Variable air volume control method, device and equipment based on image recognition

Abstract

The present invention provides a variable air volume control method, device and equipment based on image recognition. The method includes: simulating turbulent flow for a target area based on preset simulation conditions to obtain the temperature distribution of the target area; dividing the target area into grids with a predetermined size based on the temperature distribution of the target area to obtain a plurality of grid areas; determining the regional energy balance equation of each grid area based on the fitting relationship between the interface air volume and the air supply speed, the control temperature of each grid area, and the number of people in each grid area; and controlling the air supply speed with the goal of minimizing the difference between the control temperature and the design temperature of each grid area. It can dynamically control the air supply speed of the air in the grid area, and then control the interface air volume of the air in the grid area, improve the energy utilization efficiency, reduce the operation cost, ensure the comfort requirements of crowded areas, and is conducive to achieving the goal of green and low-carbon.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, specifically to the technology of air supply volume control, and more specifically, to a variable air volume control method, device and equipment based on image recognition. Background Art

[0002] Due to their large volume, high floor height and significant changes in personnel density, large-space buildings have significant energy consumption characteristics. The building heating, ventilation and air conditioning system (i.e., HVAC system) usually accounts for 40% - 60% of the total energy consumption and is the core area of energy consumption control. The phenomena of cold and hot air stratification, local overcooling or overheating are common in large spaces, while traditional air conditioning systems mostly adopt constant air volume control, which is difficult to dynamically respond to the actual needs of different areas, not only increasing energy waste but also reducing the comfort in the space. With the development of smart buildings and the increasing demand for energy conservation, the application of variable air volume control technology in large spaces has gradually become a research hotspot.

[0003] In the process of implementing the inventive concept of the present invention, it is found that for large-space areas with complex air flow distribution, the existing air supply strategies are difficult to achieve precise control, and there are still problems of low energy efficiency in some areas. In addition, the existing control system has a slow response speed and is difficult to quickly adapt to the rapid fluctuations in personnel density. Summary of the Invention

[0004] In view of this, the present invention provides a variable air volume control method, device and equipment based on image recognition.

[0005] One aspect of the present invention provides a variable air volume control method based on image recognition, including: simulating turbulent flow for a target area based on preset simulation conditions to obtain the temperature distribution of the target area; dividing the target area into grid areas of a predetermined size based on the temperature distribution of the target area to obtain a plurality of grid areas; determining the regional energy balance equation of each grid area based on the fitting relationship between the interface air volume and the air supply speed, the control temperature of each grid area, and the number of people in each grid area; and controlling the air supply speed with the optimization goal of minimizing the difference between the control temperature and the design temperature of each grid area, where the number of people in each grid area is determined by image recognition of the video data collected from the grid area.

[0006] According to an embodiment of the present invention, based on the fitting relationship between the interface air volume and the air supply speed, the control temperature of each grid area, and the number of people in each grid area, the regional energy balance equation for each grid area is determined, including: determining the first heat for the air in the grid area based on the temperature difference between the adjacent area of the grid area and the temperature of the grid area, the air parameters of the grid area, and the fitting relationship; determining the second heat for the air in the grid area based on the number of people in the grid area and the heat dissipation per person; determining the third heat for the air in the grid area based on the temperature difference between the air supply temperature and the temperature of the grid area, the air supply port area of the grid area, the air supply speed of the grid area, and the air parameters of the grid area; determining the total heat based on the first heat, the second heat, and the third heat; determining the change in the internal energy of the air in the grid area based on the air parameters of the grid area, the volume of the grid area, and the temperature of the grid area; and obtaining the regional energy balance equation based on the total heat being equal to the change in internal energy.

[0007] According to an embodiment of the present invention, with the goal of minimizing the difference between the control temperature and the design temperature of each grid area, the air supply speed is controlled, including: performing linear transformation based on the non-linear terms in the regional energy balance equation; determining the incremental form of the regional energy balance equation based on the transformed non-linear terms and the regional energy balance equation; transforming the incremental form into a matrix form; discretizing the regional energy balance equation in matrix form by the analytical method to obtain a state space equation with the increment of the control temperature of each grid area as the output; determining the state space equation after iterating a preset number of steps based on the state space equation and the measurable disturbance condition; transforming the temperature increment and the air supply speed increment within the preset number of iterations into their respective corresponding matrix forms based on the state space equation after iterating the preset number of steps; determining the variable air volume prediction equation based on the matrix forms of the temperature increment and the air supply speed increment within the preset number of iterations; determining the objective function according to the optimization goal; obtaining a quadratic programming problem with the air supply speed as the variable based on the variable air volume prediction equation and the objective function; and solving the quadratic programming problem using the interior point penalty function method to control the air supply speed.

[0008] According to an embodiment of the present invention, using the interior point penalty function method to solve the quadratic programming problem to control the air supply speed, including: transforming the quadratic programming problem into a logarithmic interior point penalty function; updating the variable using the adaptive learning rate algorithm and gradually reducing the barrier factor of the logarithmic interior point penalty function, repeating the update of the variable until the termination criterion is met to obtain the updated variable; and completing the control of the interface speed when it is determined that the preset number of iterations meets the iteration threshold.

[0009] According to an embodiment of the present invention, the variable air volume control method based on image recognition further includes: when it is determined that the respective true temperatures in the collected grid regions are successfully obtained, determining the respective true temperatures in the grid regions as the control temperatures.

[0010] According to an embodiment of the present invention, the variable air volume control method based on image recognition further includes: when it is determined that the acquisition of the respective true temperatures in the collected grid regions fails and the acquisition of the respective true temperatures in some of the grid regions in the collected grid regions is successful, based on a temperature state observer, constructing an estimation model for estimating the temperature change in the target region; using the estimation model and the respective true temperatures of some of the grid regions to predict the respective true temperatures of the other grid regions in the grid regions except for some of the grid regions, obtaining the respective predicted temperatures; determining the respective true temperatures of some of the grid regions as the respective control temperatures of some of the grid regions; and determining the respective predicted temperatures of the other grid regions as the respective control temperatures of the other grid regions.

[0011] According to an embodiment of the present invention, based on the temperature distribution of the target region, dividing the target region into a plurality of grid regions with a predetermined size includes: determining the temperature change situation in the target region according to the temperature distribution of the target region; when it is determined that the temperature change situation in the target region meets the designed temperature change condition, dividing the target region into a plurality of grid regions with a first predetermined size; when it is determined that the temperature change situation in the target region does not meet the designed temperature change condition, dividing the target region into a plurality of grid regions with a second predetermined size, wherein the first predetermined size is smaller than the second predetermined size.

[0012] According to an embodiment of the present invention, the variable air volume control method based on image recognition further includes: for each grid region: detecting the human contour in the collected video data and marking the candidate boxes; and determining the number of people according to the number of candidate boxes.

