Intelligent sensing method for ventilation operation load state of air conditioner and related equipment

Abstract

The invention discloses an air conditioner ventilation operation load state intelligent sensing method and related equipment, and relates to the field of building energy management. The method comprises the following steps: firstly, constructing a dynamic heat flow topology network comprising a monitoring sub-region and a virtual sub-region; then, combining convection, diffusion and vertical thermal buoyancy physical mechanisms, and calculating an asymmetric transfer coefficient between nodes; deducing the environment state of the virtual sub-region through the equivalent heat flow distance, and carrying out the weighted fusion of the historical state and the current space aggregation feature through a time smoothing coefficient, and generating a global correction state vector. The problem of load sensing deviation caused by a vertical thermal stratification effect and a monitoring blind area in a high and large space is effectively solved, and high-precision operation load state monitoring and precise air volume adjustment in the whole space are achieved.

Description

Technical Field

[0001] This application relates to the field of building energy management, and in particular to an intelligent sensing method and related equipment for the operating load status of air conditioning and ventilation. Background Technology

[0002] For tall building spaces such as stadiums and industrial plants, traditional air conditioning and ventilation control usually relies on PID feedback regulation using limited measuring points installed on walls or return air vents. However, due to the unique vertical thermal stratification effect and large-volume airflow lag characteristics of tall buildings, it is difficult to accurately represent the overall thermal distribution of the space by simply relying on local single-point monitoring. This results in deviations and delays in the system's response to actual load changes in different areas, reducing energy efficiency and the uniformity of environmental control.

[0003] To address the aforementioned issues of control lag and incomplete perception, a smart environmental control method based on IoT multi-source sensing and random forest algorithm has been proposed. This technology deploys a high-density sensor network within the target area to collect multi-dimensional data such as temperature and humidity in real time. It then uses signal processing techniques such as Fast Fourier Transform to extract data features, which are subsequently input into machine learning models such as random forest for training and prediction. By establishing a data-driven mapping relationship between environmental parameters and system operating status, this method combines historical data with real-time monitoring values, thus solving to some extent the problem that traditional methods cannot predict load trends and improving the adaptive capability of air conditioning and ventilation systems to cope with environmental changes.

[0004] However, when dealing with complex dynamic loads in tall spaces, the aforementioned technical solutions often rely on purely data-driven models that focus on mining statistical correlations between data. They tend to treat sensor data distributed in different spatial locations as independent variables with equal weighting, making it difficult to explicitly characterize the complex physical topology and heat and moisture transfer coupling mechanisms within tall spaces. Consequently, when faced with nonlinear changes in the thermal field caused by personnel movement or airflow migration, the accuracy of sensing the operating load status under complex dynamic conditions is reduced. Summary of the Invention

[0005] This application provides an intelligent sensing method and related equipment for the operating load status of air conditioning and ventilation systems, which can improve the accuracy of sensing the operating load status of air conditioning in a target building under complex dynamic conditions.

[0006] Firstly, this application provides an intelligent sensing method for the operating load status of air conditioning and ventilation systems, applied to a server of an air conditioning and ventilation control system. The method includes: discretizing the target building space into several monitoring sub-areas and establishing a basic structure for a dynamic heat flow topology network connecting each monitoring sub-area. The dynamic heat flow topology network includes logical edges connecting any two adjacent monitoring sub-areas, and assigns an asymmetric transfer coefficient to each logical edge; acquiring environmental monitoring data of each monitoring sub-area in real time and mapping the environmental monitoring data to the initial local state vector of the corresponding monitoring sub-area. The environmental monitoring data includes at least air temperature, relative humidity, and airflow velocity vectors; and calculating any... The convective transfer intensity between two adjacent monitoring sub-areas is determined, and the diffusion transfer intensity corresponding to adjacent monitoring sub-areas is determined based on the air temperature difference between any two adjacent monitoring sub-areas. Based on the convective transfer intensity and diffusion transfer intensity, the asymmetric transfer coefficient of the corresponding logical edge of each monitoring sub-area is calculated and updated. Based on the updated dynamic heat flux topology network, the initial local state vector of each monitoring sub-area is aggregated using neighborhood features through the asymmetric transfer coefficient to obtain the global corrected state vector. The global corrected state vector is input into the preset load mapping model, and the real-time regional load value of each monitoring sub-area is calculated through nonlinear mapping. The real-time regional load values ​​of all monitoring sub-areas are accumulated to obtain the overall operating load state of the target building space.

[0007] By adopting the above technical solution, the server first establishes a dynamic heat flow topology network reflecting the physical heat transfer relationships between each monitoring sub-region, where the asymmetric transfer coefficient accurately describes the directional characteristics of heat flow transfer. Then, the server calculates the convective transfer intensity using real-time acquired airflow velocity vectors and determines the diffusion transfer intensity using temperature differences. The combination of these two key physical parameters allows the asymmetric transfer coefficient to dynamically reflect the actual heat and moisture transfer state. Next, the server updates the dynamic heat flow topology network based on the convective and diffusion transfer intensities, improving the real-time performance and effectiveness of the physical heat transfer relationships between each monitoring sub-region within the dynamic heat flow topology network. Finally, the server integrates the local state information of each monitoring sub-region with the heat flow transfer influence of its neighboring areas through neighborhood feature aggregation, forming a globally corrected state vector containing spatial context information. This globally corrected state vector, when input into the load mapping model, accurately reflects the heat transfer coupling relationship between different regions. In summary, this solution solves the problem that traditional single-point monitoring methods struggle to characterize the overall spatial thermal distribution state, improving the accuracy of load state perception under complex dynamic operating conditions.

[0008] In conjunction with some embodiments of the first aspect, in some embodiments, an asymmetric transfer coefficient is assigned to each logical edge, specifically including: determining the positional relationship between two adjacent monitoring sub-areas in the vertical direction; if the two adjacent monitoring sub-areas are arranged vertically, and the air temperature data of the upper monitoring sub-area is higher than that of the lower monitoring sub-area, then a vertical thermal buoyancy gain coefficient is generated; the asymmetric transfer coefficient of the logical edge from the lower monitoring sub-area to the adjacent upper monitoring sub-area is corrected using the vertical thermal buoyancy gain coefficient to obtain the corrected asymmetric transfer coefficient.

[0009] By employing the above technical solution, the server first determines the vertical positional relationship between two adjacent monitoring sub-regions, enabling the identification of the vertical thermal stratification phenomenon commonly found in tall building spaces. Then, when the server detects that the air temperature data of the upper monitoring sub-region is higher than that of the lower monitoring sub-region, it generates a vertical thermal buoyancy gain coefficient. This coefficient quantifies the enhancing effect of thermal buoyancy on vertical heat transfer. Next, the server uses the vertical thermal buoyancy gain coefficient to correct the asymmetric transfer coefficient of the logical edge from the lower monitoring sub-region to the adjacent upper monitoring sub-region, ensuring that the transfer coefficient in the dynamic heat flow topology network accurately reflects the natural convection heat transfer mechanism driven by thermal buoyancy. Finally, the server uses the corrected asymmetric transfer coefficient to correctly balance the impact of vertical heat transfer during neighborhood feature aggregation. In summary, this solution solves the load prediction bias problem caused by neglecting the buoyancy effect in traditional methods, improving the accuracy of perceiving the vertical thermal stratification effect in target building spaces.

[0010] In conjunction with some embodiments of the first aspect, in some embodiments, environmental monitoring data of each monitoring sub-region is acquired in real time, and the environmental monitoring data is mapped to the initial local state vector of the corresponding monitoring sub-region. Specifically, this includes: dividing the monitoring sub-region into a measured sub-region with sensors and a virtual sub-region without sensors; the initial local state vector of the measured sub-region is directly determined by the environmental monitoring data collected by the sensors; for any virtual sub-region, adjacent measured sub-regions with logical edges connected to the virtual sub-region are determined in the dynamic heat flux topology network; the spatial Euclidean distance between the virtual sub-region and each adjacent measured sub-region is calculated; based on the environmental monitoring data of each adjacent measured sub-region and the spatial Euclidean distance, the simulated environmental monitoring data of the virtual sub-region is obtained by inverse distance weighting, and the simulated environmental monitoring data is mapped to the initial local state vector of the virtual sub-region.

