A method and system suitable for a cooling and humidifying type plant community configuration

Abstract

The application relates to a method and system suitable for temperature reduction and humidity increase type plant community configuration and relates to the technical field of ecological environment engineering, which comprises the following steps: S1, obtaining environmental data of a target area and attribute data of candidate plants; S2, constructing a space-time synergistic transpiration potential field model based on the environmental data and the attribute data, the space-time synergistic transpiration potential field model being used for simulating and quantifying space-time dynamic temperature reduction potential and humidity increase potential generated by plant community synergy in the target area under different plant configuration schemes; and S3, generating a temperature reduction and humidity increase type plant community configuration scheme of the target area by using a multi-agent reinforcement learning algorithm. The application significantly improves the scientificity and accuracy of temperature reduction and humidity increase type plant community configuration by systematically obtaining detailed environmental data of the target area and key attribute data of the candidate plants and constructing a fine space-time synergistic transpiration potential field model.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of ecological environment engineering, and in particular to a method and system suitable for cooling and humidifying plant community configuration. BACKGROUND

[0002] Improvement of urban thermal environment and promotion of human comfort are important issues faced by current urban planning and landscape design. Using plant communities for ecological cooling and humidification has become a widely concerned technical means due to its environmental friendliness and strong sustainability. Through reasonable plant configuration, local microclimate can be effectively adjusted, urban heat island effect can be alleviated, and more pleasant living and working environment can be created.

[0003] In existing cooling and humidifying plant community configuration practices, the formation of design schemes largely depends on the experience and knowledge of designers, traditional garden plant configuration principles, and qualitative cognition of plant individual ecological habits. Designers usually select and match plant species according to the general situation of the target area and common plant cooling and humidification mechanisms (such as shading and transpiration). However, this method mainly based on experience and qualitative judgment often has difficulty in accurately quantifying the overall cooling and humidification efficiency of plant communities and making prospective predictions when facing complex and variable target site environments and diversified plant characteristics. Therefore, the actual effect of the configuration scheme may deviate from the expected result, making it difficult to ensure the best microclimate regulation goal.

[0004] At the same time, the existing technology has certain limitations in considering the complex interaction between plant communities and the environment, especially the time and space dynamic energy and water exchange processes. The transpiration of plants is affected by dynamic factors such as light, temperature, humidity, and wind speed, and different plant species, different configuration densities and structures of communities have significant differences in the way and degree of microenvironment modification. Traditional configuration methods often have difficulty in precisely simulating these complex physical processes and synergistic effects between plants, such as the interception and redistribution of solar radiation by the canopy, the cumulative impact of group transpiration on local air temperature and humidity, and the possible microclimate feedback of different plant combinations in a specific spatial layout. This lack of understanding of deep mechanisms and group effects limits the ability of existing configuration schemes to maximize the ecological potential of plants and achieve fine regulation of cooling and humidification effects in a specific space.

[0005] In addition, in the process of seeking the optimal plant configuration scheme, the prior art often lacks efficient optimization means. The configuration of a cooling and humidifying type plant community is a complex decision-making problem involving multiple objectives (such as maximizing the cooling effect, maximizing the humidifying effect, controlling the cost, meeting the spatial constraints, considering plant diversity, etc.) and multiple constraints. In the face of the challenge of a large number of candidate plant species and almost infinite spatial layout combination methods, manual trial and error or optimization based on simple rules cannot systematically explore the entire solution space to find a configuration scheme with the optimal overall benefit. Therefore, the prior art still has room for improvement in generating a plant community configuration scheme that can balance multiple needs and achieve a globally or approximately globally optimal cooling and humidifying effect under specific constraint conditions. SUMMARY

[0006] The purpose of the present application is to provide a method and system suitable for cooling and humidifying type plant community configuration, which solves the problems of inaccurate prediction of cooling and humidifying effect, lack of fine consideration of plant community synergy effect, and difficulty in achieving configuration scheme optimization under complex constraints in the prior art.

[0007] In a first aspect, the present application provides a method suitable for cooling and humidifying type plant community configuration, comprising the following steps:

[0008] S1, obtaining environmental data of a target area and attribute data of candidate plants;

[0009] S2, based on the environmental data and the attribute data, constructing a spatiotemporal synergistic transpiration potential field model, the spatiotemporal synergistic transpiration potential field model being used to simulate and quantify the spatiotemporal dynamic cooling potential and humidifying potential generated by the synergistic effect of plant communities in different plant configuration schemes in the target area;

[0010] S3, using a multi-agent reinforcement learning algorithm, taking the cooling potential and humidifying potential predicted by the spatiotemporal synergistic transpiration potential field model as the optimization target, to generate a cooling and humidifying type plant community configuration scheme for the target area.

