnormal boundary data. For the corner cells of the polygon, the influence weights of each direction are carefully adjusted according to the angular relationship between them and the adjacent edges to prevent the eigenvalues ​​at the corners from changing suddenly. Based on the processed cell deformation eigenvalues, the major axis direction of each cell is determined according to the direction of its main eigenvector. Then, a two-dimensional tensor matrix is ​​constructed, which assigns higher weights to the major axis direction and lower weights to the minor axis direction. During the eigenvalue propagation process, this tensor matrix is ​​used to adjust the neighborhood weights to make the eigenvalue propagation in the main direction more efficient and enhance the spatial anisotropy characteristics. For the scalar field under the action of the constructed tensor field, the eigenvalue gradients of each cell and its neighborhood are calculated to find the abnormally high gradient areas. These areas where the gradient exceeds the threshold are smoothed to make the data more consistent with physical laws. Then, the entire scalar field is adjusted as a whole so that its total variation meets the preset conditions. While ensuring data smoothness, key features are retained and over-smoothing is avoided to lose important information.

[0067] For the optimized scalar field, reasonable maximum and minimum values ​​are set, and eigenvalues ​​that are out of range are clipped to ensure the physical rationality of the data. The entire scalar field is then normalized so that its mean and standard deviation meet the preset standards to facilitate subsequent calculations and analysis. The continuous scalar field is sampled according to a regular grid, and the value of each grid point is obtained by interpolating the eigenvalues ​​of the surrounding cells. The position, scalar value, and direction-related parameters of each sampling point are recorded. Finally, all sampling point data and related metadata are integrated together to generate the final non-uniform deformation field scalar operator file.

[0068] The present invention can accurately identify the local spatial deformation caused by temperature changes and equipment operation in key areas of the laboratory (such as hazardous chemical storage areas and high-energy consumption equipment heat dissipation areas) by calculating unit deformation parameters and constructing non-uniform deformation field scalar operators; the non-uniform deformation field scalar operator takes into account the spatial correlation and anisotropy characteristics of the grid cells. When constructing the dynamic energy efficiency optimization functional, it can be used as a physical field coupling weight to make the coupled calculation of multiple physical fields such as temperature field and airflow field more in line with the actual scene; based on the real-time sampling data to update the deformation parameters and scalar operators, the system can quickly respond to changes in the laboratory environment and equipment operation status; for example, when the load of high-energy consumption equipment suddenly increases, resulting in abnormal heat dissipation, the non-uniform deformation The field scalar operator can be updated within one sampling cycle (about 1-2 seconds), providing a timely and accurate basis for the subsequent generation of optimization instructions. Compared with the fixed parameter model, the response speed is improved by more than 50%; through discretization processing and sparse matrix storage, the computational complexity of the non-uniform deformation field scalar operator is effectively reduced while ensuring calculation accuracy; the non-uniform deformation field scalar operator comprehensively considers the geometric characteristics, topological relationships and historical data of the grid cells, and can filter out the influence of abnormal data caused by sensor noise or environmental interference. In the complex and changeable laboratory environment, the stability and reliability of the system are significantly improved, and the misjudgment rate is reduced by about 35%, ensuring the effective implementation of the energy-saving optimization control strategy.

[0069] In a preferred embodiment of the present invention, step 5, using a scalar operator as a physical field coupling weight to construct a dynamic energy efficiency optimization functional, includes: Step 51: Extract the set of laboratory equipment operating parameters and construct an adjustable control variable vector; the operating parameter set includes air conditioning power, ventilation rate, and equipment start and stop status; based on the operating parameter set, associate the environmental physical field parameter vector, which includes temperature gradient, pollutant concentration distribution, and heat flux density; Step 52: Using the non-uniform deformation field scalar operator generated in step 455 as a weight coefficient for multi-physics field interaction; using the scalar operator to weight the coupling relationship between the device energy consumption parameter and the environmental physical field parameter vector, and establishing a coupling equation between the device energy consumption and the physical field parameters; Step 53: defining equipment operation constraints, including hard constraints and soft constraints; assigning Lagrange multipliers to each constraint term based on the coupling equation and the constraints, and constructing an augmented optimization objective function including a coupling equation objective term and a constraint penalty term; Step 54 , constructing a dynamic energy efficiency optimization functional based on the adjustable control variable vector constructed in step 51 , the coupling equation established in step 52 , and the augmented optimization objective function constructed in step 53 .

[0070] In the embodiment of the present invention, the specific implementation process of constructing the dynamic energy efficiency optimization functional is as follows: In step 51, a set of equipment operating parameters is extracted from the laboratory equipment operation monitoring system, covering key indicators such as air conditioning power, ventilation rate, and equipment start and stop status. These parameters are organized into a vector of adjustable control variables. These variables are the direct control targets for subsequent optimization operations. Next, based on real-time monitoring data, a correlation is established between the set of operating parameters and the environmental physical field parameter vector, which includes elements such as temperature gradient, pollutant concentration distribution, and heat flux density. For example, changes in air conditioning power directly affect the laboratory temperature gradient, while ventilation rate is closely related to pollutant concentration distribution. By analyzing historical data and physical laws, the mode and degree of influence of each operating parameter on the physical field parameters are determined, thereby establishing a corresponding relationship between the two. For example, the influence of air conditioning power on temperature gradient: air conditioning power changes the indoor temperature distribution through cooling capacity. The greater the power, the more significant the temperature gradient change.