[0011] By adopting the above technical solution, the server first divides the monitoring sub-area into actual measurement sub-areas with deployed sensors and virtual sub-areas without deployed sensors, balancing the contradiction between sensor deployment costs and spatial coverage. Then, the server identifies adjacent actual measurement sub-areas logically connected to the virtual sub-area in the dynamic heat flux topology network and calculates the spatial Euclidean distance between the virtual sub-area and each adjacent actual measurement sub-area, providing a geometric basis for interpolation calculations. Next, based on the environmental monitoring data and spatial Euclidean distances of each adjacent actual measurement sub-area, the server calculates the simulated environmental monitoring data of the virtual sub-area through inverse distance weighting, ensuring that the closer the actual measurement sub-area is, the greater its influence on the virtual sub-area, conforming to the distance attenuation law of physical heat transfer. Finally, the server maps the simulated environmental monitoring data to the initial local state vector of the virtual sub-area, enabling the entire monitoring network to achieve full coverage monitoring of the building space with a limited number of sensors. In summary, this solution solves the problem of sensing blind spots caused by insufficient sensor coverage, reduces system deployment and maintenance costs, and ensures the spatial integrity of subsequent load calculations.

[0012] In conjunction with some embodiments of the first aspect, in some embodiments, based on the environmental monitoring data of each adjacent measured sub-region and the spatial Euclidean distance, the simulated environmental monitoring data of the virtual sub-region is obtained by inverse distance weighting calculation. Specifically, this includes: obtaining the asymmetric transfer coefficient of the logical edge connecting the virtual sub-region and each adjacent measured sub-region; correcting the spatial Euclidean distance by division using the asymmetric transfer coefficient to obtain the equivalent heat flux distance; and using the reciprocal of the equivalent heat flux distance as a weight to perform a weighted average of the environmental monitoring data of each adjacent measured sub-region to obtain the simulated environmental monitoring data of the virtual sub-region.

[0013] By employing the above technical solution, the server first obtains the asymmetric transfer coefficients of the logical edges connecting the virtual sub-region and each adjacent measured sub-region. These asymmetric transfer coefficients carry information about the actual heat transfer intensity and directionality. Then, the server corrects the spatial Euclidean distance using the asymmetric transfer coefficients to obtain the equivalent heat flow distance. This equivalent heat flow distance considers not only geometric spatial relationships but also the actual physical mechanism of heat transfer. Next, when the transfer coefficient is large, the equivalent distance decreases, indicating a stronger heat transfer capability along the corresponding path, and the virtual sub-region should be more influenced by the measured sub-region. Conversely, when the transfer coefficient is small, the equivalent distance increases, correspondingly weakening the influence weight. Finally, the server uses the reciprocal of the equivalent heat flow distance as a weight to perform a weighted average of the environmental monitoring data from each adjacent measured sub-region, ensuring that the simulated environmental monitoring data of the virtual sub-region accurately reflects the actual heat transfer patterns. In summary, this solution solves the problem of traditional geometric interpolation methods neglecting heat transfer mechanisms and improves the accuracy of virtual sub-region data reconstruction.

[0014] In conjunction with some embodiments of the first aspect, in some embodiments, the convective transfer intensity between any two adjacent monitoring sub-regions is calculated based on the airflow velocity vector in the initial local state vector of each monitoring sub-region. Specifically, this includes: based on two adjacent monitoring sub-regions, calculating the projection component of the airflow velocity vector of the upstream sub-region onto the line vector connecting the upstream sub-region to the downstream sub-region; if the projection component is positive, determining the convective transfer intensity as a first preset intensity value; if the projection component is negative, determining the convective transfer intensity as a second preset intensity value, wherein the first preset intensity value is greater than the second preset intensity value.

[0015] By employing the above technical solution, the server first calculates the projection component of the airflow velocity vector of the upstream sub-region onto the vector connecting the upstream and downstream sub-regions, based on two adjacent monitoring sub-regions. This projection component accurately determines whether the airflow flows along the connection direction between the two sub-regions. Then, when the server detects a positive projection component, it determines the convective transfer intensity as a first preset value, accurately reflecting the enhanced effect of forced convective heat transfer. Next, when the projection component is negative, the server determines the convective transfer intensity as a second preset value, avoiding erroneous convective transfer calculations, where the first preset value is greater than the second preset value. Finally, this directionality determination mechanism based on vector projection enables the server to accurately distinguish the strength of convective transfer, ensuring that the asymmetric transfer coefficient in the dynamic heat flux topology network truly reflects the airflow-driven heat transfer process. In summary, this solution solves the problem of traditional methods failing to accurately determine the directionality of airflow heat transfer, improving the physical accuracy of convective heat transfer modeling.

[0016] In conjunction with some embodiments of the first aspect, in some embodiments, based on the updated dynamic heat flux topology network, the initial local state vectors of each monitoring sub-region are aggregated using asymmetric transfer coefficients to obtain a global corrected state vector. Specifically, this includes: obtaining the historical global corrected state vectors calculated for each monitoring sub-region at the previous time; using asymmetric transfer coefficients to perform weighted summation of the initial local state vectors of each monitoring sub-region at the current time to obtain the spatial aggregation features at the current time; calculating the rate of change of the airflow velocity vector data at the current time relative to the previous time; determining a time smoothing coefficient based on the rate of change; and using the time smoothing coefficient to perform weighted fusion of the historical global corrected state vectors and the spatial aggregation features to obtain the global corrected state vector at the current time.

[0017] By employing the above technical solution, the server first obtains the historical global corrected state vectors calculated by each monitoring sub-region at the previous moment, providing a time series continuity basis for the calculation at the current moment. Then, the server uses an asymmetric transfer coefficient to weightedly sum the initial local state vectors of each monitoring sub-region at the current moment, obtaining the spatial aggregation feature at the current moment, which reflects the current spatial heat transfer relationship. Next, the server calculates the rate of change of the airflow velocity vector data at the current moment relative to the previous moment. This rate of change reflects the dynamic stability of the building's internal environment, and a time smoothing coefficient is determined based on the rate of change. Finally, the server uses the time smoothing coefficient to weightedly fuse the historical global corrected state vectors and the spatial aggregation feature to obtain the global corrected state vector at the current moment. This global corrected state vector contains both the real-time spatial heat transfer information at the current moment and maintains the smoothness and continuity of the time series. In summary, this solution solves the problem that simple spatial aggregation is easily affected by instantaneous fluctuations, improving the stability and accuracy of load state perception.

[0018] In conjunction with some embodiments of the first aspect, in some embodiments, after obtaining the overall operating load status of the target building space, the method further includes: calculating the thermal distribution variance of each monitoring sub-area in the global corrected state vector; when the monitored change in the overall operating load status exceeds a preset change threshold and the thermal distribution variance is greater than a preset uniformity threshold, selecting the monitoring sub-area with the largest real-time regional load value, and generating an airflow adjustment command for the corresponding monitoring sub-area.

[0019] By employing the above technical solution, the server first calculates the heat distribution variance of each monitoring sub-region in the global corrected state vector. This heat distribution variance accurately quantifies the degree of temperature unevenness within the entire building space. Then, the server monitors the variation in the overall operating load state. When the variation exceeds a preset variation threshold, it indicates an abnormal change in the building load. Next, the server jointly determines the load change and heat distribution uniformity. When the heat distribution variance is greater than a preset uniformity threshold, the monitoring sub-region with the largest real-time regional load value is selected. This monitoring sub-region is usually the main source of uneven heat distribution and load changes. Finally, the server generates airflow adjustment commands for the corresponding monitoring sub-region, increasing or decreasing the airflow in that region to balance the overall heat distribution and achieve precise local adjustment. In summary, this solution solves the problems of lag and excessive energy consumption in traditional global adjustment methods, improving the response accuracy and energy efficiency of the air conditioning system.

[0020] In a second aspect, this application provides a server for an air conditioning and ventilation control system. The server includes one or more processors and a memory. The memory is coupled to one or more processors and is used to store computer program code, which includes computer instructions. One or more processors invoke the computer instructions to cause the server to perform the method described in the first aspect and any possible implementation thereof.