[0013] Another aspect of the present invention provides a variable air volume control device based on image recognition, including: a simulation module for simulating turbulent flow for the target region based on preset simulation conditions to obtain the temperature distribution of the target region; a division module for dividing the target region into a plurality of grid regions with a predetermined size based on the temperature distribution of the target region; a determination module for determining the regional energy balance equation of each grid region based on the fitting relationship between the interface air volume and the air supply speed, the control temperature of each grid region, and the number of people in each grid region; and a control module for controlling the air supply speed with the optimization goal of minimizing the difference between the control temperature and the designed temperature of each grid region, wherein the number of people in each grid region is determined by image recognition of the video data of the collected grid regions.

[0014] Another aspect of the present invention provides an electronic device, including: one or more processors; a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the above method.

[0015] According to an embodiment of the present invention, by simulating turbulent flow, it is possible to determine the temperature distribution in a spatial region with complex air flow distribution. Combined with mesh generation, refined management can be achieved. Furthermore, by analyzing the regional energy balance equations of each mesh region constructed with the number of people, air supply speed, and temperature, the difference between the controlled temperature and the designed temperature is gradually minimized, and the air supply speed of the air in the mesh region is dynamically optimized, thereby controlling the interface air volume of the air in the mesh region, improving energy utilization efficiency, reducing operating costs, ensuring the comfort requirements of crowded areas, and contributing to the goal of green and low-carbon in the construction industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Through the following description of the embodiments of the present invention with reference to the drawings, the above and other objects, features, and advantages of the present invention will become clearer. In the drawings:

[0017] Figure 1 The flowchart of the variable air volume control method based on image recognition according to an embodiment of the present invention is shown;

[0018] Figure 2 The schematic diagram of mesh generation based on temperature distribution according to an embodiment of the present invention is shown;

[0019] Figure 3 The schematic diagram of detecting the human contour based on the deep learning object detection algorithm model according to an embodiment of the present invention is shown;

[0020] Figure 4 The schematic diagram of eigenvalue selection according to an embodiment of the present invention is shown;

[0021] Figure 5 The schematic diagram of the Luenberger observer control structure according to an embodiment of the present invention is shown;

[0022] Figure 6 The flowchart for variable air volume control according to an embodiment of the present invention is shown;

[0023] Figure 7 The block diagram of the variable air volume control device based on image recognition according to an embodiment of the present invention is shown. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0024] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present invention. In the following detailed description, for the sake of explanation, numerous specific details are set forth in order to provide a comprehensive understanding of the embodiments of the present invention. However, obviously, one or more embodiments can also be implemented without these specific details. In addition, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessarily obscuring the concepts of the present invention.

[0025] The terms used herein are merely for the purpose of describing specific embodiments and are not intended to limit the present invention. The terms "comprising", "including" and the like used herein indicate the presence of features, steps, operations and / or components, but do not preclude the presence or addition of one or more other features, steps, operations or components.

[0026] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification, and should not be interpreted in an idealized or overly rigid manner.

[0027] In cases where expressions similar to "at least one of A, B, and C, etc." are used, generally, it should be interpreted according to the meaning commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include, but not be limited to, a system having only A, only B, only C, having A and B, having A and C, having B and C, and / or having A, B, and C, etc.).

[0028] In the process of implementing the inventive concept, it has been found that a variable air volume control system can adjust the air supply volume according to real-time parameters such as temperature, humidity, and carbon dioxide concentration to meet the actual needs and improve the energy efficiency of the air in the grid area. Although it has advantages in reducing energy consumption, for large-space buildings such as subway stations, airport terminals, stadiums, and shopping malls, due to the complex air flow distribution in large spaces, it is difficult to achieve precise control with existing air supply strategies, and there are still problems of low energy efficiency in some areas. In addition, the existing control system has a slow response speed and is difficult to quickly adapt to the rapid fluctuations in the personnel density. To address these problems, further research on intelligent control and optimization should be carried out. For example, the present invention introduces an artificial intelligence algorithm to improve the response ability and prediction ability of the controller.

[0029] Based on this, embodiments of the present invention provide a variable air volume control method, apparatus, and device based on image recognition. The method includes: simulating turbulent flow for a target area based on preset simulation conditions to obtain the temperature distribution of the target area; dividing the target area into grids of a predetermined size based on the temperature distribution of the target area to obtain a plurality of grid areas; determining the regional energy balance equation for each grid area based on the fitting relationship between the interface air volume and the supply air velocity, the control temperature of each grid area, and the number of people in each grid area; and controlling the supply air velocity with the optimization goal of minimizing the difference between the control temperature and the design temperature of each grid area, where the number of people in each grid area is determined by performing image recognition on the video data of the collected grid area.

[0030] The following will be through Figures 1 - 6 to describe in detail the variable air volume control method based on image recognition according to the embodiments of the present invention.

[0031] Figure 1 The flowchart of the variable air volume control method based on image recognition according to the embodiments of the present invention is shown.

[0032] As Figure 1 shown, the variable air volume control method based on image recognition includes operations S110 to S140.

[0033] In operation S110, based on preset simulation conditions, simulate turbulent flow for the target area to obtain the temperature distribution of the target area.

[0034] In operation S120, divide the target area into grids of a predetermined size based on the temperature distribution of the target area to obtain a plurality of grid areas.

[0035] In operation S130, determine the regional energy balance equation for each grid area based on the fitting relationship between the interface air volume and the supply air velocity, the control temperature of each grid area, and the number of people in each grid area.

[0036] In operation S140, control the supply air velocity with the optimization goal of minimizing the difference between the control temperature and the design temperature of each grid area.

[0037] According to the embodiments of the present invention, the target area may include a plurality of supply air outlets and / or a plurality of return air outlets. The supply air outlets are used to transmit air from outside the target area into the target area, and the return air outlets are used to transmit air from inside the target area to outside the target area. The supply air velocity when the supply air outlets and the return air outlets transmit air may be the same. The target area may include, for example, but is not limited to, large space building areas such as subway stations, exhibition halls, airports, etc.

[0038] According to an embodiment of the present invention, a geometric model can be established, the target area can be divided into tetrahedral meshes, and based on preset simulation conditions, the k-ε turbulence model of computational fluid dynamics (CFD) software can be used to simulate turbulent flow to obtain the temperature distribution of the target area.

[0039] According to an embodiment of the present invention, the preset simulation conditions can be determined according to the simulation software. For example, for computational fluid dynamics (CFD) software, the preset simulation conditions can include but are not limited to: boundary conditions, the air supply openings are set as velocity inlets, the air supply velocities and air supply temperatures of each air supply opening are the same, the air supply velocities and air supply temperatures of each air supply opening are the steady-state calculation values under the design conditions, the return air opening is set as a pressure outlet, the boundaries except the air openings are all set as non-slip wall surfaces, the initial indoor temperature distribution, the outdoor meteorological parameters on a typical meteorological day, setting the solver and defining the convergence criterion, etc. The outdoor meteorological parameters such as outdoor air temperature, relative humidity, wind speed, solar radiation, etc. The initial indoor temperature distribution can be set according to the actual situation.

[0040] According to an embodiment of the present invention, the predetermined size can be determined according to the regional temperature change situation. The regional temperature change situation is inversely proportional to the predetermined size. For example, a small size is used for the area with a large temperature change trend, and a large size is used for the area with a small temperature change trend.