[0071] Related Relationships: ,in, The coefficient of influence of air conditioning power on temperature gradient (dimensionless), obtained by regression of historical data; It represents the standardized value of the ambient temperature gradient (dimensionless), reflecting the influence of the external environment; Represents the environmental impact coefficient (dimensionless), usually less than 1; represents the random error term (dimensionless), which obeys the normal distribution; represents the temperature gradient (normalized value, dimensionless).

[0072] The influence of ventilation rate on the distribution of pollutant concentration. Ventilation rate reduces indoor pollutant concentration through dilution and replacement. The correlation relationship is: ;in, 、 、 represents the fitting coefficient (dimensionless); represents the standardized value of the pollutant source term (dimensionless), which represents the pollutant generated during the experimental process; represents the random error term (dimensionless), Represents the pollutant concentration distribution (standardized value, dimensionless); the impact of the equipment start and stop status on the heat flux density. When the equipment is turned on, it releases heat, which directly increases the heat flux density. The correlation relationship is: ; in, The coefficient indicating the effect of equipment start-up and shutdown on heat flux density (dimensionless); represents the normalized value of ambient heat flux (dimensionless); represents the environmental impact coefficient (dimensionless); represents the random error term (dimensionless); is the heat flux density (normalized value, dimensionless).

[0073] In step 52, the non-uniform deformation field scalar operator generated in step 455 is introduced into the calculation and used as a weight coefficient for multi-physics field interactions. This scalar operator reflects the deformation differences at different locations and directions within the monitoring domain and can reflect the spatial characteristics of the physical field. The scalar operator is used to weight the coupling relationship between the device energy consumption parameters and the environmental physical field parameter vectors. The strength of the correlation between the device energy consumption parameters and the physical field parameters is adjusted based on the numerical value of the scalar operator at different locations. For example, in areas with large deformation differences and high scalar operator values, the weight of the impact of device energy consumption changes on the physical field parameters is increased; conversely, the weight is reduced. Through this weighting method, a coupling equation between device energy consumption and physical field parameters is established, accurately describing the quantitative relationship between the interaction between the two.

[0074] When applied in practice, the above step 52 is specifically implemented as follows: Extract detailed information for each grid cell from the output file of step 455, including its three-dimensional coordinates in the laboratory (e.g., X = 5.2 m, Y = 3.7 m, Z = 2.1 m) and the corresponding inhomogeneous deformation field scalar operator value (Ω). For example, the Ω value of a grid in the hazardous chemical storage area is 1.3, the Ω value of the grid near high-energy consumption equipment is 1.1, and the Ω value of the grid in the general experimental area is 0.8.

[0075] Laboratory functional area classification: According to the location and Ω value of the grid unit, the monitoring domain is divided into three types of core areas; high-weight areas (Ω≥1.0), including the areas around hazardous chemical storage racks and near the heat dissipation vents of high-energy-consuming equipment. The physical field changes in these areas have a greater impact on safety and energy consumption; medium-weight areas (0.8≤Ω<1.0), such as the areas around hazardous waste treatment stations and fume hood operation areas; low-weight areas (Ω<0.8), laboratory corridors, and non-critical equipment areas; for each grid unit, its connection relationship with the adjacent grids is recorded (such as grid A shares boundaries with grids B, C, and D), and the spatial impact factor is calculated based on the Ω value of the adjacent grids. For example, if Ω of grid A is 1.3 and Ω of adjacent grid B is 1.1, then the impact weight of A on B is 1.3 / 1.1≈1.18 times.

[0076] Coupling weight of air conditioning power to temperature gradient: For high-weight areas (such as hazardous chemicals storage areas), extract the Ω value of all grids in the area (such as average Ω = 1.25) and use the influence coefficient of air conditioning power as the Multiplying the original 0.7 by the Ω value gives the weighted influence coefficient of 0.7×1.25=0.875. This means that in this area, for every 10% increase in air conditioning power, the temperature gradient decreases 25% more than in ordinary areas. The normalized value of the temperature gradient is Air conditioning power The original equation is: .

[0077] Introduce the weighted processing of the scalar operator Ω: For high-weight areas (such as hazardous chemicals storage areas, Ω = 1.3), Multiply by the Ω value to get the weighting coefficient For example, the original , when Ω=1.3, .

[0078] The updated coupled equation expressions are: , where Ω=1.3, the air conditioning power increases by 10% ( When the weighting factor is increased by 0.1, the temperature gradient decreases by 0.1×0.91=0.091, which is 30% higher than that of the unweighted value (0.07).

[0079] Specific operation, if the current air conditioning power is 60% ( =0.6), ambient temperature gradient = 0.2, then when calculating the temperature gradient in this area, use 0.875×0.6 instead of the original 0.7×0.6, and add the environmental impact item 0.4×0.2 ( =0.4), and obtain the final temperature gradient value.

[0080] Low weight area (such as laboratory channel) processing, this area Ω = 0.7, will Multiply by 0.7 to get 0.49; at this time, the air conditioning power has a weaker effect on temperature regulation. For example, =0.6, the temperature gradient control term is only 0.49×0.6=0.294, which is 60% less than that in the high-weight area.

[0081] Coupling weighting of ventilation rate to pollutant concentration: Standardized values ​​of pollutant concentration and ventilation rate The original equation is: ; The weighted processing of the scalar operator Ω is introduced, and the weight modulation of the nonlinear term is used to adjust the denominator. Multiplying by Ω, we get , while the source term Multiply by Ω; for example, in hazardous waste treatment area, Ω=1.1, the original =1.6, then ; Source term After weighting .