[0021] Thirdly, embodiments of this application provide a computer-readable storage medium including instructions that, when executed on a server, cause the server to perform the method described in the first aspect and any possible implementation thereof.

[0022] Fourthly, embodiments of this application provide a computer program product containing instructions, including a computer program / instructions that, when run on a server, cause the server to perform the method described in the first aspect and any possible implementation thereof.

[0023] It is understood that the server provided in the second aspect, the computer-readable storage medium provided in the third aspect, and the computer program product provided in the fourth aspect are all used to execute the methods provided in the embodiments of this application. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods, and will not be repeated here.

[0024] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0025] 1. By adopting a technical solution of establishing a dynamic heat flow topology network connecting each monitoring sub-area and performing neighborhood feature aggregation, the server can dynamically fuse the local state of each monitoring sub-area with the heat flow transfer influence of its neighboring areas using the real-time updated asymmetric transfer coefficient, forming a global corrected state vector containing spatial context information. This effectively solves the problem in related technologies that it is difficult to truly represent the complex heat transfer and thermal distribution state inside large spaces by simply relying on local single-point monitoring or pure data-driven models, thereby achieving accurate perception of the target building space's operating load state under complex dynamic conditions.

[0026] 2. By employing a technical solution that uses the vertical thermal buoyancy gain coefficient to correct the asymmetric transfer coefficient of the logical edge from the lower monitoring sub-area to the adjacent upper monitoring sub-area, the server can quantify the enhancing effect of thermal buoyancy on vertical heat transfer. This allows the dynamic heat flow topology network to accurately reflect the natural convection heat transfer mechanism driven by thermal buoyancy, effectively solving the problem of heat flow transfer modeling deviation caused by neglecting the vertical thermal stratification effect when dealing with loads in tall spaces in related technologies. This results in the accurate capture of the unique vertical thermal stratification phenomenon in tall building spaces and the improvement of load prediction accuracy.

[0027] 3. By employing a technical solution that uses a time smoothing coefficient to weight and fuse historical global correction state vectors and spatial aggregation features, the server can dynamically adjust the weight ratio of historical information and current spatial features based on the rate of change of airflow velocity vector data. This allows for rapid response when environmental changes are drastic and noise suppression when the environment is stable. It effectively solves the problem in related technologies where relying solely on instantaneous spatial features is easily affected by sensor noise or instantaneous airflow fluctuations, leading to unstable load perception results. This results in a dual improvement in the smoothness of the load status perception process in the time domain and its adaptability to dynamic environments. Attached Figure Description

[0028] Figure 1 This is a flowchart illustrating an intelligent sensing method for the operating load status of air conditioning and ventilation in an embodiment of this application.

[0029] Figure 2 This is another flowchart illustrating an intelligent sensing method for air conditioning and ventilation operation load status in an embodiment of this application;

[0030] Figure 3 This is a schematic diagram of the physical device structure of a server for an air conditioning and ventilation control system in an embodiment of this application. Detailed Implementation

[0031] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification of this application, the singular expressions “a,” “an,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.

[0032] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0033] To facilitate understanding, the application scenarios of the embodiments of this application are described below.

[0034] In related technologies, sparse single-point monitoring combined with traditional PID feedback control or pure data-driven statistical learning models can be used to regulate building environmental parameters and estimate loads. The following describes a scenario using an intelligent sensing method for the operating load status of air conditioning and ventilation systems. For tall building spaces such as stadiums and industrial plants, traditional methods often rely on limited sensors installed on walls or return air vents for feedback regulation, or simply utilize machine learning to mine statistical correlations between data. However, due to the unique vertical thermal stratification effect and large-volume airflow lag characteristics of tall spaces, relying solely on local single-point monitoring is insufficient to accurately represent the overall thermal distribution of the space; while pure data-driven models often ignore the objectively existing physical topology and heat and moisture transfer coupling mechanisms between nodes in different areas, resulting in insufficient accuracy of load sensing and biased and delayed system responses when facing nonlinear changes in the thermal field caused by personnel movement or airflow migration.

[0035] The dynamic heat flow topology network and asymmetric transfer coefficient based on the physical heat transfer mechanism in this application embodiment are used to calculate the convection transfer intensity and diffusion transfer intensity to update the network weight in real time. Furthermore, neighborhood feature aggregation technology is used to achieve deep integration of the local state of each monitoring sub-region and the impact of spatial heat flow transfer. This not only accurately describes the dynamic transfer process of heat and moisture in complex spaces, but also provides a scenario for using the intelligent sensing method for air conditioning and ventilation operation load status described in this application. The server discretizes the target building space and establishes a dynamic heat flow topology network connecting each monitoring sub-region. Using the real-time acquired airflow velocity vector and temperature difference, the convection and diffusion transfer intensity reflecting the actual physical mechanism are calculated, and the asymmetric transfer coefficient of the logical edge is updated. Subsequently, the initial local state of each sub-region is weighted and aggregated using this coefficient to generate a globally corrected state vector containing the spatial context. Finally, this vector is input into the load mapping model to obtain the accurate overall operating load status.

[0036] As can be seen, by adopting the dynamic heat flow topology network and neighborhood feature aggregation mechanism in the embodiments of this application, while realizing the deep integration of physical heat transfer laws and environmental monitoring data in the spatiotemporal dimensions, it can also effectively solve the problem of load perception lag and inaccuracy caused by ignoring spatial physical topology and vertical thermal stratification effect in related technologies, thereby realizing high-precision intelligent perception of the air conditioning and ventilation operation load status of the target building space under complex dynamic working conditions.

[0037] To facilitate understanding, the method provided in this implementation will be described in detail below, using the above scenario as an example. Please refer to [link / reference]. Figure 1 This is a flowchart illustrating an intelligent sensing method for the operating load status of air conditioning and ventilation in an embodiment of this application.

[0038] S101. The target building space is discretized into several monitoring sub-areas, and a basic structure of a dynamic heat flow topology network connecting each monitoring sub-area is established. The dynamic heat flow topology network includes logical edges connecting any two adjacent monitoring sub-areas, and each logical edge is assigned an asymmetric transfer coefficient.

[0039] In this context, the target building space refers to the interior space of tall buildings requiring air conditioning and ventilation load monitoring, such as stadiums, industrial plants, or large exhibition halls—building spaces with vertical thermal stratification effects. Discretization refers to a mathematical processing method that divides a continuous three-dimensional building space into a finite number of independently monitorable spatial units according to a preset grid size or functional area. The monitoring sub-region represents each independent spatial unit after discretization, with each monitoring sub-region having a clear three-dimensional coordinate boundary and a unique identifier. The dynamic heat flow topology network represents a mathematical graph structure describing the heat and airflow transfer relationships between the monitoring sub-regions. The network's connectivity and weighting coefficients change dynamically based on real-time environmental parameters. The basic structure refers to the initial framework of the dynamic heat flow topology network, including the static definitions of nodes (monitoring sub-regions), edges (connectivity), and initial weights. Logical edges represent abstract connections between any two adjacent monitoring sub-regions, used to quantify the heat flow transfer relationship between the two sub-regions. The asymmetric transfer coefficient represents a numerical parameter that quantifies the degree of heat or humidity transfer from one monitoring sub-region to another. This coefficient is directional, meaning that the transfer coefficient from sub-region A to sub-region B is different from the transfer coefficient from sub-region B to sub-region A.

[0040] Specifically, the server executes this step during the system initialization phase. First, it acquires the 3D geometric model and CAD drawings of the target building space. Based on the building's height, area, and functional zoning characteristics, it divides the entire space into N×M×H 3D monitoring sub-zones using either a regular mesh method or an adaptive mesh method. Each sub-zone's dimensions range from 3m×3m×2m to 10m×10m×5m. The server assigns a unique 3D coordinate index (i, j, k) and numerical ID to each monitoring sub-zone and establishes an adjacency matrix between sub-zones. Next, the server constructs the basic graph structure G=(V, E) of the dynamic heat flux topology network, where the node set V contains all monitoring sub-zones, and the edge set E contains the connections between all pairs of adjacent sub-zones. For each logical edge e(i, j), the server initializes the asymmetric transfer coefficients α in the two opposing directions. ij and α ji The default value is 0.5. These coefficients will be dynamically updated in subsequent steps based on real-time environmental data, thereby forming a dynamic topology network infrastructure that can reflect the actual heat flow transfer pattern.