[0041] According to an embodiment of the present invention, the convective heat transfer between the wall and the air can be ignored and regarded as an unmeasurable disturbance in variable air volume control. The fitting relationship between the interface air volume and the air supply velocity, the control temperature of each grid area, and the number of people in each grid area are regarded as measurable disturbances in variable air volume control. Based on the law of conservation of energy, the regional energy balance equation of each grid area is constructed.

[0042] According to an embodiment of the present invention, the design temperature can be a design value, which can be determined by field specifications or standards. The control temperature can include the temperature actually collected by sensors.

[0043] According to an embodiment of the present invention, due to simulating turbulent flow, the temperature distribution in the spatial area with complex air flow distribution can be determined. Combined with mesh division, refined management can be realized. Furthermore, by analyzing the regional energy balance equation of each grid area constructed with the number of people, air supply velocity, and temperature, the difference between the control temperature and the design temperature is gradually minimized, the air supply velocity of the air in the control grid area is dynamically optimized, and then the interface air volume of the air in the grid area is controlled, so as to improve the energy utilization efficiency, reduce the operation cost, ensure the comfort requirements in crowded areas, and contribute to the goal of green and low-carbon in the construction industry.

[0044] According to an embodiment of the present invention, the temperature distribution is determined through CFD simulation, the air supply strategy is optimized, and refined management is carried out in combination with the requirements of the grid area, which is beneficial to effectively cope with the complex loads in large spaces. The variable air volume control method provided by the present invention is conducive to the gradual development of variable air volume control technology towards the direction of intelligence and low carbonization.

[0045] According to an embodiment of the present invention, for the operation S120 above Figure 1 in the operation S120, based on the temperature distribution of the target area, the target area is divided into grid areas of a predetermined size to obtain a plurality of grid areas, which may include operations: determining the temperature change situation in the target area according to the temperature distribution of the target area; when it is determined that the temperature change situation in the target area meets the preset temperature change condition, dividing the target area into a plurality of grid areas of a first predetermined size; when it is determined that the temperature change situation in the target area does not meet the preset temperature change condition, dividing the target area into a plurality of grid areas of a second predetermined size. The first predetermined size is smaller than the second predetermined size.

[0046] According to an embodiment of the present invention, the preset temperature change condition can be determined according to the actual situation and is used to indicate the situation where the temperature change trend reaches the preset trend value.

[0047] Figure 2 The schematic diagram of grid division based on temperature distribution according to an embodiment of the present invention is shown.

[0048] As Figure 2 shown, taking a certain target area as an example, the blank part represents the wall columns in the target area, and the other part represents the open area in the target area. For the open area, the temperature gradually increases from green to yellow and then to red. The black square can represent the divided grid area. The size of the grid area divided in the area with a large temperature change is smaller than the size of the grid area divided in the area with a gentle temperature change. The blue dot represents the air supply opening set in the target area. It should be noted that the air supply speed of each air supply opening is the same, and the air supply opening is set according to the area of the actual air supply area.

[0049] According to an embodiment of the present invention, small grids are adopted in areas with drastic temperature changes, which can capture temperature gradients more precisely and reduce calculation errors. Large grids are adopted in areas with gentle temperature changes to reduce the amount of calculation and improve the simulation efficiency. Since eddy currents are usually accompanied by strong local coupling of momentum and heat, and have a significant impact on heat exchange and momentum exchange, for example, near heat sources, cold sources, or flow separation regions. Therefore, by refining the grid based on the temperature distribution according to the present invention, sufficient resolution can be provided to more precisely capture local phenomena caused by eddy currents near the interface, avoid the eddy currents being smoothed out by numerical diffusion, and make the shape of the eddy currents closer to the real situation, thereby improving the physical rationality of the overall flow field. Through a reasonable grid division strategy, the credibility of numerical simulation can be greatly improved, the applicability and numerical convergence of the model can be enhanced, and at the same time, an optimal balance point can be found between calculation efficiency and accuracy. It can at least partially solve the problem that due to the existence of eddy currents on the interfaces of some non-grid boundaries, heat exchange is ignored because the net air volume is very small.

[0050] According to another embodiment of the present invention, the variable air volume control method based on image recognition may further include operations in addition to operations S110 to S140 as shown in Figure 1 For each grid region: perform human contour detection on the collected video data and mark candidate boxes; and determine the number of people according to the number of candidate boxes.

[0051] According to an embodiment of the present invention, a real-time vision application system can be developed by combining a programming language with computer vision and machine learning software libraries (OpenCV), and a deep learning object detection algorithm (FASTER R-CNN) is used to perform dynamic pedestrian flow detection on the video data of the real-time captured images of the camera in the target area. The real-time vision application system accesses the video stream of the real-time captured images of the camera through an interface, and uses a cross-platform software development kit (such as SDK) built based on an open-source multimedia framework (such as LibVLC) to obtain the video stream through a predetermined format transmission protocol (RTSP). Real-time frames are captured from the video stream through the class function cv2.VideoCapture of OpenCV and converted into video data in a format such as Mat.

[0052] The improved FASTER R-CNN algorithm model can be used to perform human contour detection on the collected video data and frame candidate boxes. The number of people is determined according to the number of candidate boxes.

[0053] The improved Faster R-CNN algorithm model may include a convolutional layer, a Region Proposal Network (RPN) layer, a Region of Interest (ROI) pooling layer, a classification layer, and a bounding box regression layer. The convolutional layer may extract features from video data to obtain a feature map. The RPN layer may mark regions with head features based on the feature map to obtain candidate boxes. The ROI pooling layer may scale the candidate boxes and the feature map obtained from the RPN layer to a fixed size to obtain a fixed-size feature map containing the candidate boxes. The classification layer may determine whether a candidate box contains a person by detecting the classification probability. The bounding box regression layer may adjust the position and size of the candidate box through detecting bounding box regression to make it more accurately surround the person, thereby achieving more accurate person recognition and person counting.

[0054] Figure 3 FIG. shows a schematic diagram of detecting a person's contour based on a deep learning object detection algorithm according to an embodiment of the present invention.

[0055] As Figure 3 shown, video data can be input into the convolutional layer to output a feature map, the feature map is input into the RPN layer to obtain candidate boxes, the candidate boxes and the feature map are input into the ROI pooling layer to output a fixed-size feature map containing the candidate boxes, the fixed-size feature map containing the candidate boxes is input into the classification layer to output the detection classification probability for whether the candidate box contains a person, and based on this detection classification probability, it is determined whether the candidate box contains a person. In addition, before inputting the fixed-size feature map containing the candidate boxes into the classification layer, the fixed-size feature map containing the candidate boxes can also be input into the bounding box regression layer to output an adjusted feature map containing the candidate boxes, and then input into the classification layer.

[0056] In determining the number of people, for example, after selecting a specific area through the camera screen, the video data within 30 frames is processed in a cycle of 30 frames. In each frame, the improved Faster R-CNN detects people in the image and marks candidate boxes, and at the same time, based on the correlation filtering algorithm (KCF), continuously tracks the flow of people in each grid area within each cycle. In a cycle of every 30 frames in the video stream, the system counts the number of people detected in each frame and calculates the average value of the detected numbers. Since the average value can smooth out short-term fluctuations caused by detection errors, target overlaps, or inaccurate tracking, it can provide a relatively stable statistical result.

