Method, device and equipment for predicting ozone concentration on outer facade of high-rise building and storage medium

Abstract

The application provides a method, device and equipment for predicting the ozone concentration of the outer facade of a high-rise building and a storage medium. It relates to the technical field of building environment simulation and air pollution prediction. The method first analyzes the near-wall area of the outer facade of a high-rise building using CFD technology, outputs key flow field parameters and wall ozone deposition flux gradient, and then combines the CFD simulation results with PINN to realize seamless transfer of physical quantities in the overlapping area through a dynamic coupling interface. PINN processes the free flow area, strengthens NO x reaction terms, intermediate layer embedded photolysis simplified eddy diffusion, high-rise building uses only first-order diffusion, outputs modified velocity field, ozone concentration and reaction rate. Finally, the system is trained jointly to optimize the model with a multi-objective loss function. This method has the advantages of small CFD wall flux error and efficient PINN calculation, solves the contradiction between efficiency and accuracy in prediction, supports real-time simulation, reduces data requirements, and enhances generalization.

Description

Technical Field

[0001] This application relates to the field of building environment simulation and air pollution prediction technology, and in particular to a method, device, equipment and storage medium for predicting ozone concentration on the facade of a high-rise building. Background Technology

[0002] Ozone pollution, characterized by low exposure concentrations and high health risks, has become a crucial component of air quality assessment. Outdoor ozone permeates indoors through building channels such as window gaps, elevator shaft chimney effects, and fresh air systems, becoming a major source of indoor ozone. The "Unified Standard for Civil Building Design" (GB 50352-2019) defines residential buildings taller than 24 meters as high-rise buildings. Currently, residential buildings in third- and fourth-tier cities in my country are mainly under 18 stories, approximately 8 to 18 stories, with building heights ranging from 24 to 54 meters. Buildings between 54 and 100 meters (approximately 18 to 33 stories) are the dominant type in first- and second-tier cities, accounting for over 70% of all high-rise residential buildings (according to the Ministry of Housing and Urban-Rural Development's 2022 assessment report). Buildings between 100 and 150 meters (approximately 33 to 50 stories) are common in the core areas of first- and second-tier cities, accounting for 5%. Buildings taller than 150 meters are rare, mostly being landmark luxury residences or projects on special plots. Actual measurements of ozone vertical concentration at a high tower in Beijing show that: the ground level (0-50m) is directly affected by traffic emissions (NOx+VOCs), with abundant ozone precursors, but ozone concentration is suppressed by the titration effect; the daily average concentration in this height range is 48±12 ppb, with a daily range of 35 ppb; the middle level (50-150m) is a photochemically active zone with the highest ozone generation (typical peak of 120-180 ppb), with a daily average concentration of 82±18 ppb and a daily range of 58 ppb; the upper level (>150m) is dominated by background ozone, with a daily average concentration of 110±15 ppb and a daily range of 42 ppb.

[0003] Near-surface ozone is mainly generated by NOx and VOCs under ultraviolet light. The intensity of ultraviolet radiation increases by 1.5%-2% for every 100m increase in altitude, leading to a rise in the net ozone formation rate with altitude. In atmospheric chemistry or urban microclimate studies, the concentration increment exp(0.0015H) is a simplified calculation used to describe the vertical distribution of ozone in the lower atmosphere (<500m), assuming that turbulent mixing weakens with altitude, leading to pollutant accumulation, and neglecting the time-varying effects of horizontal transport and chemical reactions. The change in ozone concentration with altitude is within H < 200m, and this exponential relationship can be simplified to a linear relationship of 1 + 0.0015H. Appendix D of the "Healthy Building Evaluation Standard" (T / ASC 02-2021) also uses the above method of linear correction for the outdoor ozone concentration at the ground level when calculating the ozone infiltration of high-rise residential buildings with a height > 54m, describing the outdoor ozone concentration at height H, with a concentration increment of 0.15% / m per unit height. 0.0015 is achieved by balancing the photochemical generation rate (Q≈2ppb / h) with the turbulent diffusion effect (Kz≈50m).2 The empirical parameters derived from / s are consistent with the ozone vertical gradient (0.1%–0.2% / m) observed in Beijing during the summer.