[0011] Preferably, the step of constructing a spatiotemporal synergistic transpiration potential field model comprises:

[0012] discretizing the target area in three-dimensional space to obtain a plurality of grid cells;

[0013] for each grid cell, coupling a canopy radiation transfer model, a plant transpiration and evaporation model, an intra-canopy aerodynamic model, and a three-dimensional heat and water vapor transport model;

[0014] wherein the canopy radiation transfer model comprises the step of calculating the shortwave radiation intensity I z at any height zz in the canopy, and the calculation formula is:

[0015] I z = top exp(-K ext ·LAI z ·G L / cos(θ sun ));

[0016] where I top is the incident shortwave radiation intensity at the top of the canopy; K ext is the extinction coefficient of the canopy; LAI z is the cumulative leaf area index from the top of the canopy to depth z; G L is the leaf inclination distribution function; and θ sun is the solar zenith angle.

[0017] Preferably, the plant transpiration and evaporation model comprises a step of calculating the latent heat flux ET L for each grid cell containing vegetation, using the Penman-Monteith equation:

[0018]

[0019] where Δ svp is the slope of the saturation vapor pressure-temperature curve; R net is the net radiation; G soil is the soil heat flux; ρ air is the air density; c p,air is the air specific heat at constant pressure; e sat (T surf ) is the saturation vapor pressure corresponding to the surface temperature T surf ; e air is the actual vapor pressure; r aero is the aerodynamic resistance; γ psych is the psychrometric constant; r canopy is the canopy resistance; F stress is the water stress factor; F group is the population synergistic transpiration impact factor.

[0020] The population synergistic transpiration impact factor is used to quantify the synergistic promotion or inhibition effect of local microenvironment changes caused by the presence of a plant population on the transpiration of individual plants.

[0021] Preferably, the step of using a multi-agent reinforcement learning algorithm comprises:

[0022] defining the configuration units of the target area as a plurality of agents;

[0023] defining a state space for each of the agents, the state space comprising micro-environment information of the agent itself, current plant configuration information, and plant configuration information of adjacent cells;

[0024] defining an action space for each of the agents, the action space comprising plant species to be planted and planting density.

[0025] Preferably, the step of employing the multi-agent reinforcement learning algorithm further comprises:

[0026] designing a reward function R k for evaluating overall benefits generated by a plant configuration scheme P(a k ) formed after all agents take joint action a k at decision time k, the reward function being calculated according to the formula:

[0027]

[0028] wherein ΔT cell,i,k+1 (P(a k )) and Δq cell,i,k+1 (P(a k )) are temperature reduction and specific humidity increase of grid cell i at the next evaluation period after performing action a k predicted based on the spatiotemporal synergistic transpiration potential field model; Cells represents a set of all grid cells within the target area; w temp and w hum are weight coefficients of temperature reduction benefits and humidity increase benefits; Cost(a k ) is configuration cost related to performing joint action a k ; Penalty(a k ) is a penalty term for violating preset constraints; Bonus div (a k ) is a reward term for improving species diversity or meeting specific aesthetic indicators.

[0029] Preferably, the step of employing the multi-agent reinforcement learning algorithm further comprises:

[0030] the plurality of agents are trained in interaction with a simulation environment in which the spatiotemporal synergistic transpiration potential field model is built-in, in the training process, the agents perform actions according to current strategies to form plant configurations, the simulation environment feeds back rewards according to the plant configurations, and the agents update their strategies according to the rewards until the strategies converge.

[0031] Preferably, the attribute data of the candidate plants comprise:

[0032] physiological parameters, morphological parameters, phenological parameters and environmental adaptability parameters of plants;

[0033] The physiological parameters include maximum stomatal conductance and its response function to environmental factors, leaf optical properties.

[0034] The morphological parameters include leaf area index, leaf area density and its vertical distribution.

[0035] Preferably, the generated cooling and humidifying type plant community configuration scheme comprises:

[0036] The species, quantity or density of the selected plants in each configuration unit in the target area and spatial layout information.