[0057] According to an embodiment of the present invention, it is possible to continuously collect images of the flow of people in a grid area within a preset period through a camera, perform detection and recognition of the images of the flow of people based on the improved Faster R-CNN algorithm, achieve real-time and accurate online statistics of the number of people in the grid area, and be able to control the air supply speed as a measurable disturbance, thereby controlling the air supply volume.

[0058] According to an embodiment of the present invention, for the above-mentioned Figure 1 In operation S130, based on the fitting relationship between the interface air volume and the air supply speed, the controlled temperature of each grid area, and the number of people in each grid area, determining the regional energy balance equation for each grid area may include operations: determining the first heat for the air in the grid area based on the difference between the temperature of the area adjacent to the grid area and the temperature of the grid area, the air parameters of the grid area, and the fitting relationship; determining the second heat for the air in the grid area based on the number of people in the grid area and the heat dissipation per person; determining the third heat for the air in the grid area based on the difference between the air supply temperature and the temperature of the grid area, the air supply opening area of the grid area, the air supply speed of the grid area, and the air parameters of the grid area; determining the total heat based on the first heat, the second heat, and the third heat; determining the change in the internal energy of the air in the grid area based on the air parameters of the grid area, the volume of the grid area, and the temperature of the grid area; and obtaining the regional energy balance equation based on the total heat being equal to the change in internal energy.

[0059] According to an embodiment of the present invention, the air parameters of the grid area may include the air density of the grid area and the specific heat capacity of the air in the grid area. The volume of the grid area may be determined by the product of the net height of the grid area and the bottom area of the grid area.

[0060] According to an embodiment of the present invention, the interface air volume and the air supply speed The fitting relationship may be , and are respectively the first-order term coefficient and the constant term of the fitting relationship formula. The convective heat transfer between the wall and the air can be ignored and regarded as an unmeasurable disturbance in the variable air volume control. Each grid area in the target area can be mapped into a two-dimensional area from the same direction. The x-axis and y-axis are used to represent the center of each grid area, and the coordinate points of the center of each grid area are used to identify the grid area. For example, for any grid area ( ), i represents the x-axis coordinate of the grid area center in the two-dimensional area, j represents the y-axis coordinate of the grid area center in the two-dimensional area, and the regional energy balance equation for each grid area may be as shown in the following formula (1):

[0061] (1)

[0062] Wherein, is the air density of the grid area, with the unit of kg / m 3 ; is the specific heat capacity of the air in the grid area, with the unit of J / (kg·K); is the net height of the grid area ( ), with the unit of m; is the bottom area of the grid region ( ), in m 2 ; is the control temperature of the grid region ( ) at the calculation time , in °C; the unit of is s; , are respectively the first-order coefficient and the constant term of the fitting relationship of the interface air volume and the air supply velocity on the interface m of the grid region ( ); each grid as shown in the above Figure 2 contains four interfaces of up, down, left and right, and m is the interfaces of up, down, left and right; is the air supply velocity of the grid region ( ) at the calculation time , in m / s; is the control temperature of the adjacent region of the interface m of the grid region ( ) at the calculation time , in °C; is the heat dissipation per person, in W / person; is the number of people in the grid region ( ) at the calculation time ; is the air supply opening area of the grid region, in m 2 ; is the air supply temperature of the grid region, in °C. The term on the left side of the equation in formula (1) can represent the change in the internal energy of the air in the grid region. The right side of the equation in formula (1) is the sum of three terms, that is, the total heat, which can represent the first heat, the second heat, and the third heat for the air in the grid region from left to right in sequence. The heat dissipation per person can represent the heat dissipated by the human body to the surrounding environment through different methods, and in the present invention, the heat dissipation per person can be determined according to empirical values.

[0063] According to the embodiments of the present invention, for the above Figure 1In operation S140, with the goal of minimizing the difference between the controlled temperature and the designed temperature of each grid region, the supply air velocity is controlled, which may include the operations of: linearly transforming based on the non-linear terms in the regional energy balance equation; determining the incremental form of the regional energy balance equation based on the transformed non-linear terms and the regional energy balance equation; transforming the incremental form into a matrix form; discretizing the regional energy balance equation in matrix form by the analytical method to obtain a state space equation with the incremental controlled temperature of each grid region as the output; determining the state space equation after a preset number of iterative steps based on the state space equation and the measurable disturbance condition; transforming the temperature increment and the supply air velocity increment within the preset number of iterative steps into their respective corresponding matrix forms based on the state space equation after the preset number of iterative steps; determining a variable air volume prediction equation based on the matrix forms of the temperature increment and the supply air velocity increment within the preset number of iterative steps; determining an objective function according to the optimization goal; obtaining a quadratic programming problem with the supply air velocity as the variable based on the variable air volume prediction equation and the objective function; and controlling the supply air velocity by solving the quadratic programming problem using the interior point penalty function method.

[0064] According to an embodiment of the present invention, the non-linear terms in the regional energy balance equation may include the product of the temperature and the supply air velocity of the grid region. For example, the non-linear terms in formula (1) , can be linearized by using the first-order Taylor expansion. For the linear terms 、 、 are directly written in the incremental form, and the specific form is shown in the following formula (2):

[0065] (2)

[0066] Wherein, 、 、 respectively correspond to the initial values of the three variables of the controlled temperature, the supply air velocity, and the number of people in the grid region, and can take the steady-state calculation values under the designed working conditions; 、 、 respectively correspond to the increments of the three variables of the controlled temperature, the supply air velocity, and the number of people in the grid region at the calculation time .

[0067] Formula (2) can be substituted into formula (1), and then subtracted from the initial form of formula (1) to obtain the incremental form of the regional energy balance equation as shown in the following formula (3):

[0068] (3)

[0069] Wherein, is the grid region ( ) at the calculation time The temperature increment at is the temperature increment of the adjacent region of the interface m of the grid region ( ) at the calculation time ; is the increment of the number of people in the grid region ( ) at the calculation time ; is the initial value of the controlled temperature of the adjacent region of the interface m of the grid region ( ); is the increment of the air supply velocity in the grid region ( ) at the calculation time . The above formula (3) can be arranged in matrix form as shown in the following formula (4):

[0070] (4)

[0071] wherein, represents the temperature increment matrix at the calculation time , with a dimension of n×1; n is the number of grid regions, represents the coefficient matrix of the temperature increments of n grid regions, with a matrix dimension of n×n; represents the coefficient matrix of the input parameter air supply velocity, with a matrix dimension of n×1; represents the increment of the air supply velocity at the calculation time , which is a numerical value; represents the coefficient matrix of the change in the number of people in each grid region of the measurable disturbance. Since each grid region only contains the measurable disturbance of its own region, therefore the matrix is a diagonal matrix; represents the matrix of the increment of the number of people at the calculation time , with a dimension of n×1.