[0041] S102. Real-time acquisition of environmental monitoring data for each monitoring sub-area, and mapping of the environmental monitoring data to the initial local state vector of the corresponding monitoring sub-area. The environmental monitoring data shall at least include air temperature, relative humidity and airflow velocity vector.

[0042] Among them, environmental monitoring data represents a set of physical quantity measurements that reflect the environmental state within the monitoring sub-area, including but not limited to temperature, humidity, airflow parameters, and pressure; the initial local state vector refers to a mathematical vector representation of the multidimensional environmental monitoring data of a certain monitoring sub-area organized according to a preset format; the airflow velocity vector refers to a three-dimensional vector describing the airflow within the monitoring sub-area, including X, Y, and Z directional components as well as the magnitude and direction information of the vector.

[0043] Specifically, the server periodically executes this step during normal system operation, receiving environmental monitoring data in real time from sensor nodes deployed in each monitoring sub-area via wireless communication protocols (such as ZigBee or WiFi). Each sensor node is equipped with a temperature sensor, a humidity sensor, and a 3D ultrasonic anemometer. The server preprocesses the received raw data, including outlier detection, noise filtering, and unit conversion. Next, the server maps the environmental monitoring data of each monitoring sub-area i into a 4D initial local state vector Si = [Ti, Hi, Vi_x, Vi_y, Vi_z], where Ti is the normalized temperature value (range 0-1), Hi is the normalized relative humidity value (range 0-1), and Vi_x, Vi_y, and Vi_z are the components of airflow velocity in the X, Y, and Z axes, respectively (unit: m / s). The server also records the timestamp and data quality identifier for each state vector to ensure that subsequent calculations use valid data from the same moment.

[0044] Optionally, in some embodiments, the server can also obtain a real-time heat map of the personnel density distribution inside the target building space, and map the heat map to each of the monitoring sub-areas to generate personnel heat dissipation intensity values ​​for the corresponding monitoring sub-areas.

[0045] When generating the initial local state vector, the personnel heat dissipation intensity value is extended to the initial local state vector as an independent dimension data feature;

[0046] When calculating the global corrected state vector, a corresponding source term gain coefficient is generated based on the personnel heat dissipation intensity value, and the spatial aggregation characteristics of the monitoring sub-region at the current moment are corrected using the source term gain coefficient to compensate for the nonlinear sensible heat and latent heat gain caused by personnel gathering.

[0047] S103. Based on the airflow velocity vector in the initial local state vector of each monitoring sub-region, calculate the convection transfer intensity between any two adjacent monitoring sub-regions, and determine the diffusion transfer intensity corresponding to the adjacent monitoring sub-regions based on the air temperature difference between any two adjacent monitoring sub-regions.

[0048] Among them, convective transfer intensity represents the ability of heat to be transferred from one monitoring sub-region to an adjacent monitoring sub-region due to airflow motion, and this intensity is directly related to the direction and magnitude of airflow velocity; diffusion transfer intensity represents the ability of heat caused by temperature gradient to be transferred between adjacent monitoring sub-regions through molecular diffusion or turbulent diffusion mechanisms.

[0049] Specifically, after obtaining the initial local state vectors of all monitoring sub-regions, the server performs this step. First, it traverses all logical edges in the dynamic heat flux topology network and calculates the convection transfer intensity for each pair of adjacent monitoring sub-regions (i, j). The server extracts the airflow velocity vector Vi = [Vi_x, Vi_y, Vi_z] of monitoring sub-region i and calculates the unit direction vector d from sub-region i to sub-region j. ij =[(x j -x i ) / |r ij |,(y j -y i ) / |r ij |,(z j -z i ) / |r ij |], where (x) i y i , z i ) and (x j y j , z j ) are the geometric center coordinates of the two sub-regions, |r ij | represents the distance between the two sub-intervals. Next, the server calculates the convection transfer intensity using vector dot product. This formula ensures that positive convection transfer only occurs when the airflow points from i to j. For calculating diffusion transfer intensity, the server first calculates the temperature difference ΔT. ij =|T i -T j Then, the diffusion transfer intensity is calculated according to the physical diffusion law. , where k is a diffusion coefficient constant (typically 0.1-0.3). The server performs the above calculations on all adjacent sub-region pairs in the network to generate convection transfer intensity matrix and diffusion transfer intensity matrix.

[0050] Optionally, in some embodiments, the server can calculate the projection component of the airflow velocity vector of the upstream sub-region onto the line vector connecting the upstream sub-region to the downstream sub-region based on two adjacent monitoring sub-regions;

[0051] If the projected component is positive, the convective transfer intensity is determined to be the first preset intensity value;

[0052] If the projected component is negative, the convective transfer intensity is determined to be the second preset intensity value, and the first preset intensity value is greater than the second preset intensity value.

[0053] S104. Based on the convection transfer intensity and diffusion transfer intensity, calculate and update the asymmetric transfer coefficient of the corresponding logical edge of each monitoring sub-region;

[0054] Specifically, the server executes this step immediately after calculating the convection and diffusion propagation strengths, using a weighted fusion method to calculate the asymmetric propagation coefficient of each logical edge. The server first sets the convection weight w. conv =0.7 and diffusion weight w diff =0.3 (These two preset weight thresholds are used to balance the relative importance of convection and diffusion heat transfer mechanisms, and can be adjusted according to building type and season), then according to the formula The asymmetric transfer coefficients from monitoring sub-region i to monitoring sub-region j are calculated. To ensure numerical stability and the rationality of physical meaning, the server normalizes the weights of all outgoing edges for each monitoring sub-region i. The server also sets a threshold δmax = 0.3 to limit the rate of change of the transfer coefficients (this threshold is used to prevent abrupt changes in coefficients that could lead to system instability; smoothing is performed when the difference between the new coefficient and the coefficient at the previous time step exceeds this threshold), and applies a time smoothing algorithm to avoid drastic fluctuations in coefficients. Finally, the server stores the updated asymmetric transfer coefficient matrix in the dynamic heat flux topology network, completing the real-time adjustment of the network structure.

[0055] Optionally, in some embodiments, the server can also obtain real-time solar radiation intensity and incident angle data of the location of the target building space, and identify the boundary monitoring sub-area located at the building boundary in the dynamic heat flow topology network.

[0056] The radiative heat flux injection amount of each boundary monitoring sub-region is calculated based on the incident angle and the spatial orientation of the boundary monitoring sub-region;

[0057] Before updating the asymmetric transfer coefficient of the logical edge corresponding to each of the monitoring sub-regions, the asymmetric transfer coefficient of the logical edge from the boundary monitoring sub-region to the adjacent internal monitoring sub-region is increased according to the amount of radiative heat flow injection, so as to characterize the directional penetration process of external radiative heat flow into the internal space.

[0058] Optionally, in some embodiments, the server may also perform sparsification pruning on the dynamic heat flow topology network after updating the asymmetric transfer coefficients of the logical edges corresponding to each monitoring sub-region.

[0059] Specifically, this includes: marking logical edges whose asymmetric transmission coefficient is lower than a preset weak transmission threshold as invalid edges and temporarily blocking them from the dynamic heat flow topology network;

[0060] When performing the neighborhood feature aggregation, weighted calculations are only performed on the retained valid logical edges;

[0061] When the temperature difference between the monitoring sub-regions at both ends of the invalid edge exceeds the preset activation threshold, the connection state of the logical edge is restored, thereby reducing the computational overhead of the server for feature aggregation through graph neural networks while ensuring the perception accuracy.

[0062] S105. Based on the updated dynamic heat flux topology network, the initial local state vector of each monitoring sub-region is aggregated with neighborhood features by using the asymmetric transfer coefficient to obtain the global corrected state vector.

[0063] Among them, the updated dynamic heat flux topology network refers to the heat flux topology network structure that includes the latest asymmetric transfer coefficient; neighborhood feature aggregation refers to the mathematical operation process of weighting and merging the state information of a certain monitoring sub-region and its neighboring monitoring sub-regions according to the transfer coefficient; the global corrected state vector represents the comprehensive state vector that can reflect the influence of the surrounding environment on the monitoring sub-region after the neighborhood information fusion.