[0004] The actual variation of ozone concentration with altitude is complex, and the differences can vary significantly with time, weather conditions, geographical location, and the characteristics of the building itself. Multiple factors affect the vertical distribution of ozone concentration along the height of a building: ① Ozone sources: The ground area near busy urban areas with high industrial emissions is an ozone production zone. At the same time, ozone near the ground is easily consumed by building surfaces, vegetation, soil, and some chemical reactions; ozone at higher floors is first affected by the vertical upward diffusion of ozone concentration near the ground, and secondly, stratospheric ozone is sometimes transported downwards, affecting the concentration in the upper troposphere; ozone generated in cities is carried downwind by wind and mixed in higher layers. ② Atmospheric Boundary Layer and Mixing: During the day, sunlight heats the Earth's surface, generating updrafts. The atmospheric boundary layer develops vigorously, resulting in strong vertical mixing. At this time, ozone and its precursors are thoroughly mixed, and the concentration is relatively uniform from the ground to the upper atmosphere with a small vertical gradient. Especially in areas with many tall buildings, enhanced turbulence and mixing also lead to small vertical concentration differences. At night and from dawn, due to the depletion of ground-based ozone and the titration of nitrogen oxides, the ozone concentration near the ground usually drops to a very low level. The inversion layer formed by the cooling of the ground inhibits vertical mixing. Above the inversion layer is a layer of air that was thoroughly mixed the previous day (residual layer). Because ozone does not come into contact with the ground-based depletion sources and the reaction of precursors is weakened, the concentration is significantly higher than the ground concentration. ③ Urban Canopy Effect: Dense buildings alter the local wind field and turbulence structure. Poor ventilation may lead to the accumulation of precursors and ozone, resulting in a complex concentration distribution.

[0005] In summary, ozone concentrations on higher floors (at or above the bottom of the residual ozone layer) are typically significantly higher than on lower floors and at ground level. This difference is particularly pronounced on clear nights. In winter, with lower ozone background levels and more frequent and deeper temperature inversions, higher ozone concentrations on higher floors are more common at night. Residents on higher floors may be exposed to higher ozone concentrations at night and in the early morning than reported by ground-based monitoring stations (which are typically installed 2-10 meters above the ground). Therefore, when assessing the health risks to residents of high-rise buildings, one cannot rely solely on data from ground-based monitoring stations, especially during nighttime hours.

[0006] Understanding the vertical distribution of ozone is crucial for accurately assessing air pollutant exposure levels for residents on different floors. However, continuous, multi-point ozone concentration monitoring on the facades of high-rise buildings—such as installing fixed monitoring stations at different heights or portable devices on suspended platforms or window cleaning machines—is costly and technically challenging. Using drones with sensors for vertical profile measurements is limited by regulations and flight time constraints, while vertical gradient monitoring using high towers lacks specificity due to location limitations. Therefore, simulation and prediction methods have become the most feasible alternative. They can assess the vertical distribution patterns of ozone on the facades of high-rise buildings with low cost and high spatiotemporal accuracy, overcoming the limitations of physical monitoring methods and providing a scientific basis for refined health risk management and optimized building environment design.

[0007] PINN is a method that embeds physical laws into deep learning models, particularly suitable for scenarios with sparse data but well-defined physical mechanisms. Currently, PINN is mostly applied to ozone concentration field simulation in near-surface / urban canopy areas. Regional transport studies include: Peking University using PINN with a three-dimensional wind-driven advection equation as physical constraints to quantify the inter-city transport flux of ozone in the Beijing-Tianjin-Hebei region; Nanjing University using PINN combined with chemical generation terms and advection diffusion equations to distinguish between local emissions and regional transport contributions, finding that regional transport contributed approximately 25-35% in summer events in the Yangtze River Delta; and Fudan University using a PINN + chemical transport model (CTM) coupling method to perform attribution inversion on extreme ozone pollution events in the Yangtze River Delta in 2022, concluding that 45-55% was due to local photochemical contributions, 20-25% to horizontal regional transport, and 30-38% to upper-level sinking transport (3-8 km).

[0008] Traditional data-driven models such as LSTM are used to predict the vertical distribution of ozone, but they suffer from poor generalization ability in sparse regions of data at altitudes of 5 km and above, and cannot guarantee that the prediction results conform to atmospheric physicochemical laws (such as photochemical reaction equilibrium and vertical diffusion constraints). Patent CN113920489B proposes using PINN to predict ozone at an altitude of 5 km, reducing the error by 42% compared to the LSTM method. Tsinghua University compensated for the lack of vertical observation data by using physical constraints (advection diffusion equation + photochemical relaxation term) and PINN to fuse sparse ozone radiosonde data, reducing the root mean square error of ozone prediction at altitudes of 3-6 km by about 40% compared to the pure data-driven model LSTM. Nanjing University of Information Science and Technology used PINN to input radiosonde data from the Yangtze River Delta stations and obtained an ozone prediction error of about 9% at 4-7 km. The Institute of Atmospheric Physics, Chinese Academy of Sciences, developed the OzonePINN-3D model based on DeepXDE, which predicted the vertical distribution of ozone in China from 0 to 15 km.

[0009] Computational fluid dynamics (CFD) simulations can provide more accurate predictions than empirical formulas.

[0010] For example, Chinese patent application CN114265854A embeds an ozone generation weighting function into the Realizable k-ε model, densifies the mesh in high-concentration gradient regions, and replaces the traditional ozone prediction model WRF-Chem, improving accuracy at the urban microscale and analyzing turbulence and vertical gradient changes caused by building clusters. Chinese patent application CN113887746A, to assess the impact of newly constructed high-rise buildings on street ozone diffusion, addresses the insufficient accuracy of the traditional ozone diffusion model AERMOD in urban building clusters by employing a modified k-ωSST model to solve the turbulent adsorption of ozone on building facades. It couples a simplified photochemical mechanism (CB05-S) to dynamically calculate the NOx-VOCs-O3 reaction chain, outputting spatial distribution data of ozone concentration along the building facade or the vertical direction of the street. The aforementioned RANS models are suitable for steady-state ozone distribution calculations, mostly simulating long-term accumulation of concentration near the ground or building canopy height, but have limited accuracy in simulating transient turbulence.