[0072] The time step at the calculation time can be , , from to . The above matrix form can be discretized by the analytical method to obtain the following formula (5):

[0073] (5)

[0074] Assume that the air supply velocity and the measurable disturbance are step functions. From to , the air supply velocity and the measurable disturbance remain unchanged at and respectively. Then, the above formula (5) can be transformed into the following formula (6):

[0075] (6)

[0076] Integrate the inner function of the above formula (6) First, translate it to the left by , and then perform a symmetry transformation. Then, the above formula (6) is transformed into a state - space equation with the increment of the control temperature of each grid region as the output, as shown in the following formula (7):

[0077] (7)

[0078] Wherein, represents the temperature increment matrix at the k - th time step, with a dimension of n×1; represents the coefficient matrix of the state variables (temperature increments) of n grid regions, which describes the evolution process of the system from the k - th time step to the k + 1 - th time step without input. The dimension of the matrix is n×n; represents the coefficient matrix of the input parameter supply air velocity, which represents the influence degree of the input on the change of the system state. The dimension of the matrix is n×1; represents the increment of the supply air velocity at the k - th time step, which is a numerical value; represents the coefficient matrix of the change in the number of people in each grid region of the measurable disturbance, which represents the influence degree of the measurable disturbance on the change of the system state. The dimension of the matrix is n×n; represents the increment matrix of the number of people at the k - th time step, with a dimension of n×1; represents the output variable, which is the temperature increment at the k - th time step. In the present invention, the dimension is N×1, where N is the number of grid regions with measurable temperature; represents the output matrix, which describes the contribution of the state variable to the output, that is, how the state is mapped to the output of the system , and in the present invention, the dimension is N×n; represents the influence matrix of the input variable on the output variable, which is zero in the present invention; represents the influence matrix of the measurable disturbance on the output variable, which is zero in the present invention. The temperature at the k + 1 - th time step is the temperature at the k - th time step plus the temperature increment at the k + 1 - th time step, that is = + . According to another embodiment of the present invention, in addition to the operations S110~S140 as shown above Figure 1 , the variable air volume control method based on image recognition may further include an operation: when it is determined that the real temperature in each grid region is successfully collected, determine the real temperature in each grid region as the control temperature.

[0079] In the process of implementing the embodiments of the present invention in real time, it is found that when the space of the target area is large, the cost, quantity, and accuracy of installing sensors will all be limited. If temperature measurement points can only be arranged in a limited area, the obtained true temperature measurement values can only reflect the temperature information of part of the area, and cannot directly provide the complete state of the entire space. In this case, if directly relying on the measurement data for control, it will cause the system response to lag or be inaccurate.

[0080] Based on this, in another embodiment of the present invention, in addition to the operations S110~S140 shown above, the variable air volume control method based on image recognition may further include the operations: when it is determined that the acquisition of the true temperature of each grid area in the acquisition grid area fails and the acquisition of the true temperature of some grid areas in the acquisition grid area is successful, an estimation model for estimating the temperature change in the target area is constructed based on a temperature state observer; and the true temperature of each grid area in the grid area except the some grid areas is predicted by using the estimation model and the true temperature of each of the some grid areas to obtain the respective predicted temperatures; the true temperature of each of the some grid areas is determined as the control temperature of each of the some grid areas; and the predicted temperature of each of the other grid areas is determined as the control temperature of each of the other grid areas. Figure 1 According to the embodiments of the present invention, in order to ensure that the state estimator can accurately estimate the temperature of all areas through the temperature of the measurable area, the matrix pair

[0081] needs to be observable. The judgment basis for observability is that the matrix is row full rank, and the number of rows of the output matrix should be as small as possible to estimate the temperature of all areas with the least number of measurement points installed. The output matrix is a screening matrix with dimensions of N×n, where N is the number of measurable temperature areas and n is the number of grid areas. The matrix is as shown in the following formula (8): As shown in the following formula (8):

[0082] (8)

[0083] A temperature state observer, such as a Luenberger observer, can be used to establish an estimation model for estimating the temperature change in the target area. The discretized standard form is as shown in the following formula (9):

[0084] (9)

[0085] Among them, represents the estimated temperature increment matrix at the (k + 1)-th time step; represents the estimated temperature increment matrix at the k-th time step; represents the gain matrix; Denotes the output calculated from the estimated state variables at the k-th time step.

[0086] The estimated temperature at the (k + 1)-th time step is the output calculated from the estimated state variables at the (k + 1)-th time step plus the estimated temperature at the k-th time step, i.e., = + .

[0087] The second formula in Equation (9) can be substituted into the first formula in Equation (9) to obtain the following as shown in Equation (10):

[0088] (10)

[0089] The first formula in Equation (7) can be subtracted from Equation (10), and the second formula in Equation (7) can be substituted to obtain the following as shown in Equation (11):

[0090] (11)

[0091] Where, is the error of the estimation model at the k-th time step , and the following as shown in Equation (12) can be obtained:

[0092] (12)

[0093] is the error of the estimation model at the (k + 1)-th time step.

[0094] Figure 4 Shows a schematic diagram of eigenvalue selection according to an embodiment of the present invention.

[0095] Since designing an estimation model can ensure the stability of variable air volume control, for a discrete system, it is required that all eigenvalues of the matrix must satisfy that the modulus is less than 1 (inside the unit circle in the complex plane) to ensure the asymptotic stability of the system. The dynamics of the corresponding estimation model are determined by the error. Therefore, the eigenvalues of the matrix must also satisfy that the modulus is less than 1 and have no common eigenvalues with the matrix .

[0096] Arrange all eigenvalues of the matrix in a sector region with a radius less than 1 as shown in Figure 4 . This region is bounded by two straight lines radiating from the origin with an angle of . The larger the angle, the larger the overshoot. If all matrices If the eigenvalues are configured at a certain point or concentrated in a very small area, then usually the response speed is slow and the execution signal amplitude is large. Therefore, all matrices should have their eigenvalues evenly configured on Figure 4 a circle with a radius of r within the sector area shown in Figure 4 . The larger the radius of the circle, the faster the response speed. In the present invention, the sector arc with a radius r less than 1 and a central angle of (n - 1)° symmetric about the real axis is evenly divided into n - 1 segments, and n / 2 pairs of conjugate complex eigenvalues are obtained by taking the coordinates of each node, as shown in

[0097] The Lyapunov equation can be used to calculate a reasonable gain matrix such that the eigenvalues of matrix are the selected conjugate complex eigenvalues. According to the duality theorem, the matrix pair is controllable. An observable matrix pair is constructed. Matrix F is a modal matrix constructed from n / 2 pairs of conjugate complex eigenvalues. The real part of the eigenvalues of matrix F is on the main diagonal position, and the imaginary part is on the off - diagonal position. Its dimension is n×n, while matrix has a dimension of N×n, and for each block corresponding to matrix F, there is at least one column that is not all zero. Matrix F is shown in the following formula (13):

[0098] (13)

[0099] Solve the Lyapunov equation as shown in the following formula (14). If matrix G is singular, replace it with a new matrix and solve it again. If matrix G is non - singular, calculate the gain matrix from the following formula (15) such that the eigenvalues of matrix are the same as the desired n / 2 pairs of conjugate complex eigenvalues of matrix F, thus completing the establishment of the estimation model for estimating the temperature change in the target area.