[0064] Specifically, the server performs this step after the dynamic heat flux topology network update is complete, using a graph neural network message passing mechanism to aggregate neighborhood features for each monitoring sub-region. The server first constructs a neighborhood message set for each monitoring sub-region i, then iterates through all its neighboring monitoring sub-regions j∈N(i) to calculate the neighborhood influence message m. ij =α ji ×S j , where α ji S represents the asymmetric transfer coefficient from subregion j to subregion i. j This is the initial local state vector for monitoring sub-region j. Next, the server processes all neighborhood messages m of monitoring sub-region i. ij The total neighborhood influence vector is obtained by aggregation and summation. Then, the server uses a learnable feature transformation to fuse its own state and neighborhood influence, and calculates the global corrected state vector S'. i =σ(W self ×S i +W neigh ×M i +b i ), where W self and W neigh b is the pre-trained weight matrix (4×8 dimensions). i Let σ be the bias vector and σ be the ReLU activation function. The server performs the above aggregation process in parallel on all monitored sub-regions, generating a set of corrected state vectors {S'1, S'2, ..., S'} containing global spatial context information. N These vectors not only contain local environmental information, but also incorporate the effects of heat flow transfer from adjacent areas, thus overcoming the limitations of traditional single-point monitoring.

[0065] S106. Input the global correction state vector into the preset load mapping model, calculate the real-time regional load value of each monitoring sub-area through nonlinear mapping, and accumulate the real-time regional load values ​​of all monitoring sub-areas to obtain the overall operating load status of the target building space.

[0066] Among them, the preset load mapping model refers to a machine learning model that has been pre-trained to map environmental state vectors to load demand values, usually using a multilayer perceptron or recurrent neural network structure; nonlinear mapping refers to the mathematical transformation process of transforming input features into output results through neural network layers containing activation functions; real-time regional load value represents the air conditioning load demand of a certain monitoring sub-area at the current moment, including sensible heat load and latent heat load, in kilowatts (kW); overall operating load status represents the overall load demand status of the entire target building space, used to guide the operation and control of the air conditioning and ventilation system.

[0067] Specifically, after obtaining the global corrected state vectors of all monitored sub-regions, the server performs this step, first by modifying the global corrected state vector S' of each monitored sub-region i. i The load is input into a pre-defined load mapping model for inference calculations. This load mapping model employs a three-layer fully connected neural network structure. The input layer has 8 nodes corresponding to the corrected state vector dimension, the hidden layer has 16 nodes using the ReLU activation function, and the output layer has 2 nodes outputting the predicted values ​​of sensible heat load and latent heat load, respectively. The server performs forward propagation calculations to obtain the real-time regional load value of monitoring sub-area i, and checks the reasonableness of the prediction results, setting an upper limit threshold (this threshold is used to prevent abnormal prediction values; when the predicted load exceeds this threshold, it is truncated) and a lower limit threshold (e.g., 0kW) to ensure physical reasonableness. Next, the server performs weighted accumulation of the real-time regional load values ​​of all monitoring sub-areas to calculate the overall operating load status. Finally, the server transmits the calculated overall operating load status to the air conditioning and ventilation control system, updates the historical load database, and generates a load distribution heat map for maintenance personnel to refer to, completing a full intelligent sensing cycle.

[0068] Optionally, in some embodiments, the training process of the preset load mapping model includes the following steps:

[0069] The system collects operational data of the target building space over a pre-defined historical time period (e.g., the past year's operating cycle). This operational data includes historical environmental monitoring data (air temperature, relative humidity, airflow velocity vector) and corresponding historical actual load values ​​for each monitoring sub-area. The historical actual load values ​​can be directly read from heat meters installed at the end vents of each area, or calculated using the air enthalpy difference method based on historical return air enthalpy, supply air enthalpy, and supply air volume, and used as supervisory labels for model training.

[0070] Using the methods described in steps S101 to S105 above, a dynamic heat flux topology network for historical moments is constructed based on historical environmental monitoring data. The asymmetric transfer coefficient of each monitoring sub-region at each historical moment is calculated, and then neighborhood feature aggregation is performed to generate a historical global corrected state vector corresponding to each monitoring sub-region. This historical global corrected state vector is then used as the input feature for model training.

[0071] Construct a multilayer perceptron (MLP) or deep neural network (DNN) as the initial load mapping model. Set the number of neurons in the input layer to be consistent with the dimension of the global correction state vector (e.g., 8-dimensional), and the number of neurons in the output layer to correspond to the type of load value (e.g., 2-dimensional, corresponding to sensible heat load and latent heat load respectively). Initialize the weight matrix and bias vector of the hidden layers.

[0072] The historical input feature vectors are batch-input into the initial neural network model for forward propagation, outputting predicted load values. A loss function is constructed to calculate the mean squared error (MSE) between the predicted load values ​​and the historical true load values. Based on the MSE, the gradient is calculated using the back propagation algorithm, and the weight matrix and bias vector in the network model are updated using the Adam optimizer or stochastic gradient descent (SGD). The above training steps are repeated until the value of the loss function converges to a preset error range or reaches a preset number of iterations, resulting in a well-trained load mapping model.

[0073] In this embodiment, by establishing a dynamic heat flow topology network connecting each monitoring sub-region and updating the asymmetric transfer coefficient based on the convection transfer intensity and diffusion transfer intensity, and then performing neighborhood feature aggregation on the initial local state vector of each monitoring sub-region, a globally corrected state vector that incorporates the influence of heat flow transfer in the neighborhood space can be generated. This effectively solves the problem that traditional single-point monitoring is difficult to accurately represent the complex thermal distribution state in tall spaces, thereby achieving accurate perception of the overall operating load state of the target building space.

[0074] Based on the above embodiments, the method provided in this embodiment will be described in further detail below. Please refer to... Figure 2 This is another flowchart illustrating an intelligent sensing method for the operating load status of air conditioning and ventilation in this application.

[0075] S201. The target building space is discretized into several monitoring sub-areas, and a basic structure of a dynamic heat flow topology network connecting each monitoring sub-area is established. The dynamic heat flow topology network includes logical edges connecting any two adjacent monitoring sub-areas, and each logical edge is assigned an asymmetric transfer coefficient.

[0076] This step specifically includes:

[0077] Determine the vertical positional relationship between two adjacent monitoring sub-areas;

[0078] If two adjacent monitoring sub-areas are arranged vertically, and the air temperature data of the upper monitoring sub-area is higher than that of the lower monitoring sub-area, then a vertical thermal buoyancy gain coefficient is generated.

[0079] The asymmetric transfer coefficient of the logical edge from the lower monitoring sub-region to the adjacent upper monitoring sub-region is corrected using the vertical thermal buoyancy gain coefficient, resulting in the corrected asymmetric transfer coefficient.

[0080] Among them, positional relationship refers to the relative spatial position of two monitoring sub-areas in a three-dimensional coordinate system; vertical arrangement means that the two monitoring sub-areas are distributed vertically and their horizontal positions are basically overlapping; the monitoring sub-area located above refers to the monitoring sub-area with a larger vertical coordinate value; the monitoring sub-area located below refers to the monitoring sub-area with a smaller vertical coordinate value; the vertical thermal buoyancy gain coefficient is a correction parameter used to quantify the enhancement effect of thermal buoyancy on heat transfer in the vertical direction, and this coefficient is calculated based on the temperature difference and the vertical height difference.

[0081] Specifically, the server executes this step during the system initialization phase. First, following step S101, it divides the monitoring sub-regions, constructs logical edges, and initializes the asymmetric transfer coefficients corresponding to each logical edge. Next, the server traverses all adjacent monitoring sub-region pairs, determining their vertical positional relationship by comparing their three-dimensional coordinates. When two sub-regions satisfy a horizontal deviation less than a preset horizontal threshold (e.g., δh = 1.0 meters) and a vertical coordinate difference greater than a preset vertical threshold (e.g., δv = 2.0 meters), they are considered vertically aligned. The server acquires real-time temperature data for the vertically aligned sub-region pairs. When the temperature T of the upper sub-region... u Temperature T is higher than that of the sub-region below. d At that time, according to the formula γ=1+η×(T) u -T d ) / T ref Calculate the vertical thermal buoyancy gain coefficient, where η is the buoyancy sensitivity factor (e.g., a value of 2.0), and T ref The reference temperature is (e.g., 298K). Finally, the server uses the vertical thermal buoyancy gain coefficient to correct the asymmetric transfer coefficient of the logical edge pointing from the lower sub-region to the upper sub-region, with the correction formula being α'. ij =α ij ×γ yields the corrected asymmetric transfer coefficient that reflects the vertical thermal buoyancy effect, and updates the weight values ​​of the corresponding logical edges in the dynamic heat flux topology network.