[0037] Preferably, the method further comprises:

[0038] Long-term monitoring of the implemented plant community configuration scheme to obtain actual operation effect data and site environment dynamic change data;

[0039] When the actual operation effect data deviates from the expected or the site environment changes, the monitoring data and change data are taken as new inputs to re-run the multi-agent reinforcement learning algorithm for re-optimization to generate an adjusted plant community configuration scheme.

[0040] In a second aspect, the present application provides a system suitable for cooling and humidifying type plant community configuration, comprising:

[0041] A data acquisition module for acquiring environmental data of a target area and attribute data of candidate plants;

[0042] A spatiotemporal synergistic transpiration potential field modeling module connected with the data acquisition module, for constructing and running a spatiotemporal synergistic transpiration potential field model based on the environmental data and the attribute data, to simulate and quantify the spatiotemporal dynamic cooling potential and humidifying potential generated by the synergistic effect of plant communities under different plant configuration schemes in the target area;

[0043] An intelligent optimization configuration module connected with the spatiotemporal synergistic transpiration potential field modeling module, for using a multi-agent reinforcement learning algorithm to generate a cooling and humidifying type plant community configuration scheme for the target area, taking the cooling potential and humidifying potential predicted by the spatiotemporal synergistic transpiration potential field model as the optimization target;

[0044] A scheme output module connected with the intelligent optimization configuration module, for outputting the cooling and humidifying type plant community configuration scheme.

[0045] In summary, the present application includes at least one of the following beneficial technical effects:

[0046] 1. This invention significantly improves the scientific rigor and accuracy of cooling and humidifying plant community configuration by systematically acquiring detailed environmental data and key attribute data of candidate plants in the target area, and constructing a refined spatiotemporal transpiration potential field model. This analytical method based on detailed data and mechanistic models allows plant configuration to move beyond mere experience and enable quantitative design based on the microclimate conditions of a specific site and the physiological and ecological characteristics of the plants, thereby ensuring the relevance and effectiveness of the configuration scheme.

[0047] 2. This invention utilizes a constructed spatiotemporal synergistic transpiration potential field model to quantitatively predict and assess the overall cooling and humidifying potentials resulting from the interactions among plant communities under different plant community configurations. This model not only considers the transpiration characteristics of individual plants but, more importantly, simulates the comprehensive impact of plant communities on microenvironmental factors such as light, wind, temperature, and humidity under specific spatial arrangements. This allows for the identification and optimization of synergistic effects among plants, achieving more efficient regulation of regional microclimates.

[0048] 3. This invention employs a multi-agent reinforcement learning algorithm, using the output of a spatiotemporal collaborative transpiration potential field model as the core optimization objective. Combined with various constraints in actual configurations (such as cost, space limitations, and species suitability), it can intelligently explore and generate optimized cooling and humidifying plant community configuration schemes. This intelligent optimization method can efficiently search for high-quality solutions that balance cooling and humidifying effects with practical feasibility from a vast array of possible combinations, overcoming the limitations of traditional design methods when facing complex multi-objective optimization problems.

[0049] 4. This invention couples the dynamic physiological processes of plants, the spatial structure of communities, and the spatiotemporal changes of environmental factors within a unified modeling and optimization framework. This enables the generated plant community configuration scheme to better adapt to the dynamic environmental characteristics of the target area and achieve a rational distribution of cooling and humidifying benefits in space. This design approach, which considers spatiotemporal synergy, helps to achieve the long-term effective functioning of plant ecosystems and the continuous improvement of the microclimate in specific areas, thereby enhancing the environmental benefits and sustainability of plant community configuration. Attached Figure Description

[0050] Figure 1 This is a flowchart of the method used in this application;

[0051] Figure 2 This is the system architecture diagram of this application. Detailed Implementation

[0052] The following is in conjunction with the appendix Figure 1 This application will be described in further detail below.

[0053] Embodiment: A method suitable for configuring a cooling and humidifying type plant community, comprising the following steps:

[0054] S1, obtaining environmental data of a target area and attribute data of candidate plants;

[0055] In the method, first, step S1 is performed: obtaining environmental data of a target area and attribute data of candidate plants. This step is the starting point of the entire methodology, and the accuracy and completeness of its output directly affect the realism of subsequent model simulation and the effectiveness of optimization results.