[0082] The coupled equation expression is: ; When the ventilation rate When , the unweighted denominator is , after weighting , the calculated value of pollutant concentration decreased by 4.3%, indicating that the ventilation rate needs to be increased to maintain the purification effect; the treatment of violent reaction areas (such as hazardous waste treatment areas, Ω=1.1) has an impact coefficient on the ventilation rate (original 1.6) weighted: 1.6 × 1.1 = 1.76. When the ventilation rate =0.5, the denominator changes from 1+1.6×0.5=1.8 to 1+1.76×0.5=1.88, and the calculated pollutant concentration decreases by about 4.3%, indicating that a higher ventilation rate is required to achieve the same purification effect; at the same time, the pollutant source term Multiplying by Ω = 1.1, if the pollutant source term generated by the original experiment is 0.5, then the weighted value is 0.55, which reflects the characteristic of faster diffusion of pollutants in high deformation areas.

[0083] Treat in a well-ventilated area (e.g. near an exhaust vent, Ω=0.9). The weight is 1.6×0.9=1.44, When =0.5, the denominator=1+1.44×0.5=1.72, which makes it easier to reduce the pollutant concentration than the standard case. At this time, the ventilation efficiency is improved by about 4.5%.

[0084] Coupling weight of equipment start and stop on heat flux density, normalized value of heat flux density Device start / stop status The original equation is: .

[0085] The weighted processing of the scalar operator Ω is introduced, and the device influences the weight modulation: right Multiplying by Ω, we get , for example, high energy consumption equipment area ,Original ,but , when the three devices are turned on at the same time, the total heat flow contribution is .

[0086] The coupled equation expression is: , when the device area When a single device is turned on, its contribution to the heat flux density is 20% higher than that in the normal area (Ω = 1.0), reflecting the enhanced effect of heat dissipation of the equipment in the high deformation area.

[0087] Equipment-intensive areas (e.g., three high-energy-consuming devices running simultaneously, where the grid Ω = 1.2) are handled, and the heat flow influence coefficient of a single device is (original 0.4) weighted to 0.4×1.2=0.48, the total heat flow contribution of the three devices is 3×0.48=1.44; if the ambient heat flow =0.3, =0.2, the total heat flux density is 1.44 + 0.2 × 0.3 = 1.5, which is 19% higher than the unweighted case (3 × 0.4 + 0.06 = 1.26), reflecting the enhanced effect of equipment heat dissipation in high deformation areas. The edge area of ​​the equipment (2 meters away from the equipment, Ω = 0.8) is processed. The weighting is 0.4×0.8=0.32, and the influence of the equipment heat flow is weakened. At this time, the total contribution of the three devices is 0.96, which is 33% less than the core area, reflecting the attenuation of heat flow with spatial distance. The cross-influence of the temperature field and the ventilation field. In areas with high Ω values ​​(such as the hazardous chemical area with Ω=1.3), the temperature gradient changes caused by air conditioning control will further enhance the impact of ventilation on pollutants: Multiply it by 1.1 (temperature-ventilation coupling factor), The final weight is 1.6×1.3×1.1=2.288. =0.6, denominator = 1 + 2.288 × 0.6 ≈ 2.37, the calculated pollutant concentration is 5% lower than that when weighted individually, reflecting the promoting effect of temperature change on airflow.

[0088] The synergistic effect of the heat flow field and the temperature field, when the equipment area Ω = 1.1 and the temperature gradient is high, the influence coefficient of the air conditioning on the temperature In addition to multiplying by Ω = 1.1, a heat flow feedback factor of 0.1 is added, that is, = 0.7 × 1.1 × 1.1 = 0.847. This means that when equipment heat dissipation increases, the air conditioner needs to operate at a higher efficiency to maintain a stable temperature. The scalar operator is refreshed in real time, acquiring new temperature and infrared thermal imaging data from the sensor network every second and recalculating the Ω value for each grid. For example, when a high-energy-consuming device is turned on, the Ω value of the grid near its heat dissipation outlet rapidly increases from 0.9 to 1.1, and the system immediately detects this change.

[0089] Dynamic adjustment process of weight coefficient: If the Ω value of a grid changes by more than 0.1 (e.g., from 1.0 to 1.1), the grid's 、 and Weighted values ​​(such as From 0.7×1.0=0.7 to 0.7×1.1=0.77); the weight coefficient of the adjacent grid (distance ≤ 1 meter) is adjusted by 50% of the Ω change ratio (such as the adjacent grid Increase 0.7 × 0.1 × 0.5 = 0.035), triggering the local reconstruction of the coupling equations in this area, which takes ≤ 20 ms.

[0090] Forced correction of weights under abnormal working conditions: When the temperature in the hazardous chemicals storage area exceeds the limit (e.g. Ω>1.5), the system automatically increases the air conditioning influence coefficient of the area. Forced to increase to 1.5 times the original value (regardless of the current Ω value), while the ventilation influence coefficient Increased to 1.3 times to ensure safety first; for example, the original =0.7, and after forced weighting it becomes 1.05, which improves the air conditioning control efficiency by 50%. Through the above steps, the scalar operator realizes the refined control of the coupling relationship between equipment energy consumption and physical fields: it automatically increases the control weight in high-risk areas and reduces the computational load in ordinary areas, ultimately improving the spatial response accuracy of the coupling equation by 40%, controlling the control delay in key areas within 50ms, and reducing unnecessary calculations by more than 35%.