[0082] S202. The monitoring sub-area is divided into a measured sub-area with sensors and a virtual sub-area without sensors; the initial local state vector of the measured sub-area is determined directly from the environmental monitoring data collected by the sensors.

[0083] Among them, "deploying sensors" refers to the physical installation of environmental parameter detection equipment such as temperature, humidity, and airflow speed in a specific monitoring sub-area; "measured sub-area" refers to a monitoring sub-area where sensor equipment has been installed and can directly collect environmental monitoring data; "no sensors deployed" refers to a state where sensor equipment has not been installed in certain monitoring sub-areas due to cost, technology, or physical limitations; and "virtual sub-area" refers to a monitoring sub-area where sensor equipment has not been installed and environmental parameters need to be inferred through numerical calculation methods.

[0084] Specifically, the server executes this step during the data acquisition phase of the system operation. First, it reads the pre-stored sensor deployment configuration file, which records the sensor installation status identifier for each monitoring sub-area. Based on the sensor deployment cost budget and building space characteristics, the server divides the total number of N monitoring sub-areas into two subsets: the set of measured sub-areas V... real and virtual subregion set V virtual For each subregion i∈V in the set of measured subregions. real The server receives environmental monitoring data from the corresponding sensor nodes in real time via ZigBee or LoRa wireless communication protocols, and performs data validity verification and outlier filtering. The server directly maps the valid environmental monitoring data collected in each measured sub-region into a 4-dimensional initial local state vector, forming the basic data of the real environmental state of the measured sub-region.

[0085] S203. For any virtual sub-region, determine the adjacent measured sub-regions that are logically connected to the virtual sub-region in the dynamic heat flux topology network; calculate the spatial Euclidean distance between the virtual sub-region and each adjacent measured sub-region.

[0086] Among them, the spatial Euclidean distance represents the straight-line distance between the geometric center points of two monitoring sub-regions in the three-dimensional coordinate system.

[0087] Specifically, the server executes this step immediately after completing the data collection of the measured sub-region, and performs this step on the virtual sub-region set V. virtual Each virtual sub-region in the network undergoes data reconstruction. The server first queries the adjacency matrix of the dynamic heat flux topology network for all neighboring nodes directly connected to the virtual sub-region, and then selects those belonging to the measured sub-region set V. real The nodes form an adjacent set of measured subregions N(v) = {r1, r2, ..., r...} mNext, the server obtains the three-dimensional coordinates of the virtual sub-region and the three-dimensional coordinates of each adjacent measured sub-region, and calculates the geometric straight-line distance between the virtual sub-region and each adjacent measured sub-region one by one according to the spatial Euclidean distance calculation formula.

[0088] S204. Based on the environmental monitoring data and spatial Euclidean distance of each adjacent measured sub-region, the simulated environmental monitoring data of the virtual sub-region is obtained by inverse distance weighting calculation, specifically including:

[0089] Obtain the asymmetric transfer coefficients of the logical edges connecting the virtual sub-region and each adjacent measured sub-region;

[0090] The equivalent heat flux distance is obtained by dividing the spatial Euclidean distance using the asymmetric transfer coefficient.

[0091] Using the reciprocal of the equivalent heat flux distance as a weight, the environmental monitoring data of each adjacent measured sub-region are weighted and averaged to obtain the simulated environmental monitoring data of the virtual sub-region.

[0092] Among them, simulated environmental monitoring data represents the environmental parameter values ​​of virtual sub-regions obtained through numerical calculation methods, which are used to replace the data collected by actual sensors; equivalent heat flow distance represents the corrected distance value after considering the physical mechanism of heat transfer. This distance not only reflects the geometric spatial relationship, but also reflects the directionality and intensity characteristics of heat flow transfer.

[0093] Specifically, the server executes this step immediately after obtaining the spatial Euclidean distances between the virtual sub-region and each adjacent measured sub-region, employing a physically enhanced distance-inverse weighted interpolation algorithm to calculate the simulated environmental monitoring data for the virtual sub-region. The server first extracts the asymmetric transfer coefficients from the current dynamic heat flux topology network, which represent the logical edges connecting each adjacent measured sub-region r and the virtual sub-region v. These coefficients reflect the intensity of heat flux transfer from the measured sub-regions to the virtual sub-region. Next, the server calculates the equivalent heat flux distance using the formula... For each adjacent measured sub-region, a division correction is performed, where ε is a small constant to prevent the denominator from being zero (usually taken as 1×10⁻). 5 ),when When D is large eq A decrease indicates a shorter heat flow distance, thus increasing the weight of the measured sub-region in the weighted calculation; conversely, a larger distance indicates a shorter distance. When D is smaller eq An increase indicates a greater heat flow distance, resulting in a decrease in weight. Then, the server calculates the weight coefficient w for each adjacent measured sub-region. r =1 / [D eq [v, r)]², employing an inverse square distance ratio to enhance the influence weight of nearby measured points. Finally, the server performs a weighted average calculation on the environmental monitoring data of all adjacent measured sub-regions to obtain the simulated environmental monitoring data of the virtual sub-region.

[0094] S205, and map the simulated environmental monitoring data into the initial local state vector of the virtual sub-region.

[0095] S206. Based on the airflow velocity vector in the initial local state vector of each monitoring sub-region, calculate the convection transfer intensity between any two adjacent monitoring sub-regions, and determine the diffusion transfer intensity corresponding to the adjacent monitoring sub-regions based on the air temperature difference between any two adjacent monitoring sub-regions.

[0096] S207. Based on the convection transfer intensity and diffusion transfer intensity, calculate and update the asymmetric transfer coefficient of the corresponding logical edge of each monitoring sub-region;

[0097] Steps S206 and S207 and Figure 1 Steps S103 and S104 in the embodiment are described similarly and will not be repeated here.

[0098] S208. Obtain the historical global correction state vector calculated by each monitoring sub-region at the previous moment;

[0099] Among them, the previous moment refers to the previous time sampling point relative to the current processing moment. The time interval is usually the system's preset sampling period, such as 30 seconds, 1 minute or 2 minutes. The historical global correction state vector refers to the comprehensive state vector of each monitoring sub-region obtained after neighborhood feature aggregation and spatiotemporal fusion processing at the previous time sampling point. This vector integrates spatial neighborhood influence and time series features.

[0100] Specifically, the server executes this step immediately after updating the asymmetric transfer coefficients at the current moment, as a preparatory stage for spatiotemporal feature fusion calculation. The server first determines the timestamp of the current moment t, then calculates the timestamp of the previous moment t-1, where the time interval Δt is determined by the system's preset data processing cycle. The server queries the global state vector historical cache database for the record with timestamp t-1, extracting the global corrected state vectors of all N monitoring sub-regions at the previous moment.

[0101] S209. Using the asymmetric transfer coefficient, the initial local state vectors of each monitoring sub-region at the current time are weighted and summed to obtain the spatial aggregation characteristics at the current time.

[0102] The current moment indicates the time stamp of the real-time data sampling point that the system is currently processing.

[0103] Specifically, the server executes this step immediately after obtaining the historical global corrected state vector, performing spatial neighborhood aggregation processing on the initial local state vectors of each monitoring sub-region at the current time. The server first reads the initial local state vectors of each monitoring sub-region at the current time t, as well as the asymmetric transfer coefficient matrix updated in step S207. Next, the server performs neighborhood-weighted aggregation calculations for each monitoring sub-region. The server uses parallel computing to process the spatial aggregation calculations of all monitoring sub-regions simultaneously, utilizing GPU acceleration for matrix operations to improve computational efficiency. Finally, the server obtains the spatial aggregation feature vector at the current time, which integrates the spatial neighborhood influence of each monitoring sub-region based on the real-time heat flow topology.