[0011] Chinese patent application CN110647721A, based on Large Eddy Simulation (LES), performs real-time calculations of pollutant diffusion in building complexes, representing a high-precision CFD simulation for pollutant diffusion in urban microenvironments. Chinese patent application CN114265854A uses the EBI algorithm to simplify the calculation of ozone generation / depletion rates and employs LES to simulate the flow field at the microscale of building complexes / urban blocks. While LES can capture instantaneous turbulent structures, it requires significant computation and sophisticated hardware.

[0012] However, neither PINN nor CFD simulation techniques have been used in the prediction of vertical ozone concentration distribution along the height of individual buildings.

[0013] The application of CFD in building ozone simulation suffers from problems such as oversimplification of models and low computational efficiency. Using CFD to construct the flow field on the building facade to simulate the vertical concentration distribution of ozone requires analytical turbulent structures with a grid size of 0.1-0.5m, resulting in high computational costs.

[0014] PINN embeds physical laws as constraints into neural network training, directly mapping spatial coordinates (x,y,z,t) to the pollutant concentration field through a coordinate network MLP, avoiding the tens of millions of grid divisions required by CFD, thus improving computational speed and prediction speed. However, PINN struggles to accurately capture high gradient changes at the micrometer scale in viscous subsurfaces, resulting in lower prediction accuracy than CFD in high gradient regions near the wall. Furthermore, the coupling loss function between the fluid equations (NS) and the RACM2 chemical mechanism has a competitive optimization conflict, leading to turbulence-chemical interactions (such as NO)... x -O nonlinear feedback) error amplification. Summary of the Invention

[0015] This application provides a method, device, equipment, and storage medium for predicting ozone concentration on the facade of high-rise buildings. Addressing the problems of low computational efficiency and difficulty in coupling chemical mechanisms in CFD models, as well as insufficient prediction accuracy and incomplete turbulence-chemical interaction modeling in the PINN model in the high gradient region near the wall, a PINN-CFD fusion architecture is constructed to collaboratively simulate and calculate the vertical ozone concentration on the facade of high-rise buildings in a partitioned manner. The CFD module accurately analyzes the turbulent structure in the near-wall region, providing reliable wall flux and turbulence parameters. The PINN module employs a hierarchical modeling strategy to handle the free flow region, enhancing NOx reaction terms in the near-surface layer, embedding photolysis simplification and eddy diffusion in the intermediate layer, and using only first-order diffusion in the high-rise section. The final output includes a corrected velocity field, ozone concentration, and reaction rate. A dynamic coupling interface algorithm enables seamless transfer of physical quantities between CFD and PINN in overlapping regions. This fusion architecture combines the small wall flux error of CFD with the computational efficiency of PINN, resolving the efficiency-accuracy contradiction in the dynamic prediction of ozone concentration distribution on the facade of high-rise buildings, and providing core technical support for green building operation and maintenance and smart city management.

[0016] Firstly, this application provides a method for predicting ozone concentration on the exterior facade of a high-rise building, comprising:

[0017] A CFD module is constructed to perform high-resolution simulation of the near-wall area of ​​the facade of a high-rise building. Boundary layer mesh is used to solve the ozone transport equation and output key flow field parameters and ozone concentration field. At the same time, the wall ozone deposition flux is calculated to provide boundary conditions for the PINN module.

[0018] A PINN module is constructed, and a hierarchical modeling strategy is used to process the free flow region through the PINN module, wherein the near-surface layer enhances NO. x The reaction term uses photolysis simplification and eddy diffusion in the middle layer and only first-order diffusion in the upper layer, outputting a corrected velocity field, ozone concentration and reaction rate.

[0019] The physical quantity transfer between the CFD module and the PINN module is realized through a dynamic coupling interface using a residual adaptive weighting algorithm. The CFD module transmits the flow field parameters of the overlapping area to the PINN module, and the PINN module feeds back the concentration gradient to the CFD module as the top boundary condition.

[0020] A joint training system is constructed to train an ozone concentration prediction model. The ozone concentration prediction model includes a CFD module, a PINN module, and a dynamic coupling interface. During the training process, RMSE and error analysis are calculated. If the upper layer error exceeds a set threshold, the number of PINN intermediate layer nodes is increased or an adaptive weighted loss is introduced to optimize the ozone concentration prediction model until the preset accuracy requirements are met.

[0021] In one possible design, the formula for calculating the wall ozone deposition flux is:

[0022]

[0023] In the formula, J dep Where D is the wall ozone deposition flux, and D is the diffusion coefficient. The gradient of ozone concentration along the wall normal is represented by wall, where wall represents the wall surface.