[0091] Step 53 above classifies and quantifies equipment operation constraints, specifically including hard constraints and soft constraints. Hard constraints (physical safety mandatory restrictions) include equipment power boundary constraints, environmental safety threshold constraints, and equipment start-stop logic constraints. For example, equipment power boundary constraints limit the maximum power of air conditioners to 90% of the rated power (to avoid overload); the minimum speed of ventilation equipment must maintain negative pressure in the laboratory; environmental safety threshold constraints require that the temperature of hazardous chemical storage areas must be kept constant. , heat flux (to avoid chemical spoilage); contaminant concentrations , equipment start-stop logic constraints, high energy consumption equipment start interval (To prevent current shock), the ventilation system must be turned on 5 minutes before the experimental equipment is started , where t represents the time variable, which is used to represent the time point in the ventilation system; represents a function of time t, representing the third control input or state variable (such as voltage or flow that varies with time); Representing variables At the current time The value for the next 5 minutes is used for prediction or delayed control scenarios.

[0092] Soft constraints (adjustable targets for energy efficiency optimization) include energy cost targets, temperature and humidity uniformity targets, and equipment life optimization targets; among them, energy cost targets, the overall energy consumption of the laboratory ,in, is the baseline energy consumption, =10% is the fluctuation tolerance, The overall energy consumption of the laboratory; the temperature and humidity uniformity target, the temperature deviation in key areas , humidity deviation (Ensure the accuracy of precision instruments); equipment life optimization goals, cumulative operating time of high-energy consumption equipment (extend service life), start-stop frequency f ≤3 times / day, of which, Indicates the actual cumulative operating time of high-energy-consuming equipment; Indicates the rated design operating time of the equipment.

[0093] Lagrange multiplier allocation and augmented objective function construction, specifically including: The mathematical expression of constraints includes hard constraints and soft constraints; hard constraints are converted into equality constraints: Power Cap: ,in, Indicates the rated input or output parameters of the device (such as rated voltage, power, etc.); Temperature safety: ,in, Indicates is the input function; represents a function with temperature T as input; T represents the temperature variable; Soft constraints are converted into penalty items: Energy consumption target: ; Temperature and humidity uniformity: ,in, is the temperature weight coefficient (taken as 10), where Indicated by energy is the input function; Indicated by temperature difference is the input function.

[0094] Constructing augmented optimization objective function by combining physical field coupling effects weighted by scalar operators : ; in, 、 is the weight coefficient of coupling between energy consumption and physical field, Indicates the initial time point; Indicates the end time point; represents the power consumption of device i, affecting energy consumption E and temperature T; represents the Lagrange multiplier of the jth constraint (such as life constraint); Indicates the A mathematical expression for an inequality constraint; The mathematical expression representing the kth equality constraint is the same as the inequality constraint ≤0 corresponds to; is a hard-constrained Lagrange multiplier (e.g., λ1 = 1000 to enforce power limits); is the soft constraint penalty factor (e.g. Adjust energy consumption deviation); The non-uniform deformation field scalar operator generated in step 455 is used to weight the physical field influence of key areas (e.g., strengthening the temperature constraint when Ω=1.3 in the hazardous chemicals area); when the hard constraint approaches the threshold (e.g., temperature T≥24°C), the corresponding multiplier is automatically increased (λ2 increases from 500 to 1000) to improve the constraint priority; the soft constraint is adaptively adjusted according to the real-time energy consumption data (e.g., when the energy consumption exceeds 10%, from 50 to 80), balancing energy conservation and safety goals.

[0095] The above step 54 integrates the control variables, coupling equations and constraints to form a dynamic energy efficiency optimization functional formula ,in, ; in, is a static polygon monitoring domain, is the time domain (sampling period 1 second), Indicates the monitoring domain V in static polygon and the time domain [ , ] double integral on ; 、 and are weight functions related to control variables, system state parameters, and constraints, respectively, and are used to quantify the impact of different factors on energy efficiency; Indicates the control variable To optimize the object, seek its optimal trajectory in the time domain; is the controlled variable energy consumption item; Physics terms weighted by scalar operators; is the constraint penalty term.

[0096] The monitoring domain is divided into N grid cells (e.g., 0.5m×0.5m resolution). The control variables of each cell i are associated with the physical field through the structured grid in step 3: Air conditioning power Mapped to the temperature gradient of each unit ; ventilation rate Mapped to pollutant concentrations ; in, and are the influence coefficients of unit i, is the scalar operator value of unit i (such as hazardous chemicals area ),in, represents the aging coefficient of device i (e.g. the aging rate of semiconductor devices is 0.5% per year); Indicates the operating status of device i; Indicates the sensitivity of temperature to aging (e.g., the aging rate doubles for every 10°C increase in temperature); represents the heat exchange power between device i and the environment; represents the lifetime constraint penalty factor of device i; Indicates the effect of load rate on life (e.g. accelerated aging during overload operation); represents the energy input power of device i; Represents the factor that converts actual ventilation rate into effective pollutant removal rate.