[0104] S210. Calculate the rate of change of the current airflow velocity vector data relative to the previous moment;

[0105] Specifically, the server executes this step immediately after completing the spatial aggregation feature calculation for the current moment. It assesses the dynamic stability of the building's internal environment by analyzing the temporal variation characteristics of the airflow velocity vector. The server extracts airflow velocity vector data from the initial local state vectors of each monitoring sub-area at the current moment, including velocity components in three spatial directions, as well as the overall magnitude and direction information of the vector. Simultaneously, the server retrieves airflow velocity vector data for the same monitoring sub-area from the historical data cache at the previous moment. The server calculates the magnitude of the airflow velocity vector change for each monitoring sub-area, quantifying the degree of velocity intensity change by comparing the difference in vector magnitude between the current and previous moments; and assesses the degree of airflow direction deflection by analyzing the change in vector direction angle. The server comprehensively considers both the magnitude and direction changes to calculate the overall rate of change index for each sub-area, and then performs statistical analysis on the rate of change for all monitoring sub-areas across the entire site, including calculating characteristic values ​​such as the average rate of change, the maximum rate of change, and the standard deviation of the rate of change.

[0106] S211. Determine the time smoothing coefficient based on the rate of change, and use the time smoothing coefficient to perform weighted fusion of the historical global corrected state vector and spatial aggregation features to obtain the global corrected state vector at the current moment.

[0107] The time smoothing coefficient represents a weighting parameter used to control the fusion ratio of historical state information and current spatial aggregation features, and its value is usually between 0 and 1.

[0108] Specifically, the server executes this step immediately after calculating the airflow velocity change rate, achieving adaptive spatiotemporal feature fusion to generate the current global corrected state vector. Based on the comprehensive change rate index obtained in the previous step, the server dynamically determines the value of the time smoothing coefficient through a preset nonlinear mapping function. The design principle of this mapping function is: when the change rate is large, the time smoothing coefficient tends to a larger value, indicating that the system relies more on the spatial aggregation features of the current moment, reducing the influence of historical state information to quickly respond to dynamic environmental changes; when the change rate is small, the time smoothing coefficient tends to a smaller value, indicating that the system retains more historical state information, eliminating sensor noise and instantaneous fluctuations through the continuity of the time series. The server sets a minimum value for the time smoothing coefficient to ensure a certain degree of state update capability even in a completely static environment, avoiding an overly sluggish system response. After obtaining the time smoothing coefficient, the server performs spatiotemporal weighted fusion calculations for each monitoring sub-region, linearly combining the current spatial aggregation features with the historical global corrected state vector from the previous moment according to the calculated weight ratio. Finally, the server generates a global corrected state vector that integrates spatial neighborhood relationships and time series features.

[0109] S212. Input the global correction state vector into the preset load mapping model, calculate the real-time regional load value of each monitoring sub-area through nonlinear mapping, and accumulate the real-time regional load values ​​of all monitoring sub-areas to obtain the overall operating load status of the target building space.

[0110] This step and Figure 1 The description of step S106 in the embodiment is similar and will not be repeated here.

[0111] S213. Calculate the thermal distribution variance of each monitoring sub-region in the global corrected state vector;

[0112] Among them, the thermal distribution variance represents the degree of dispersion of the temperature state value of each monitoring sub-area relative to the average temperature state, and is used to measure the non-uniformity of the thermal environment distribution within the entire target building space.

[0113] Specifically, the server executes this step immediately after completing the calculation of the global corrected state vector at the current moment, which is used to assess the uniformity of the thermal environment distribution inside the building. The server first extracts temperature component data from the global corrected state vectors of all monitoring sub-regions, forming a temperature distribution vector T_global=[T1', T2', ..., T...]. n '], where each element T i'Represents the corrected temperature value of monitoring sub-region i after neighborhood feature aggregation and spatiotemporal fusion. Next, the server calculates the average temperature state value of the entire building space, obtaining the global temperature mean T_mean by arithmetically averaging the corrected temperature values ​​of all monitoring sub-regions. Then, the server calculates the temperature deviation value for each monitoring sub-region, i.e., the difference between the corrected temperature value of each sub-region and the global mean, and squares the deviation value to eliminate the influence of positive and negative signs. The server sums all the squared deviation values ​​and divides them by the total number of monitoring sub-regions to obtain the thermal distribution variance value.

[0114] S214. When the overall operating load status change exceeds the preset change threshold and the heat distribution variance is greater than the preset uniformity threshold, the monitoring sub-area with the largest real-time regional load value is selected, and an air volume adjustment command is generated for the corresponding monitoring sub-area.

[0115] Among them, the change amplitude represents the degree of change of the overall operating load state relative to the previous moment or the baseline value, usually expressed as the absolute value difference or the relative change rate; the preset change threshold is a critical value used to judge whether the load state has changed significantly, with a typical value of 5%-15% of the total load. Its function is to avoid the system over-responding to small load fluctuations, and the setting scheme is usually determined based on the statistical analysis of historical operating data; the preset uniformity threshold is a critical variance value used to judge whether the heat distribution is uniform. The setting of this threshold is based on comfort requirements and building characteristics, with a typical value of 2.0-5.0 square degrees Celsius. Its function is to ensure that local adjustment is only initiated when the heat distribution is severely uneven; the maximum real-time regional load value refers to the load value corresponding to the sub-area with the highest load demand among all monitored sub-areas at the current moment; the air volume adjustment command represents the control signal sent by the server to the air conditioning and ventilation equipment to guide the increase or decrease of the air supply volume in a specific area.

[0116] Specifically, the server executes this step immediately after calculating the heat distribution variance, enabling intelligent adjustment decisions based on load state changes and heat distribution uniformity. The server first obtains the current overall operating load state value and compares it with historical baseline load values ​​or the load value from the previous moment, calculating the absolute change amplitude or relative change rate. The server compares the calculated change amplitude with a preset change threshold to determine if the critical condition for adjustment has been met. Simultaneously, the server compares the heat distribution variance calculated in step S213 with a preset uniformity threshold to assess the uniformity of the current thermal environment distribution. When the server detects that both conditions are met simultaneously—that is, the load change amplitude exceeds the preset change threshold and the heat distribution variance is greater than the preset uniformity threshold—it indicates a significant load change within the building accompanied by uneven heat distribution, at which point a local precise adjustment mechanism is activated. The server iterates through the real-time regional load values ​​of all monitoring sub-areas, identifying the monitoring sub-area with the maximum load demand and its corresponding load value and spatial location information through numerical comparison operations. Next, based on the location and current load value of the maximum load sub-zone, and considering the building's ventilation system topology, the server calculates the required airflow adjustment, including the adjustment direction (increase or decrease) and adjustment magnitude (specific airflow value). Finally, the server generates an airflow adjustment command containing the target monitoring sub-zone identifier, the adjusted airflow value, and the execution priority, and sends it to the corresponding damper controller or frequency converter via a communication interface to achieve precise airflow adjustment in key areas.

[0117] In this embodiment, by using the vertical thermal buoyancy gain coefficient to correct the asymmetric transfer coefficient, calculating the simulated environmental monitoring data of the virtual sub-region through the equivalent heat flow distance, and using the time smoothing coefficient to weight and fuse the historical global corrected state vector and spatial aggregation features, not only is the vertical thermal stratification effect accurately quantified and full-space coverage monitoring achieved, but the real-time performance and stability of load perception are also taken into account. This effectively solves the problem of perception blind spots and fluctuation interference caused by the complex and changeable thermal environment of large spaces, thereby achieving high-precision overall operating load status perception and precise airflow adjustment command generation.

[0118] The server in the embodiments of this invention is described below from the perspective of hardware processing. Please refer to [link / reference]. Figure 3 This is a schematic diagram of the physical device structure of a server for an air conditioning and ventilation control system in this application embodiment.