[0100] (14)

[0101] (15)

[0102] Since and are zero in the present invention, the control structure diagram of the Luenberger observer can be as shown in Figure 5 . It can be based on and As the input of Equations (7) and (9), the true system of the k-th time step is fitted by the estimation model of the k-th time step in Equation (9), and then the error of the estimation model is determined, and the gain matrix is calculated according to Equation (12). , such that the matrix has the same eigenvalues as the matrix F, which are the desired n / 2 pairs of conjugate complex eigenvalues. By updating the feedback, the estimation model can perfectly fit the true system.

[0103] According to the embodiments of the present invention, a state estimator is introduced to estimate all values based on the true temperature of partial area temperature measurement points, that is, the estimation model is corrected using the deviation between the true temperature and the estimated value of the measurable area temperature in the target area, which can solve the problem of insufficient temperature measurement points in the target area and reduce the cost according to the sensors. Especially when the temperature in the target area changes rapidly, based on these limited measurement data, the global state can be speculated or estimated, and the state variables that cannot be directly measured can be supplemented, enabling the control system to use the estimated global state for optimal control and avoiding errors in the control strategy due to lack of comprehensive information.

[0104] According to the embodiments of the present invention, in the variable air volume control system of the actual target area, it is assumed that both the prediction time domain and the control time domain of the variable air volume constraint control are to , the measurable disturbances at the k-th time step and before are known, while the measurable disturbances after the k-th time step are unknown. Assuming that the measurable disturbances after the k-th time step remain unchanged, then for the number of people, there is , and thus the following Equation (16) can be obtained:

[0105] (16)

[0106] Estimate the temperature increments of all grid areas through the estimation model, and based on the state estimation value at the k-th time step and the predicted temperature , the and of each grid area at the (k + l)-th time step within the prediction time domain l can be derived from Equation (7) as shown in the following Equation (17), where represents calculating the (k + l)-th time step from the k-th time step:

[0107] (17)

[0108] Write the predicted temperature vector to within the prediction time domain and the input variable (air supply speed increment) in matrix form as shown in the following Equation (18) as and a matrix, each predicted temperature vector has a dimension of the state variable dimension n×1. Therefore, the matrix composed of the predicted temperature vectors is a column vector of n×l, that is, nl×1.

[0109] (18)

[0110] Based on equations (17) and (18), the variable air volume prediction equation within the prediction horizon can be obtained as shown in (19). Among them, each element in the coefficient matrices , , corresponding to the temperature increment, temperature, and occupancy increment is a square matrix of n×n. In particular, each element in is an identity matrix of n×n, each element in is an ordinary square matrix of n×n, and the coefficient matrix corresponding to the supply air velocity increment is an l×1 lower triangular matrix, and its elements are column vectors of n×1.

[0111] (19)

[0112] The optimization objective can be to make the temperature of each grid area within the target area in all prediction horizons equal to the design temperature. The controlled temperature is the actual temperature or the predicted temperature. The objective function is shown in the following equation (20):

[0113] (20)

[0114] Among them, is the prediction horizon to the weight coefficient of the nth grid area within, is the weight coefficient matrix of each grid area within the target area in the prediction horizon, which is a diagonal matrix of nl×nl, and the diagonal elements are the weight coefficients of each prediction error at different times in the objective function; is the design temperature of the nth grid area within the prediction horizon, is the design temperature matrix of each grid area within the target area in the prediction horizon, which is a column vector of nl, allowing different design values for each output at different times; , are the constraint column vectors of the supply air velocity, with a dimension of l, determined based on the air volume change range in the large space; , are the constraint column vectors of the supply air velocity increment, with a dimension of l, to avoid wind pressure fluctuations caused by large adjustments of the supply air velocity.

[0115] Substituting the variable air volume prediction equation (19) into the objective function (20) and removing the constant term that has nothing to do with the control variable, it can be arranged into a quadratic programming form with the air supply speed increment as the control variable, as shown in the following formula (21):

[0116] (21)

[0117] The constraint conditions of the air supply speed and its increment can be arranged as shown in the following formula (22), where both the I matrix and the O matrix are square matrices with dimensions of the control time domain to and the I matrix is the identity matrix, and the O matrix is a lower triangular matrix with elements all being 1, is the optimal value of the air supply speed calculated at the (k - 1)-th time step and is a known value at the k-th time step.

[0118] (22)

[0119] The objective function of the quadratic programming problem is a quadratic function, and the constraint conditions are linear inequalities. The interior point penalty function method is used to solve it. Based on formulas (21) and (22), the quadratic programming problem with the air supply speed as the variable is as shown in the following formula (23).

[0120] (23)

[0121] where, is a quadratic function.

[0122] According to an embodiment of the present invention, using the interior point penalty function method to solve the quadratic programming problem and control the air supply volume may include operations: transforming the quadratic programming problem into a logarithmic interior point penalty function; updating the variable using the adaptive learning rate algorithm, and gradually reducing the barrier factor of the logarithmic interior point penalty function, repeating the variable update until the termination criterion is met to obtain the updated variable; and completing the control of the interface air volume when it is determined that the preset number of iterations meets the iteration threshold.

[0123] According to an embodiment of the present invention, the interior point penalty function method can substitute the constraint as a penalty term into the objective function, thereby transforming the constrained optimization problem into an unconstrained optimization problem. The interior point penalty function method is applicable to solving inequality constraint problems. The so-called interior point means that the assumed initial point is within the constraint range. Transforming the above formula (23) into the logarithmic interior point penalty function form is as shown in the following formula (24):

[0124] (24)

[0125] where, is the interior point penalty function, is the barrier factor; is a row vector with a dimension of 4l, and all elements are 1.

[0126] The solution principle of the interior point penalty function is to assume an initial optimization control variable within the constraint range , and update it iteratively to make the penalty function tend to zero. When the optimization variable approaches the constraint boundary, the logarithmic term of the penalty function increases rapidly, thus avoiding exceeding the boundary. Let the barrier factor gradually decrease, and the optimal solution under the current barrier factor is used as the starting point for the next optimization until the termination criterion is met:

[0127] (25)

[0128] where is a constant used to determine whether the optimization has converged and can be a small constant

[0129] For a given barrier factor, the control variable is updated iteratively using the adaptive learning rate algorithm (Adam). The calculation formulas for the first and second moments of the Adam algorithm and the update method for the coefficients to be determined in the model are expressed as shown in the following formula (26):

[0130] (26)

[0131] where are the first and second moments of the gradient of the coefficients to be determined in the Adam algorithm respectively are the decay rate factors of the first and second moments respectively, usually taking 0.9 and 0.99 is the initial learning rate of the coefficients to be determined, taking 0.001 is the smoothing control parameter for updating the coefficients to be determined, taking 10 -6 ; is the gradient of the coefficients to be determined are the coefficients to be determined, including the weight coefficient and the bias

[0132] The derivative of the logarithmic interior point penalty function with respect to the control variable is shown in the following formula (27):

[0133] (27)

[0134] where is the Jacobian matrix of the column vector and the column vector of control variables , with a dimension of 4l×l, and the general formula for the elements of is shown in formula (28); is the Jacobian matrix of the column vector and the column vector , with a dimension of 4l×4l and being a diagonal matrix, and the general formula for the elements on the diagonal is shown in the following formula (29).