[0024] In one possible design, the near-surface layer employs 2nd-order chemical PINN, whose physical constraints satisfy the following reaction-diffusion equation:

[0025]

[0026] In the formula, Let t be the ozone concentration, t be the time, and u be the flow velocity vector. Here, D is the gradient operator, and R is the diffusion coefficient. chem This represents the rate of a chemical reaction.

[0027] In one possible design, the intermediate layer constructs a photo-simplified + eddy-diffused PINN, whose physical constraints satisfy the equation:

[0028]

[0029] In the formula, v t The eddy diffusion coefficient is... Let be the gradient operator, D be the diffusion coefficient, C be the ozone concentration, and J(x) be the photolysis rate. denoted as nitrogen dioxide concentration, and u as the flow velocity vector.

[0030] In one possible design, the dynamic coupling interface calculates the relative error ω of the overlap region concentration during data transmission and controls ω < 5%; wherein the formula for calculating the relative error of the overlap region concentration is:

[0031]

[0032] In the formula, C PINN For the concentration simulated by the PINN module, C CFD The concentration is from the CFD simulation.

[0033] In one possible design, when the upper layer adopts a first-order diffusion PINN, only the advection-diffusion process is considered, complex chemical reactions are ignored, and the transport law of ozone in the upper layer free flow region is described by a first-order diffusion equation.

[0034] In one possible design, the input parameters of the joint training system include spatial coordinates, timestamps, meteorological parameters, building geometric features, and pollution source data; the output parameters include ozone concentration field, flow velocity field, reaction rate, and gradient characteristics. The meteorological parameters include wind speed, solar radiation intensity, temperature, relative humidity, and background ozone concentration; the building geometric features include wall curvature, building orientation, and surface roughness; and the pollution source data includes NO... x Emission intensity and near-ground ozone precursor concentration.

[0035] Secondly, this application provides a device for predicting ozone concentration on the exterior facade of a high-rise building, the device comprising:

[0036] The CFD module construction unit is configured to construct a CFD module, which performs high-resolution simulation of the near-wall area of ​​the facade of a high-rise building. The CFD module uses boundary layer mesh to solve and couple the ozone transport equation, outputting key flow field parameters and ozone concentration field. At the same time, it calculates the wall ozone deposition flux and provides boundary conditions for the PINN module.

[0037] The PINN module building unit is configured to build PINN modules, which employ a hierarchical modeling strategy to process the free flow region, wherein the near-surface layer enhances NO. x The reaction term uses photolysis simplification and eddy diffusion in the middle layer and only first-order diffusion in the upper layer, outputting a corrected velocity field, ozone concentration and reaction rate.

[0038] The dynamic coupling unit is configured to transfer physical quantities between the CFD module and the PINN module through a residual adaptive weighting algorithm via a dynamic coupling interface, wherein the CFD module transmits the flow field parameters of the overlapping region to the PINN module, and the PINN module feeds back the concentration gradient to the CFD module as the top boundary condition.

[0039] The joint training unit is configured to build a joint training system to train an ozone concentration prediction model. The ozone concentration prediction model includes a CFD module, a PINN module, and a dynamic coupling interface. During the training process, RMSE and error analysis are calculated. If the upper layer error exceeds a set threshold, the number of PINN intermediate layer nodes is increased or an adaptive weighted loss is introduced to optimize the ozone concentration prediction model until the preset accuracy requirements are met.

[0040] Thirdly, embodiments of this application provide an electronic device, including: at least one processor and a memory; the memory stores computer execution instructions; the at least one processor executes the computer execution instructions stored in the memory, causing the at least one processor to perform the method for predicting ozone concentration on the facade of a high-rise building as described in the first aspect and various possible designs of the first aspect.

[0041] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions. When a processor executes the computer-executable instructions, it implements the method for predicting ozone concentration on the facade of a high-rise building as described in the first aspect and various possible designs of the first aspect.

[0042] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the method for predicting ozone concentration on the facade of a high-rise building as described in the first aspect and various possible designs of the first aspect.

[0043] The method, apparatus, equipment, and storage medium for predicting ozone concentration on the exterior facade of high-rise buildings provided in this application have at least the following beneficial effects:

[0044] 1) Improved computational efficiency: Compared to traditional CFD, prediction speed is 10 times faster (under the same hardware), and real-time dynamic simulation is supported;

[0045] 2) Reduced data requirements: Only 5% of key point monitoring data (such as rooftops and central facades) are needed to achieve >90% prediction accuracy;

[0046] 3) Enhanced generalization: adaptable to different building forms (tower / slab), meteorological conditions, and urban background pollution scenarios. Attached Figure Description

[0047] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0048] Figure 1 A schematic diagram illustrating the implementation principle of a method for predicting ozone concentration on the exterior facade of a high-rise building, provided in this application embodiment.