[0097] Using the moving average perspective (MPC) strategy, the next 10 minutes are divided into 600 time steps (△t = 1 second), and the functional is updated at each time step: ; in, represents the kth sub-goal (e.g., the energy consumption cost at the kth hour); is a function of time t, representing the actively adjustable system input, as in step 53 , and ; It is a function of time t, used to weight the impact of the physical field on the critical area. Ω(t)>1 means strengthening the constraints in the area (for example, when Ω=1.3 in the hazardous chemicals area, the temperature constraint priority is increased); Ω(t)<1 means weakening the constraints (for example, when Ω<0.8 in the non-critical area, a coarse grid is used for calculation); It is a function of time t, which represents the energy consumption of the system during operation; Represents the constraints, which are a function of time t and include all constraints on system operation (such as temperature, pressure, life, etc.).

[0098] The finite element method is used to discretize the functional and transform it into a nonlinear programming problem (NLP). Based on a structured grid, each triangular element corresponds to a finite element node; the backward Euler method is used to discretize the differential equation, such as the temperature gradient change rate. ,in, represents the power consumption of device 1 at the current time t; Indicates that device 1 was at the last moment Power consumption; Indicates the time step; the solver is IPOPT (Interior PointOPTimizer), and the convergence accuracy is set to 10 −6 , with a maximum number of iterations of 200 times; key areas are prioritized, and the grid unit weights of hazardous chemical storage areas and high-energy consumption equipment areas are increased by 30% to ensure that safety constraints are met first; computational efficiency is optimized, and coarse grid calculations (1m×1m) are used in non-critical areas (Ω<0.8), reducing the number of nodes by 50%, and the solution time is controlled within 50ms.

[0099] This method establishes a coupling relationship between equipment operating parameters and physical field parameters and uses a weighted non-uniform deformation field scalar operator to enable the optimization functional to accurately reflect the actual laboratory operation. Compared with traditional optimization methods, the energy efficiency control accuracy is improved by approximately 30%, which can effectively reduce energy waste. Hard and soft constraints are distinguished and integrated into the augmented optimization objective function via Lagrange multipliers, allowing the optimization process to meet hard requirements such as safety and environmental protection while also taking into account soft objectives such as energy consumption cost and equipment lifespan. The dynamic energy efficiency optimization functional is constructed based on real-time operating parameters and physical field data, and can quickly respond to changes in the laboratory environment and equipment status. Through the control of the optimization functional, resources such as air conditioning power and ventilation rate are rationally allocated to avoid overuse or uneven distribution.

[0100] In a preferred embodiment of the present invention, step 6, combining the real-time changes in environmental parameters to solve the final solution of the dynamic energy efficiency optimization functional and generate an optimization instruction set, includes: Step 61, obtaining the dynamic energy efficiency optimization functional and its function structure constructed in step 54; collecting environmental physical field parameters in real time; calculating the variational increment vector of the environmental physical field parameters collected in real time relative to the previous sampling period; Step 62: Using the variational increment vector as the disturbance input, a gradient descent path is constructed in the parameter space of the dynamic energy efficiency optimization functional. Based on the equipment operation constraints defined in step 53 and their corresponding Lagrange multipliers, the projected gradient method is used to iteratively solve the problem until convergence to a functional minimum point that satisfies all equipment operation constraints. Step 63: When the iterative solution of step 62 converges to a minimum point, the adjustable control variable vector corresponding to the minimum point is analyzed; the analyzed adjustable control variable vector is converted into a physical quantity instruction executable by the laboratory equipment; Step 64, based on the physical quantity instructions converted in step 63, encode the instructions according to the address of the target device, add timestamps and priority tags; based on the execution dependencies between the instructions, build a directed instruction tree containing the instruction execution timing relationship; integrate all instruction information and output an optimized instruction set.

[0101] In an embodiment of the present invention, the above step 61 is based on the dynamic energy efficiency optimization functional (including energy consumption terms, physical field weighting terms, and constraint penalty terms, referred to as the objective function) and constraint conditions constructed in step 54; environmental physical field parameters such as temperature, humidity, and gas concentration, as well as operating parameters such as equipment power and speed are collected in real time through sensors; the environmental parameters of the current sampling period are compared with the data of the previous period, and the difference of each parameter (such as temperature change, concentration change rate) is calculated to form a vector describing the dynamic change of the parameter; for example, if the temperature of the hazardous chemicals storage area rises from 24°C to 24.5°C, the temperature variation increment is +0.5°C; if the ventilation rate drops from 500m³ / h to 480m³ / h, the variation increment is -20m³ / h. 3 / h.

[0102] Step 62 above obtains the variation increment vector of the current environmental parameters from step 61 (e.g., if the temperature rises by 0.5°C and the ventilation volume decreases by 20m 3 / h); using the optimal control variable vector from the previous cycle (e.g., an air conditioning power coefficient of 0.8 and a ventilation rate coefficient of 0.6) as a starting point, set the learning rate (e.g., 0.1, representing the step size of each iteration), the convergence threshold (e.g., the rate of change of the function value <0.5%), and the maximum number of iterations (e.g., 100), and analyze the impact of the variational increment on the various components of the objective function. For example, if rising temperature increases the weight of the cooling energy consumption term, lowering the air conditioning temperature setpoint may more significantly reduce total energy consumption. If reduced ventilation volume leads to an increase in the concentration of hazardous chemicals, increasing the weight of the safety penalty term, then increasing the ventilation rate should be prioritized. Taking into account the sensitivity of each parameter, determine the direction that will most rapidly reduce the objective function. For example, if reducing air conditioning power reduces energy consumption but may cause the temperature to exceed the limit (triggering the penalty term), the gradient direction needs to balance these two factors, potentially pointing to a combination of "slightly reducing power and slightly increasing cooling efficiency."