[0119] It should be noted that, Figure 3 The server structure shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0120] like Figure 3As shown, the server includes a CPU 301, which can perform various appropriate actions and processes based on a program stored in the read-only memory ROM 302 or a program loaded from the storage section 308 into the random access memory RAM 303, such as performing the methods described in the above embodiments. The RAM 303 also stores various programs and data required for system operation. The CPU 301, ROM 302, and RAM 303 are interconnected via a bus 304. An I / O interface 305 is also connected to the bus 304.

[0121] The following components are connected to I / O interface 305: input section 306 including audio input devices, push-button switches, etc.; output section 307 including a liquid crystal display (LCD) and audio output devices, indicator lights, etc.; storage section 308 including a hard disk, etc.; and communication section 309 including a network interface card such as a LAN (Local Area Network) card, modem, etc. Communication section 309 performs communication processing via a network such as the Internet. Drive 310 is also connected to I / O interface 305 as needed. Removable media 311, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 310 as needed so that computer programs read from them can be installed into storage section 308 as needed.

[0122] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program / instructions carried on a computer-readable medium, the computer program / instructions containing computer program / instructions for performing the methods shown in the flowcharts. In such embodiments, the computer program / instructions can be downloaded and installed from a network via communication section 309, and / or installed from removable medium 311. When the computer program is executed by CPU 301, it performs the various functions defined in the present invention.

[0123] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0124] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Each block in a flowchart or block diagram may represent a module, program segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those shown in the drawings.

[0125] Specifically, the server in this embodiment includes a processor and a memory. The memory stores a computer program. When the computer program is executed by the processor, it implements the intelligent sensing method for the operating load status of air conditioning and ventilation provided in the above embodiment.

[0126] In another aspect, the present invention also provides a computer-readable storage medium, which may be included in the server described in the above embodiments; or it may exist independently and not assembled into the server. The storage medium carries one or more computer programs that, when executed by a processor of the server, cause the server to implement the intelligent sensing method for air conditioning ventilation operation load status provided in the above embodiments.

[0127] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0128] As used in the above embodiments, depending on the context, the term "when..." can be interpreted as meaning "if...", "after...", "in response to determining...", or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if (the stated condition or event) is interpreted as meaning "if determining...", "in response to determining...", "when (the stated condition or event) is detected", or "in response to detecting (the stated condition or event)".

[0129] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A method for intelligent sensing of air conditioning and ventilation operation load status, applied to a server in an air conditioning and ventilation control system, characterized in that, include: The target building space is discretized into several monitoring sub-areas, and a basic structure of a dynamic heat flow topology network connecting each monitoring sub-area is established. The dynamic heat flow topology network includes logical edges connecting any two adjacent monitoring sub-areas, and each logical edge is assigned an asymmetric transfer coefficient. Real-time acquisition of environmental monitoring data for each monitoring sub-region, and mapping of the environmental monitoring data to the initial local state vector of the corresponding monitoring sub-region, wherein the environmental monitoring data includes at least air temperature, relative humidity and airflow velocity vector; Based on the airflow velocity vector in the initial local state vector of each monitoring sub-region, the convective transfer intensity between any two adjacent monitoring sub-regions is calculated, and the diffusion transfer intensity corresponding to the adjacent monitoring sub-regions is determined based on the air temperature difference between any two adjacent monitoring sub-regions. Based on the convection transfer intensity and the diffusion transfer intensity, the asymmetric transfer coefficient of the logical edge corresponding to each monitoring sub-region is calculated and updated; Based on the updated dynamic heat flux topology network, the initial local state vector of each monitoring sub-region is aggregated using the asymmetric transfer coefficient to obtain a global corrected state vector. The global correction state vector is input into a preset load mapping model, and the real-time regional load value of each monitoring sub-area is obtained through nonlinear mapping calculation. The real-time regional load values ​​of all monitoring sub-areas are accumulated to obtain the overall operating load status of the target building space.

2. The method according to claim 1, characterized in that, Assigning an asymmetric transitivity coefficient to each logical edge specifically includes: Determine the vertical positional relationship between two adjacent monitoring sub-areas; If two adjacent monitoring sub-areas are arranged vertically, and the air temperature data of the monitoring sub-area above is higher than that of the monitoring sub-area below, then a vertical thermal buoyancy gain coefficient is generated. The asymmetric transfer coefficient of the logical edge from the lower monitoring sub-region to the adjacent upper monitoring sub-region is corrected using the vertical thermal buoyancy gain coefficient to obtain the corrected asymmetric transfer coefficient.

3. The method according to claim 2, characterized in that, The real-time acquisition of environmental monitoring data for each monitoring sub-region and mapping the environmental monitoring data to the initial local state vector of the corresponding monitoring sub-region specifically includes: The monitoring sub-region is divided into a measured sub-region with sensors deployed and a virtual sub-region without sensors deployed; the initial local state vector of the measured sub-region is directly determined by the environmental monitoring data collected by the sensors. For any of the virtual sub-regions, determine the adjacent measured sub-regions that are logically connected to the virtual sub-region in the dynamic heat flux topology network; Calculate the spatial Euclidean distance between the virtual sub-region and each of the adjacent measured sub-regions; Based on the environmental monitoring data of each of the adjacent measured sub-regions and the spatial Euclidean distance, the simulated environmental monitoring data of the virtual sub-region is obtained by inverse distance weighting, and the simulated environmental monitoring data is mapped to the initial local state vector of the virtual sub-region.

4. The method according to claim 3, characterized in that, The simulated environmental monitoring data of the virtual sub-region is obtained by calculating the environmental monitoring data of each of the adjacent measured sub-regions and the spatial Euclidean distance through inverse distance weighting, specifically including: Obtain the asymmetric transfer coefficients of the logical edges connecting the virtual sub-region and each of the adjacent measured sub-regions; The equivalent heat flux distance is obtained by dividing the spatial Euclidean distance using the asymmetric transfer coefficient. Using the reciprocal of the equivalent heat flux distance as a weight, the environmental monitoring data of each of the adjacent measured sub-regions are weighted and averaged to obtain the simulated environmental monitoring data of the virtual sub-region.

5. The method according to claim 1, characterized in that, The step of calculating the convective transfer intensity between any two adjacent monitoring sub-regions based on the airflow velocity vector in the initial local state vector of each monitoring sub-region specifically includes: Based on two adjacent monitoring sub-regions, calculate the projection component of the airflow velocity vector of the upstream sub-region onto the vector connecting the upstream sub-region to the downstream sub-region; If the projection component is positive, the convective transmission intensity is determined to be a first preset intensity value; If the projection component is negative, the convective transmission intensity is determined to be a second preset intensity value, where the first preset intensity value is greater than the second preset intensity value.

6. The method according to claim 4 or 5, characterized in that, The updated dynamic heat flux topology network, based on the asymmetric transfer coefficient, performs neighborhood feature aggregation on the initial local state vector of each monitoring sub-region to obtain a globally corrected state vector, specifically including: Obtain the historical global correction state vector calculated for each of the monitoring sub-regions at the previous time step; The initial local state vectors of each monitoring sub-region at the current time are weighted and summed using the asymmetric transfer coefficients to obtain the spatial aggregation features at the current time. Calculate the rate of change of the airflow velocity vector data at the current moment relative to the previous moment; The time smoothing coefficient is determined based on the rate of change, and the historical global corrected state vector and the spatial aggregation feature are weighted and fused using the time smoothing coefficient to obtain the global corrected state vector at the current moment.

7. The method according to claim 6, characterized in that, After obtaining the overall operational load status of the target building space, the method further includes: Calculate the variance of the heat distribution in each monitoring sub-region of the global corrected state vector; When the overall operating load status changes beyond a preset change threshold and the heat distribution variance is greater than a preset uniformity threshold, the monitoring sub-area with the largest real-time regional load value is selected, and an airflow adjustment command is generated for the corresponding monitoring sub-area.

8. A server for an air conditioning and ventilation control system, characterized in that, The server includes: One or more processors and a memory; the memory is coupled to the one or more processors, the memory being used to store computer program code, the computer program code including computer instructions, the one or more processors invoking the computer instructions to cause the server to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium storing computer instructions, characterized in that, When the computer instructions are executed on the server, the server causes the server to perform the method as described in any one of claims 1-7.

10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are run on the server, the server causes the server to perform the method as described in any one of claims 1-7.

Images