[0049] Figure 2 A flowchart illustrating a method for predicting ozone concentration on the exterior facade of a high-rise building, provided in an embodiment of this application.

[0050] Figure 3 A flowchart of high-precision CFD boundary layer analysis provided for embodiments of this application;

[0051] Figure 4 A flowchart for modeling the PINN step reaction provided in this application embodiment;

[0052] Figure 5 A flowchart of dynamic coupling training provided for embodiments of this application;

[0053] Figure 6A structural diagram of the ozone concentration prediction device for the exterior facade of a high-rise building provided in this application embodiment.

[0054] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0055] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0056] The collection, storage, use, processing, transmission, provision, and disclosure of financial data or user data involved in the technical solution of this application all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0057] It should be noted that in the embodiments of this application, certain software, components, models and other existing solutions in the industry may be mentioned. These should be regarded as exemplary and are only intended to illustrate the feasibility of implementing the technical solution of this application. However, it does not mean that the applicant has used or necessarily used the solution.

[0058] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0059] This application provides a method for predicting ozone concentration on the exterior facade of a high-rise building, such as... Figure 1The diagram illustrates the implementation principle of a method for predicting ozone concentration on the facade of a high-rise building, as provided in this application. The method first uses CFD technology to perform high-precision analysis of the near-wall region (boundary layer) of the high-rise building facade, outputting key flow field parameters and the wall ozone deposition flux gradient. Subsequently, the CFD simulation results are combined with a Physical Information Neural Network (PINN), achieving seamless transfer of physical quantities in overlapping regions through a dynamic coupling interface algorithm. The PINN module processes the free flow region, employing a hierarchical modeling strategy, dividing it into near-surface, intermediate, and high-rise layers. Specifically, the near-surface layer strengthens the NOx reaction term, the intermediate layer embeds photolysis simplification and eddy diffusion, and the high-rise layer only uses first-order diffusion, ultimately outputting the corrected velocity field, ozone concentration, and reaction rate. The model is optimized using a multi-objective loss function based on physical residuals, data matching, and mass conservation through a joint training system. This method combines the advantages of small wall flux error in CFD and the computational efficiency of PINN, effectively resolving the efficiency-accuracy contradiction in the dynamic prediction of ozone concentration distribution on the facade of high-rise buildings. It supports real-time dynamic simulation, significantly reduces data requirements, and enhances generalization.

[0060] In some embodiments, a PINN-CFD fusion architecture is established. This PINN-CFD fusion architecture is used to realize a method for predicting ozone concentration on the facade of high-rise buildings. It achieves efficient prediction of ozone infiltration on the facade of buildings through multi-module collaboration. Its core components include: (1) The CFD module is responsible for high-resolution simulation of the near-wall area. It uses boundary layer mesh to solve and couple the ozone transport equation, outputs key flow field parameters and ozone concentration field, and calculates the wall ozone deposition flux to provide boundary conditions for the subsequent PINN module; (2) The PINN module uses a hierarchical modeling strategy to process In the free flow region, the NOx reaction term is enhanced in the near-surface layer, the intermediate layer is embedded with photolysis simplification + eddy diffusion, and only the first-order diffusion is used in the upper layer. The final output is the corrected velocity field, ozone concentration and reaction rate; (3) The dynamic coupling interface uses the residual adaptive weighting algorithm to control the data exchange between CFD and PINN (CFD→PINN transmits the flow field parameters of the overlapping area, and PINN→CFD feeds back the concentration gradient as the top boundary condition); (4) The joint training system calculates RMSE and performs error analysis. If the error in the upper layer is significant, the number of nodes in the intermediate layer of PINN is increased or an adaptive weighted loss is introduced.

[0061] Specifically, such as Figure 2 The diagram shown is a flowchart of a method for predicting ozone concentration on the exterior facade of a high-rise building, provided in an embodiment of this application. The method for predicting ozone concentration on the exterior facade of a high-rise building includes the following steps S100-S400.

[0062] S100: Construct a CFD module to perform high-resolution simulation of the near-wall area of ​​the facade of a high-rise building. Use boundary layer mesh to solve and couple the ozone transport equation to output key flow field parameters and ozone concentration field. At the same time, calculate the ozone deposition flux on the wall to provide boundary conditions for the PINN module.

[0063] The purpose of step S100 is to achieve high-precision resolution of the CFD boundary layer. In some embodiments, such as Figure 3 As shown, step S100 specifically includes the following steps S101-S103.

[0064] S101: Input parameter preparation.

[0065] Input the CAD building geometry model; boundary conditions include inlet wind speed profile, ambient ozone concentration, wall roughness, and turbulence intensity.

[0066] S102: Mesh generation and solution.

[0067] Mesh generation: Boundary layer mesh (y+<1) is used in the near-wall region, with mesh size Δx≤0.1m; Solution setup: The flow field is solved using the k-ε model, coupled with the ozone transport equation.

[0068] S103: Output data.