[0135] (28)

[0136] (29)

[0137] Among them, is the element in the i-th row and j-th column of the matrix ; is the i-th element of the diagonal of the matrix ; is all the elements in the i-th row of the matrix W; is the i-th element of the column vector Z; is the j-th element of the control variable column vector .

[0138] According to the embodiments of the present invention, the difference between the temperature of each grid area in the target area and the designed temperature is determined as the objective function, constraints are set for the air supply volume and the air supply speed, the variable air volume control optimization problem based on the state estimator is transformed into a standard quadratic programming form, the quadratic programming problem is solved in real time by the interior point penalty function method, and variable air volume adjustment is performed to achieve air supply on demand, effectively avoiding the energy consumption problem of one-size-fits-all air supply in the entire space of the target area.

[0139] Figure 6 Shows a flowchart for variable air volume control according to an embodiment of the present invention.

[0140] As Figure 6 shown, the method for variable air volume control in this embodiment includes operations S601 to S614.

[0141] In operation S601, the first-order term coefficient and the constant term of the fitting relationship between the air volume and the air supply speed of the input interface are input.

[0142] In operation S602, the control time domain and the time step are initialized, and the initial values and initial increments of the air supply speed, the number of people, and the control temperature in each grid area are assumed.

[0143] In operation S603, the coefficient matrices , , of the state space equation are calculated.

[0144] In operation S604, the gain matrix L is calculated.

[0145] In operation S605, the temperature variable at the k-th time step is predicted through the estimation model.

[0146] In operation S606, a quadratic programming problem is constructed based on the temperature variable at the k-th time step.

[0147] In operation S607, initialize the obstacle factor μ 0 and the control variable ΔV(k) 0 .

[0148] In operation S608, adopt the Adam algorithm to update ΔV(k) 0 to obtain the updated control variable ΔV(k).

[0149] In operation S609, determine whether f(ΔV(k), μ 0 ) ≤ ε. If it is determined to be yes, execute operation S610. If it is determined to be no, repeat operation S608.

[0150] In operation S610, determine whether μ 0 Iln(WΔV(k) - Z) ≤ ε. If it is determined to be yes, execute operation S612 and output the control variable acting on the air in the grid area. If it is determined to be no, execute operation S611, set μ = 0.5μ 0 and then repeat operation S608.

[0151] In operation S613, determine whether k = k off . If it is determined to be yes, end the calculation. k off = k + l. If it is determined to be no, execute operation S614, set k′ = k + 1 and then repeat operation S605.

[0152] According to the embodiments of the present invention, the variable air volume control method based on image recognition proposed by the present invention can solve the core problems in environmental control and energy management of large - space buildings (such as subway stations, exhibition halls, airports, etc.). By using real - time image recognition technology to obtain the number of people, and combining the fitting relationship between grid temperature, interface air volume and supply air speed, the operation parameters of the air in the grid area are dynamically optimized, realizing on - demand energy supply and precise control, significantly improving energy utilization efficiency, reducing operation costs, and at the same time ensuring the comfort requirements of crowded areas, which is expected to help the construction industry achieve green and low - carbon goals. Taking the subway station as an example, the variable air volume control method based on image recognition proposed by the present invention can dynamically adjust the supply air volume of the station hall according to peak and off - peak hours, optimize the air flow organization, not only improve the comfort of passengers, but also reduce unnecessary energy consumption. Compared with traditional feedback control, the present invention can achieve forward - looking regulation by predicting temperature changes, optimize the operation parameters of the air in the grid area in advance, reduce the response delay of the system, and reduce energy waste caused by frequent adjustments.

[0153] Figure 7 Shows a block diagram of the variable air volume control device based on image recognition according to an embodiment of the present invention.

[0154] As shown Figure 7 in FIG. 1, the variable air volume control device 700 based on image recognition includes an analog module 710, a division module 720, a determination module 730, and a control module 740.

[0155] The analog module 710 is configured to simulate turbulent flow for a target area based on preset simulation conditions to obtain the temperature distribution of the target area. The division module 720 is configured to divide the target area into grids of a predetermined size based on the temperature distribution of the target area to obtain a plurality of grid areas. The determination module 730 is configured to determine the regional energy balance equation of each grid area based on the fitting relationship between the interface air volume and the air supply speed, the control temperature of each grid area, and the number of people in each grid area. The control module 740 is configured to control the air supply speed with the optimization objective of minimizing the difference between the control temperature of each grid area and the designed temperature control temperature, where the number of people in each grid area is determined by image recognition of the video data of the collected grid area.

[0156] According to an embodiment of the present invention, the variable air volume control device 700 based on image recognition further includes a first temperature determination module. The first temperature determination module is configured to determine the actual temperature of each grid area as the control temperature when the actual temperature of each grid area in the collected grid area is successfully determined.

[0157] According to an embodiment of the present invention, the variable air volume control device 700 based on image recognition further includes: a construction module, a prediction module, a second temperature determination module, and a third temperature determination module. The construction module is configured to construct an estimation model for estimating the temperature change in the target area based on a temperature state observer when the determination of the actual temperature of each grid area in the collected grid area fails and the determination of the actual temperature of some grid areas in the collected grid area is successful. The prediction module is configured to predict the actual temperature of other grid areas in the grid area except the partial grid areas by using the estimation model and the actual temperature of the partial grid areas to obtain the respective predicted temperatures. The second temperature determination module is configured to determine the actual temperature of each partial grid area as the control temperature of each partial grid area. The third temperature determination module is configured to determine the predicted temperature of each other grid area as the control temperature of each other grid area.

[0158] According to an embodiment of the present invention, the variable air volume control device 700 based on image recognition further includes: a detection module and a counting module. The detection module is configured to, for each grid area: detect the human contour of the collected video data and mark the candidate boxes. The counting module is configured to determine the number of people according to the number of candidate boxes.

[0159] According to an embodiment of the present invention, any multiple of the simulation module 710, the partitioning module 720, the determination module 730, and the control module 740 may be combined and implemented in one module, or any one of them may be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules may be combined with at least part of the functions of other modules and implemented in one module. According to an embodiment of the present disclosure, at least one of the simulation module 710, the partitioning module 720, the determination module 730, and the control module 740 may be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on chip, a system on substrate, a system on package, an application specific integrated circuit (ASIC), or any other reasonable way of integrating or packaging circuits, etc., implemented by hardware or firmware, or implemented in any one of the three implementation manners of software, hardware, and firmware, or in an appropriate combination of any several of them. Alternatively, at least one of the simulation module 710, the partitioning module 720, the determination module 730, and the control module 740 may be at least partially implemented as a computer program module, and when the computer program module is run, the corresponding functions may be executed.