[0103] Based on the gradient direction, tentative control variable updates are calculated. For example, the original air conditioning power coefficient of 0.8 is changed to a tentative value of 0.75 (a shift of 0.05 along the gradient direction); the original ventilation rate coefficient of 0.6 is changed to a tentative value of 0.65 (increasing ventilation to cope with rising concentrations). All tentative adjustments are combined into a new control variable vector. If the tentative air conditioning power exceeds 90% of the rated value, it is forced to be corrected to 0.9. If reducing ventilation may cause the hazardous chemical concentration to exceed the safety limit, the minimum ventilation rate coefficient is maintained at 0.6. For non-strict constraints (such as energy budgets), if the tentative solution exceeds the budget but the safety risk is controllable, a slight overshoot can be accepted and optimized in subsequent iterations.

[0104] Iterations are continued until the objective function value no longer decreases significantly (e.g., the function value change rate between two adjacent iterations is less than 0.5%), and all constraints (e.g., temperature and power boundaries) are met. At this point, the functional minimum point is reached. The present invention uses projection operations to ensure that the safety threshold (e.g., hazardous chemical temperature exceeds the limit) will not be exceeded during the optimization process. The misjudgment rate is reduced by 35%. Gradient descent combined with constrained projection reduces the number of iterations by 40% compared to blind search, and the single solution time is controlled within 50ms.

[0105] In step 63, the adjustable control variable vector is extracted from the converged minimum point, for example, the optimal air conditioning power is 75% of the rated value, the ventilation rate is 600m 3 / h, a high-energy-consuming device maintains its operating state; converts abstract control variables (such as normalized power coefficient) into physical quantities that can be executed by the device. For example, a control variable of 0.75 corresponds to an actual air conditioner power of 7.5kW (if the rated power is 10kW); a ventilation rate coefficient of 0.6 corresponds to 600m 3 / h (rated 1000m³ / h); the present invention converts mathematical solutions into actual control actions to ensure that the energy efficiency optimization strategy is operational, for example, reducing air conditioning energy consumption by 15% while meeting temperature constraints.

[0106] In step 64 above, physical quantity instructions are encoded by device address (e.g., air conditioner ID-001, ventilator ID-003); timestamps (accurate to the second) are added to mark the execution sequence; priority labels (e.g., high, medium, low) are assigned based on safety priority (e.g., cooling instructions for hazardous chemical areas > ventilation instructions for general areas); dependencies between instructions are analyzed (e.g., the ventilator must be started before the hazardous waste treatment equipment is turned on), and the execution order is represented in the form of a directed graph. For example, instruction A (starting the ventilator) is a predecessor node of instruction B (starting the hazardous waste treatment equipment), ensuring that instruction B is triggered after A completes. All encoded instructions, along with timestamps, priorities, and timing relationships, are aggregated to generate a structured, optimized instruction set. For example, these are packaged in the format of "device ID - instruction content - execution time - priority" and sent to the execution terminal. The directed instruction tree of the present invention avoids multi-device control conflicts. For example, it prevents the leakage of pollutants caused by the operation of hazardous waste treatment equipment when the ventilator is not started. The timestamp and priority mechanism ensure that critical instructions (such as emergency cooling when the temperature of hazardous chemicals exceeds the limit) are executed first, and non-critical instructions (such as lighting adjustment in ordinary areas) are delayed, thereby improving resource utilization by 30%.

[0107] like Figure 2 As shown, an embodiment of the present invention proposes a laboratory energy-saving optimization control system driven by an energy efficiency model, comprising: The acquisition module is used to collect multi-physical field data and equipment operating parameters of the laboratory environment, perform spatiotemporal fusion processing, and generate a fused data set; the processing module is used to identify the core coordinates of the temperature-exceeding area in the hazardous chemical storage area as the first point, identify the coordinates of the peak position of the oxidation reaction rate in the hazardous waste treatment area as the second point, and identify the centroid coordinates of the boundary of the abnormal heat dissipation area of ​​high-energy-consuming equipment as the third point based on the fused data set; a key point coordinate set is constructed based on the first point, the second point, and the third point, and the key points are connected to generate a static polygon monitoring domain; the generation module is used to perform grid discretization processing with the static polygon monitoring domain as the boundary to generate structured grid data containing vertex coordinates and unit adjacency relationships; the calculation module is used to calculate the deformation parameters of each unit based on the grid unit characteristics of the area where the key point is located, and fuse to generate a non-uniform deformation field scalar operator; the construction module is used to use the scalar operator as the physical field coupling weight to construct a dynamic energy efficiency optimization functional; the distribution module is used to solve the final solution of the dynamic energy efficiency optimization functional and generate an optimization instruction set in combination with the real-time changes in environmental parameters; and the optimization instruction set is distributed to the execution terminal.