[0069] The output data includes the near-wall region velocity field (u, v, w), pressure field (p), and ozone concentration field C. O3 And the wall ozone deposition flux. The formula for calculating the wall ozone deposition flux is:

[0070]

[0071] In the formula, J dep Where D is the wall ozone deposition flux, and D is the diffusion coefficient. The gradient of ozone concentration along the wall normal is represented by wall, where wall represents the wall surface.

[0072] S200: Construct the PINN module, and use a hierarchical modeling strategy to handle the free flow region through the PINN module, in which NO is enhanced in the near-surface layer. x The reaction term uses photolysis simplification and eddy diffusion in the middle layer and only first-order diffusion in the upper layer, outputting a corrected velocity field, ozone concentration and reaction rate.

[0073] The purpose of step S200 is to achieve PINN modeling of the stepwise reaction. In some embodiments, such as Figure 4 As shown, step S200 includes the following steps S201-S203.

[0074] S201: Pre-training phase.

[0075] Input CFD overlapping region data, train a lightweight PINN subnetwork, calculate the flow field-concentration mapping relationship in the Kinneh overlapping region, and output the subnetwork weights as the initial parameters of PINN.

[0076] S202: Construct a fully coupled PINN architecture.

[0077] The inputs to this fully coupled PINN architecture are: spatial coordinates (x, y, z), time (t), overlapping region data (u, C) provided by CFD, and environmental data.

[0078] Table 1 Stepwise Reaction Intercalation

[0079]

[0080] As shown in Table 1, the hierarchical construction method is as follows: 2nd-order chemical PINN for the near-surface layer (0-50m), with physical constraints satisfied, and the loss function including PDE residuals and data matching errors, focusing on the chemical kinetic equations of the surface layer. The reaction-diffusion equation is expressed as:

[0081]

[0082] In the formula, Let t be the ozone concentration, t be the time, and u be the flow velocity vector. Here, D is the gradient operator, and R is the diffusion coefficient. chem This represents the rate of a chemical reaction.

[0083] Intermediate layer (50-200m): photolysis simplification + eddy diffusion PINN, physical constraints: simplified reaction term (photolysis dominant), superimposed eddy diffusion coefficient v t The loss function focuses on the PDE residuals of the optical decomposition-eddy diffusion coupling process. The physical constraint equations for the intermediate layer are expressed as follows:

[0084]

[0085] In the formula, v t The eddy diffusion coefficient is... Let be the gradient operator, D be the diffusion coefficient, C be the ozone concentration, and J(x) be the photolysis rate. denoted as nitrogen dioxide concentration, and u as the flow velocity vector.

[0086] Top-level (>200m) 1st-order diffusion PINN, physical constraints: only solve the advection-diffusion equations, neglecting chemical reactions; loss function: focus on the PDE residuals of simple diffusion processes in the upper layers.

[0087] The output of this fully coupled PINN architecture is: the trained PINN model can predict the vertical gradient of ozone concentration at any location C(x,y,z,t) in the free flow region.

[0088] S203: Coupled Training.

[0089] The parameters of each layer are updated alternately, with priority given to training the ground layer (60% loss weight → 30% loss weight for the middle layer → 10% loss weight for the top layer).

[0090] S300: The physical quantity transfer between the CFD module and the PINN module is realized through a residual adaptive weighting algorithm via a dynamic coupling interface. The CFD module transmits the flow field parameters of the overlapping area to the PINN module, and the PINN module feeds back the concentration gradient to the CFD module as the top boundary condition.

[0091] The purpose of step S300 is to achieve dynamically coupled training. In some embodiments, such as Figure 5 As shown, step S300 includes the following steps S301-S303.

[0092] S301: Data input: CFD overlap region time-averaged data (u, C), PINN prediction value in the overlap region.

[0093] S302: Alternating training.

[0094] In overlapping regions (e.g., 1-1.5 times the building height), shared grid points are established, and CFD data is extracted as soft boundary conditions for PINN. With the CFD results fixed, PINN is trained to minimize L. CFD-overlap Using the high-level concentrations predicted by PINN as the top boundary conditions for CFD, the CFD solution is updated. The relative error of the concentration in the overlapping region ω < 5%. The formula for calculating the relative error of the concentration in the overlapping region is as follows:

[0095]

[0096] In the formula, C PINN For the concentration simulated by the PINN module, C CFD The concentration is from the CFD simulation.

[0097] S303: Results output: Coupled convergent PINN-CFD joint model, flow field and concentration field after coordination in the overlapping region.

[0098] S400: Construct a joint training system to train the ozone concentration prediction model. The ozone concentration prediction model includes a CFD module, a PINN module, and a dynamic coupling interface. During the training process, RMSE and error analysis are calculated. If the upper layer error exceeds the set threshold, the number of PINN intermediate layer nodes is increased or an adaptive weighted loss is introduced to optimize the ozone concentration prediction model until the preset accuracy requirements are met.