[0160] It should be noted that the part of the variable air volume control device based on image recognition in the embodiment of the present invention corresponds to the part of the variable air volume control method based on image recognition in the embodiment of the present invention. For the description of the part of the variable air volume control device based on image recognition, please refer to the part of the variable air volume control method based on image recognition, which will not be elaborated here.

[0161] Those skilled in the art can understand that the features described in the various embodiments of the present invention can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in the present invention. In particular, without departing from the spirit and teachings of the present invention, the features described in the various embodiments of the present invention can be combined and / or combined in various ways. All such combinations and / or combinations fall within the scope of the present invention.

[0162] The embodiments of the present invention have been described above. However, these embodiments are only for illustrative purposes and not for limiting the scope of the present invention. Although the embodiments have been described separately above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. The scope of the present invention is defined by the appended claims and their equivalents. Without departing from the scope of the present invention, those skilled in the art can make various substitutions and modifications, and all such substitutions and modifications should fall within the scope of the present invention.

Claims

1. A variable air volume control method based on image recognition, characterized in that: The variable air volume control method comprises: Based on preset simulation conditions, simulating turbulent flow in a target area to obtain a temperature distribution in the target area; Based on the temperature distribution of the target area, dividing the target area into grids of a predetermined size to obtain a plurality of grid areas; Determining a regional energy balance equation for each of the grid areas based on a fitting relationship between the interface air volume and the air supply speed, the control temperature of each of the grid areas, and the number of personnel in each of the grid areas; and The air supply speed is controlled with the optimization goal of minimizing the difference between the control temperature and the design temperature of each grid area, wherein the number of people in each grid area is determined based on image recognition of the video data collected from the grid area.

2. The variable air volume control method according to claim 1, characterized in that: The regional energy balance equation of each grid area is determined based on the fitting relationship between the interface air volume and the air supply speed, the control temperature of each grid area, and the number of personnel in each grid area, including: Determine a first heat quantity for the air in the grid area based on a difference between a temperature of an area adjacent to the grid area and a temperature of the grid area, an air parameter of the grid area, and the fitting relationship; Determining a second heat quantity for the air in the grid area based on the number of people in the grid area and the average heat dissipation per person; Determine a third heat amount for the air in the grid area based on a difference between the air supply temperature of the grid area and the temperature of the grid area, an air supply port area of ​​the grid area, an air supply speed of the grid area, and an air parameter of the grid area; determining a total amount of heat based on the first amount of heat, the second amount of heat, and the third amount of heat; Determining a change in the internal energy of the air in the grid area based on the air parameters of the grid area, the volume of the grid area, and the temperature of the grid area; and Based on the total heat being equal to the change in internal energy, the regional energy balance equation is obtained.

3. The variable air volume control method according to claim 1, characterized in that: The control of the air supply speed with minimizing the difference between the control temperature and the design temperature of each grid area as an optimization goal comprises: Performing linear transformation based on nonlinear terms in the regional energy balance equation; Determining an incremental form of the regional energy balance equation based on the transformed nonlinear term and the regional energy balance equation; converting the incremental form into a matrix form; Discretizing the regional energy balance equation in matrix form by analytical method to obtain a state space equation with the increment of the control temperature of each grid area as output; Based on the state-space equation and the measurable disturbance condition, determining the state-space equation after iterating a preset number of steps; Based on the state space equation after the preset number of iteration steps, the temperature increment and the air supply speed increment within the preset number of iteration steps are converted into respective corresponding matrix forms; Determine a variable air volume prediction equation based on the matrix forms corresponding to the temperature increment and the air supply speed increment within the preset number of iteration steps; According to the optimization goal, determining the objective function; According to the variable air volume prediction equation and the objective function, a quadratic programming problem with the air supply speed as a variable is obtained; and The quadratic programming problem is solved by using an interior point penalty function method to control the air supply speed.

4. The variable air volume control method according to claim 3, characterized in that: The method of solving the quadratic programming problem by using the interior point penalty function method to control the air supply speed includes: Converting the quadratic programming problem into a logarithmic interior point penalty function; Using an adaptive learning rate algorithm to update the variable, gradually reducing the barrier factor of the logarithmic interior point penalty function, and repeatedly updating the variable until a termination criterion is met, thereby obtaining an updated variable; and When it is determined that the preset number of iteration steps meets the iteration threshold, the control of the air supply speed is completed.

5. The variable air volume control method according to claim 1, characterized in that: The variable air volume control method further includes: in a case where it is determined that the real temperatures of each grid area are collected successfully, determining the real temperatures of each grid area as the control temperature.

6. The method according to claim 1, characterized in that The method further comprises: In the case where it is determined that the acquisition of the respective real temperatures within the grid area fails, but the acquisition of the respective real temperatures of some grid areas within the grid area succeeds, constructing an estimation model for estimating the temperature change within the target area based on a temperature state observer; Using the estimation model and the respective true temperatures of the partial grid areas, predicting the respective true temperatures of the other grid areas within the grid area except the partial grid areas to obtain respective predicted temperatures; determining the respective real temperatures of the partial grid areas as the respective control temperatures of the partial grid areas; and The predicted temperature of each of the other grid areas is determined as the control temperature of each of the other grid areas.

7. The variable air volume control method according to any one of claims 1 to 6, characterized in that: The step of dividing the target area into grids of a predetermined size based on the temperature distribution of the target area to obtain a plurality of grid areas includes: Determining temperature changes within the target area according to the temperature distribution of the target area; When it is determined that the temperature change in the target area meets the preset temperature change condition, dividing the target area into a plurality of grid areas of a first predetermined size; When it is determined that the temperature change in the target area does not meet the preset temperature change condition, the target area is divided into a plurality of grid areas of a second predetermined size, wherein the first predetermined size is smaller than the second predetermined size.

8. The variable air volume control method according to any one of claims 1 to 6, characterized in that: The variable air volume control method further comprises: For each of the grid regions: Performing person contour detection on the collected video data and marking candidate frames; and The number of personnel is determined according to the number of the candidate boxes.

9. A variable air volume control device based on image recognition, characterized in that: The variable air volume control device comprises: A simulation module, used to simulate turbulent flow in a target area based on preset simulation conditions to obtain a temperature distribution in the target area; A division module, used for dividing the target area into grids of predetermined sizes based on the temperature distribution of the target area to obtain a plurality of grid areas; a determination module, configured to determine a regional energy balance equation for each of the grid areas based on a fitting relationship between the interface air volume and the air supply speed, a control temperature of each of the grid areas, and the number of personnel in each of the grid areas; and A control module is used to control the air supply speed with the optimization goal of minimizing the difference between the control temperature and the design temperature of each grid area, wherein the number of people in each grid area is determined based on image recognition of the video data collected from the grid area.

10. An electronic device, comprising: one or more processors; a memory for storing one or more computer programs, It is characterized in that the one or more processors execute the one or more computer programs to implement the steps of the variable air volume control method according to any one of claims 1 to 8.

Images