[0108] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A laboratory energy-saving optimization control method based on energy efficiency model driving, characterized in that: The method comprises: Collect multi-physics field data of laboratory environment and equipment operating parameters, perform spatiotemporal fusion processing, and generate fused data sets; Based on the fused data set, the core coordinates of the temperature-exceeding area in the hazardous chemical storage area are identified as the first point, the coordinates of the peak position of the oxidation reaction rate in the hazardous waste treatment area are identified as the second point, and the coordinates of the centroid of the boundary of the abnormal heat dissipation area of ​​high-energy-consuming equipment are identified as the third point. A key point coordinate set is constructed based on the first, second, and third points, and the key points are connected to generate a static polygon monitoring domain. Using the static polygon monitoring domain as the boundary, the grid is discretized to generate structured grid data containing vertex coordinates and unit adjacency relationships; Based on the grid cell characteristics of the key point area, the deformation parameters of each cell are calculated and fused to generate a non-uniform deformation field scalar operator; Using scalar operators as physical field coupling weights, a dynamic energy efficiency optimization functional is constructed; Combined with the real-time changes in environmental parameters, the final solution of the dynamic energy efficiency optimization functional is solved and an optimization instruction set is generated; Send the optimized instruction set to the execution terminal.

2. The laboratory energy-saving optimization control method based on energy efficiency model driving according to claim 1 is characterized in that: The first point is determined by the following process: Extract the temperature monitoring data of hazardous chemical storage units from the fused dataset, identify the spatial area where the temperature continuously exceeds the preset safety threshold, calculate the minimum bounding rectangle of the spatial area, and use its center coordinates as the first point; The second point is determined by the following process: Analyze the gas concentration time series data of the hazardous waste treatment area in the fused dataset; calculate the gas concentration change rate of adjacent time slices, locate the spatial position with the maximum change rate, and take the weighted centroid coordinates within a preset radius as the second point with this position as the center; The third point is determined by the following process: Acquire infrared thermal imaging data of high-energy consumption equipment, segment the abnormal heat dissipation area based on the temperature gradient threshold, extract the closed boundary of the area, and calculate the coordinates of the centroid of the closed boundary polygon as the third point.

3. The laboratory energy-saving optimization control method based on energy efficiency model drive according to claim 2 is characterized in that: Construct a key point coordinate set based on the first point, the second point, and the third point, and connect the key points to generate a static polygon monitoring domain, including: Taking the key point coordinate set consisting of the first point, the second point and the third point as input, performing homogeneous transformation processing on the coordinates of each key point in three-dimensional space to obtain a transformed key node set in a unified coordinate system; For the transformed key node set, calculate the center of mass O of all nodes projected on the XOY plane, establish a polar coordinate system with O as the pole, calculate the azimuth of each node relative to O, and sort all nodes in increasing order of azimuth to obtain an ordered node sequence; Based on the ordered node sequence, a stack structure is initialized and a starting node with the smallest polar angle is pushed into the stack structure; the remaining nodes in the sequence are traversed in sequence, a vector cross product sign of the two nodes at the top of the stack and the currently traversed node is calculated, and the top element of the stack is iteratively adjusted according to the cross product sign until the ordered node sequence is traversed, and the node sequence retained in the stack is the vertex set constituting the convex hull; The vertex set of the convex hull is spatially sorted according to the right-hand spiral rule, and the connecting edges between adjacent vertices are smoothed to generate the static polygon monitoring domain for mesh discretization.

4. The laboratory energy-saving optimization control method based on energy efficiency model drive according to claim 3 is characterized in that: Using the static polygon monitoring domain as the boundary, grid discretization is performed to generate structured grid data containing vertex coordinates and unit adjacency relationships, including: Taking the static polygon monitoring domain as the input domain, a discrete point set is uniformly inserted along the boundary of the static polygon monitoring domain, and the vertices of the static polygon monitoring domain and the inserted boundary discrete points together constitute the initial discrete node set; Based on the initial discrete node set, the Delaunay triangulation algorithm is used for processing, including: generating a convex hull boundary containing all the initial discrete nodes, iteratively inserting internal diagonals based on the criterion of maximizing the minimum internal angle of the unit, and decomposing the static polygon monitoring domain into a set of non-overlapping triangular units; Assign homogeneous coordinate attributes to each triangle unit in the triangle unit set, that is, expand the three-dimensional space coordinates of the three vertices of each triangle unit into four-dimensional homogeneous coordinate form; Based on the geometric relationship between each unit in the triangle unit set, the triangle unit pairs with shared edges are identified, and a sparse adjacency matrix describing the adjacency topological relationship between all triangle units is constructed accordingly; The four-dimensional homogeneous coordinate data of vertices of all units in the triangular unit set are integrated with the sparse adjacency matrix data to generate structured mesh data.

5. The laboratory energy-saving optimization control method based on energy efficiency model drive according to claim 4 is characterized in that: Based on the grid cell characteristics of the key point area, the deformation parameters of each cell are calculated and fused to generate the non-uniform deformation field scalar operator, including: Based on the structured mesh data, the four-dimensional homogeneous coordinates of the vertices of each triangular unit are extracted; Based on the four-dimensional homogeneous coordinates of the vertices and combined with the historical grid data set stored under steady-state conditions in the laboratory environment, the area tensor of each triangular unit in the structured grid data under steady-state conditions is calculated and stored as the reference area tensor set corresponding to each unit; For the current sampling period, the instantaneous area tensor of each triangle unit at the current moment is calculated based on the four-dimensional homogeneous coordinates of the vertices obtained in real time from the structured mesh data; For each triangular element in the structured grid data, the difference tensor between its instantaneous area tensor in the current sampling period and its corresponding reference area tensor is calculated; the Frobenius norm of the difference tensor is solved and the Frobenius norm is divided by the Frobenius norm of the element reference area tensor to obtain the corresponding normalized instantaneous deformation change rate; The normalized instantaneous deformation change rates of all triangular elements in the structured grid data are used as element deformation parameters and fused to generate a non-uniform deformation field scalar operator covering the entire static polygon monitoring domain.