[0099] Spatiotemporal modeling of architectural scenes is achieved through a joint training system, which includes an input layer and an output layer. Specifically:

[0100] Input layer: spatial coordinates (x, y, z), timestamp (t), meteorological parameters (wind speed, solar radiation intensity, temperature, relative humidity, background ozone concentration), building geometric features (wall curvature, building orientation, surface roughness), pollution source data (NOx emission intensity, near-ground ozone precursor concentration);

[0101] Output layer: ozone concentration field, flow rate field, reaction rate (near-wall titration rate, photochemical generation rate), gradient characteristics (wall ozone deposition flux, vertical concentration gradient).

[0102] This application also provides a device for predicting ozone concentration on the exterior facade of a high-rise building, such as... Figure 6 As shown, the ozone concentration prediction device for the exterior facade of this high-rise building includes:

[0103] CFD module building unit 601 is configured to build a CFD module, which performs high-resolution simulation of the near-wall area of ​​the facade of a high-rise building, uses boundary layer mesh to solve and couple the ozone transport equation, outputs key flow field parameters and ozone concentration field, and calculates the wall ozone deposition flux to provide boundary conditions for the PINN module.

[0104] PINN module building unit 602 is configured to build PINN modules, which employ a hierarchical modeling strategy to process the free flow region, wherein the near-surface layer enhances NO. x The reaction term uses photolysis simplification and eddy diffusion in the middle layer and only first-order diffusion in the upper layer, outputting a corrected velocity field, ozone concentration and reaction rate.

[0105] The dynamic coupling unit 603 is configured to transfer physical quantities between the CFD module and the PINN module through a residual adaptive weighting algorithm via a dynamic coupling interface, wherein the CFD module transmits flow field parameters of the overlapping region to the PINN module, and the PINN module feeds back the concentration gradient to the CFD module as the top boundary condition.

[0106] The joint training unit 604 is configured to build a joint training system to train an ozone concentration prediction model. The ozone concentration prediction model includes a CFD module, a PINN module, and a dynamic coupling interface. During the training process, RMSE and error analysis are calculated. If the upper layer error exceeds a set threshold, the number of PINN intermediate layer nodes is increased or an adaptive weighted loss is introduced to optimize the ozone concentration prediction model until the preset accuracy requirement is met.

[0107] This application provides an electronic device. The electronic device may include a processor and a memory, wherein the processor and the memory can communicate; exemplarily, the processor and the memory communicate via a communication bus.

[0108] The processor executes computer execution instructions stored in memory, causing the processor to perform the schemes in the above embodiments. The processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0109] Communication buses can be Peripheral Component Interconnect (PCI) buses or Extended Industry Standard Architecture (EISA) buses, etc. System buses can be divided into address buses, data buses, control buses, etc. Transceivers are used to enable communication between database access devices and other computers (e.g., clients, read-write libraries, and read-only libraries). Memory may include random access memory (RAM) and may also include non-volatile memory.

[0110] The electronic device provided in this application embodiment can be the terminal device described in the above embodiments.

[0111] This application also provides a computer-readable storage medium storing computer instructions. When the computer instructions are executed on a computer, the computer performs the technical solution of the method for predicting ozone concentration on the facade of a high-rise building as described in the above embodiments.

[0112] This application also provides a computer program product, which includes a computer program stored in a computer-readable storage medium. At least one processor can read the computer program from the computer-readable storage medium. When the at least one processor executes the computer program, it can implement the technical solution of the method for predicting ozone concentration on the facade of high-rise buildings in the above embodiments.

[0113] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or modules, and may be electrical, mechanical, or other forms.

[0114] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to implement the solution of this embodiment according to actual needs.

[0115] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit. The unit composed of the above modules can be implemented in hardware or in the form of hardware plus software functional units.

[0116] The integrated modules described above, implemented as software functional modules, can be stored in a computer-readable storage medium. These software functional modules, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute some steps of the methods of the various embodiments of this application.

[0117] It should be understood that the aforementioned processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly manifested as execution by a hardware processor, or execution by a combination of hardware and software modules within the processor.

[0118] The memory may include high-speed RAM, and may also include non-volatile storage (NVM), such as at least one disk storage device, and may also be a USB flash drive, external hard drive, read-only memory, disk or optical disc, etc.

[0119] Buses can be Industry Standard Architecture (ISA) buses, Peripheral Component Interconnect (PCI) buses, or Extended Industry Standard Architecture (EISA) buses, etc. Buses can be categorized into address buses, data buses, control buses, etc.

[0120] The aforementioned storage medium can be implemented from any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The storage medium can be any available medium accessible to general-purpose or special-purpose computers.

[0121] An exemplary storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. Alternatively, the storage medium can be an integral part of the processor. The processor and storage medium can reside in an Application-Specific Integrated Circuit (ASIC). Alternatively, the processor and storage medium can exist as discrete components in an electronic control unit or main control device.

[0122] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0123] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. 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 or all of the technical features therein. 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.