6. The laboratory energy-saving optimization control method based on energy efficiency model drive according to claim 5 is characterized in that: The normalized instantaneous deformation change rate of all triangular elements in the structured grid data is used as the element deformation parameter, and the non-uniform deformation field scalar operator covering the entire static polygon monitoring domain is generated by fusion, including: Based on the sparse adjacency matrix contained in the structured grid data, a topological weight matrix describing the spatial correlation between grid cells is constructed; using the topological weight matrix, the discrete Laplace-Beltrami operator is calculated; Based on the normalized instantaneous deformation change rate of each unit and the discrete Laplace-Beltrami operator, the covariance matrix of the deformation gradients between adjacent units is calculated; Perform eigenvalue decomposition on the covariance matrix to extract the principal eigenvector corresponding to the maximum eigenvalue; based on the principal eigenvector, calculate the eigenvector direction angle on the two-dimensional monitoring plane; The eigenvector direction angle is converted into a directional weight factor, and the normalized instantaneous deformation change rate is weightedly projected by the directional weight factor, and mapped to the local coordinate system defined by the main eigenvector to obtain the fused unit deformation eigenvalue; The unit deformation eigenvalues ​​obtained by nonlinear fusion processing of all units are integrated to generate an inhomogeneous deformation field scalar operator with spatial anisotropy characteristics.

7. The laboratory energy-saving optimization control method based on energy efficiency model drive according to claim 6 is characterized in that: Using scalar operators as physical field coupling weights, a dynamic energy efficiency optimization functional is constructed, including: Extract the set of laboratory equipment operating parameters and construct a vector of adjustable control variables; The non-uniform deformation field scalar operator is used as the weight coefficient of the multi-physics field interaction; the scalar operator is used to weight the coupling relationship between the device energy consumption parameters and the environmental physical field parameter vectors, and the coupling equation between the device energy consumption and the physical field parameters is established; Defining equipment operation constraints, the constraints including hard constraints and soft constraints; assigning a Lagrange multiplier to each constraint term based on the coupling equation and the constraints, and constructing an augmented optimization objective function including a coupling equation objective term and a constraint penalty term; Based on the adjustable control variable vector, coupling equations and augmented optimization objective function, a dynamic energy efficiency optimization functional is constructed.

8. The laboratory energy-saving optimization control method based on energy efficiency model drive according to claim 7 is characterized in that: Combined with the real-time changes in environmental parameters, the final solution of the dynamic energy efficiency optimization functional is solved and an optimization instruction set is generated, including: Obtain dynamic energy efficiency optimization functionals and collect environmental physical field parameters in real time; calculate the variational increment vector of the environmental physical field parameters collected in real time relative to the previous sampling period; Using the variational increment vector as the perturbation input, a gradient descent path is constructed in the parameter space of the dynamic energy efficiency optimization functional. Based on the equipment operation constraints and their corresponding Lagrange multipliers, the projected gradient method is used to iteratively solve the problem until it converges to the functional minimum point that satisfies all equipment operation constraints. When the iterative solution converges to a minimum point, the adjustable control variable vector corresponding to the minimum point is analyzed; the analyzed adjustable control variable vector is converted into a physical quantity instruction executable by the laboratory equipment; Based on physical quantity instructions, instructions are encoded according to the address of the target device, and timestamps and priority tags are added; based on the execution dependencies between instructions, a directed instruction tree containing the instruction execution timing relationship is constructed; all instruction information is integrated and an optimized instruction set is output.

9. The laboratory energy-saving optimization control method based on energy efficiency model drive according to claim 8 is characterized in that: The operating parameter set includes air conditioning power, ventilation rate, and equipment start and stop status; based on the operating parameter set, the environmental physical field parameter vector is associated, and the environmental physical field parameter vector includes temperature gradient, pollutant concentration distribution, and heat flux density.

10. A laboratory energy-saving optimization control system driven by an energy efficiency model, characterized in that: The system is used to perform the method according to any one of claims 1 to 9, comprising: The acquisition module is used to collect multi-physics field data of the laboratory environment and equipment operating parameters, perform spatiotemporal fusion processing, and generate a fused data set; A processing module is used to identify the core coordinates of the temperature-exceeding area in the hazardous chemical storage area as the first point based on the fused data set, identify the coordinates of the peak position of the oxidation reaction rate in the hazardous waste treatment area as the second point, and identify the centroid coordinates of the boundary of the abnormal heat dissipation area of ​​the high-energy-consuming equipment as the third point; construct a key point coordinate set based on the first point, the second point, and the third point, and connect the key points to generate a static polygon monitoring domain; A generation module is used to perform grid discretization processing based on the static polygon monitoring domain as the boundary, and generate structured grid data containing vertex coordinates and unit adjacency relationships; The calculation module is used to calculate the deformation parameters of each unit based on the grid unit characteristics of the key point area, and fuse them to generate a non-uniform deformation field scalar operator; A building block for constructing dynamic energy efficiency optimization functionals using scalar operators as physical field coupling weights; The sending module is used to combine the real-time changes in environmental parameters to solve the final solution of the dynamic energy efficiency optimization functional and generate an optimization instruction set; and send the optimization instruction set to the execution terminal.

Images