Claims

1. A method for predicting ozone concentration on the exterior facade of a high-rise building, characterized in that, The method includes: A CFD module is constructed to perform high-resolution simulation of the near-wall area of ​​the facade of a high-rise building. Boundary layer mesh is used to solve the ozone transport equation and output key flow field parameters and ozone concentration field. At the same time, the wall ozone deposition flux is calculated to provide boundary conditions for the PINN module. A PINN module is constructed, and a hierarchical modeling strategy is used to process the free flow region through the PINN module, wherein the near-surface layer enhances NO. x The reaction term uses photolysis simplification and eddy diffusion in the middle layer and only first-order diffusion in the upper layer, outputting a corrected velocity field, ozone concentration and reaction rate. The physical quantity transfer between the CFD module and the PINN module is realized through a dynamic coupling interface using a residual adaptive weighting algorithm. The CFD module transmits the flow field parameters of the overlapping area to the PINN module, and the PINN module feeds back the concentration gradient to the CFD module as the top boundary condition. A joint training system is constructed to train the ozone concentration prediction model. The ozone concentration prediction model includes a CFD module, a PINN module, and a dynamic coupling interface. During the training process, RMSE and error analysis are calculated. If the high-level error exceeds a set threshold, the number of PINN intermediate layer nodes is increased or an adaptive weighted loss is introduced to optimize the ozone concentration prediction model until the preset accuracy requirements are met. The near-surface layer employs 2nd-order chemical PINN, and its physical constraints satisfy the following reaction-diffusion equation: In the formula, This refers to ozone concentration. t For time, u For flow velocity vectors, For gradient operators, D The diffusion coefficient is... R chem It represents the rate of a chemical reaction. The optically simplified + eddy diffusion PINN constructed in the intermediate layer has physical constraints that satisfy the following equation: In the formula, v t The eddy diffusion coefficient is... For gradient operators, D The diffusion coefficient is... J ( x () represents the photolysis rate. This refers to the concentration of nitrogen dioxide. u It is the velocity vector; When the upper layer adopts the first-order diffusion PINN, only the advection-diffusion process is considered, and complex chemical reactions are ignored. The transport law of ozone in the upper layer free flow region is described by the first-order diffusion equation.

2. The method for predicting ozone concentration on the facade of a high-rise building according to claim 1, characterized in that, The formula for calculating the wall ozone deposition flux is: In the formula, J dep For wall ozone deposition flux, D The diffusion coefficient is... This represents the ozone concentration gradient along the wall normal. wall Represents the wall.

3. The method for predicting ozone concentration on the facade of a high-rise building according to claim 1, characterized in that, The dynamic coupling interface calculates the relative error of the concentration in the overlapping area during data transmission. and control <5%; wherein, the formula for calculating the relative error of the concentration in the overlapping region is: In the formula, C PINN The concentration simulated by the PINN module. C CFD The concentration is from the CFD simulation.

4. The method for predicting ozone concentration on the facade of a high-rise building according to claim 1, characterized in that, The input parameters of the joint training system include spatial coordinates, timestamps, meteorological parameters, building geometric features, and pollution source data. The output parameters include ozone concentration field, flow velocity field, reaction rate, and gradient characteristics. The meteorological parameters include wind speed, solar radiation intensity, temperature, relative humidity, and background ozone concentration. The building geometric features include wall curvature, building orientation, and surface roughness. The pollution source data includes NO... x Emission intensity and near-ground ozone precursor concentration.

5. A device for predicting ozone concentration on the facade of a high-rise building, used to implement the method as described in any one of claims 1-4, characterized in that, The device includes: The CFD module construction unit is configured to construct a CFD module, which performs high-resolution simulation of the near-wall area of ​​the facade of a high-rise building. The CFD module uses boundary layer mesh to solve and couple the ozone transport equation, outputting key flow field parameters and ozone concentration field. At the same time, it calculates the wall ozone deposition flux and provides boundary conditions for the PINN module. The PINN module building unit is configured to build PINN modules, which employ a hierarchical modeling strategy to process the free flow region, wherein the near-surface layer enhances NO. x The reaction term uses photolysis simplification and eddy diffusion in the middle layer and only first-order diffusion in the upper layer, outputting a corrected velocity field, ozone concentration and reaction rate. The dynamic coupling unit is configured to transfer physical quantities between the CFD module and the PINN module through a residual adaptive weighting algorithm via a dynamic coupling interface, wherein the CFD module transmits the flow field parameters of the overlapping region to the PINN module, and the PINN module feeds back the concentration gradient to the CFD module as the top boundary condition. The joint training unit is configured to build a joint training system to train an ozone concentration prediction model. The ozone concentration prediction model includes a CFD module, a PINN module, and a dynamic coupling interface. During the training process, RMSE and error analysis are calculated. If the upper layer error exceeds a set threshold, the number of PINN intermediate layer nodes is increased or an adaptive weighted loss is introduced to optimize the ozone concentration prediction model until the preset accuracy requirements are met.

6. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the method for predicting ozone concentration on the facade of a high-rise building as described in any one of claims 1-4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method for predicting ozone concentration on the facade of a high-rise building as described in any one of claims 1-4.

Images