[0056] When obtaining environmental data of the target area, exemplarily, it can include collecting geographic spatial information. The geographic spatial information is used to accurately describe the physical boundaries and topographic features of the configuration area. For example, vector boundary data and digital elevation model (DEM) data of the target area can be obtained by consulting geographic information system (GIS) database, analyzing high-resolution remote sensing images, or conducting field survey, etc. These data are the geometric basis for subsequent spatial analysis and model grid division.

[0057] Further, when obtaining geographic spatial information, it can also include recording the spatial position, geometric form and attributes of existing buildings, roads, hard paving, water bodies and other important structures in the area. These information is crucial for evaluating the influence of existing environment on plant growth, and considering the effects of shading, reflection, etc. in the model. All geographic spatial data need to be georeferenced to ensure their spatial consistency in a unified coordinate reference system, and appropriate spatial resolution should be selected according to the simulation accuracy requirement.

[0058] When obtaining environmental data of the target area, it also includes collecting meteorological data. Meteorological data is used to represent the climate background conditions of the target area at a specific time scale, and is a key input for driving plant physiological processes and microclimate changes. Exemplarily, it can be obtained from standard meteorological observation stations near the target area, authoritative regional meteorological model reanalysis data sets (such as ERA5, NCEPFNL, etc.), or by monitoring on site through small automatic meteorological stations in the target area.

[0059] Preferably, the collected meteorological data should cover a long time series, such as historical data of the last 5 to 10 years, and specific representative periods, such as typical summer high temperature days or specific seasonal hourly or higher time frequency meteorological parameters. These parameters should at least include: total solar radiation (Rs), unit: watt per square meter (W / m2); air temperature (T), unit: Celsius or Kelvin (K); air relative humidity (RH), unit: percentage; wind speed (u), unit: meter per second (m / s). s,in ), unit: watt per square meter (W / m 2 ); air temperature (T air,ambient ), unit: Celsius or Kelvin (K); air relative humidity (RH ambient ), unit: percentage; wind speed (u ambientwind speed (ws), in meters per second; and wind direction (wd ambient ), in degrees. For the acquired raw meteorological data, strict quality control is needed, such as removing obvious outliers, filling in missing data by using appropriate interpolation or regression methods, to ensure the data quality.

[0060] When acquiring the environmental data of the target area, soil data can also be included. Soil data is of great significance for assessing the water and nutrient conditions available to plants and for simulating the surface energy balance. For example, soil sampling in the target configuration area and laboratory analysis, or consulting local soil survey reports, soil databases, etc. can be used to obtain the data.

[0061] Preferably, the investigated soil data should include the physicochemical property parameters of the surface layer and the main plant root activity layer of the target area. These parameters can include soil type (such as sandy soil, loamy soil, clay), soil texture (such as the percentage of each particle size), soil bulk density (g / cm 3 ), soil porosity, soil field capacity, soil wilting point water content, and soil saturated hydraulic conductivity, etc. These parameters will directly affect the calculation of plant water stress factors in subsequent models and the simulation of soil water dynamics.

[0062] When acquiring the environmental data of the target area, spatial constraints can also be included. Spatial constraints define the actual feasible boundaries and limiting factors of plant configuration, and are an important basis for ensuring the feasibility of the configuration scheme. For example, it is necessary to define and digitize the range of the area available for plant configuration (i.e. specific planting plots or planting areas), the maximum growth height limit allowed for planting plants in each area (possibly due to line of sight, safety or regulatory requirements), the precise distribution location of important underground pipelines (such as water supply, drainage, power, gas, communication pipelines) and their safety distance requirements, as well as the necessary traffic access or sight corridor requirements above ground, etc.

[0063] In step S1, the attribute data of the candidate plants is acquired, aiming to establish a comprehensive plant feature information library to provide a basis for subsequent plant selection and model parameterization. First, according to the climate conditions, soil characteristics, and expected cooling and humidification targets of the target area, as well as the local native plant resources, a preliminary selection of candidate plant species with known or potential good cooling and humidification effects and capable of adapting to the local growing environment is made. These candidate plants can include trees, shrubs, herbaceous plants, and vines, etc. of different life forms.

[0064] Subsequently, for each selected candidate plant, its key physioecological parameters, morphological feature parameters, phenological rhythm parameters, and environmental adaptability parameters are systematically collected and quantified, which is the core of plant attribute data acquisition.

[0065] When acquiring the physiological parameters, exemplarily, the maximum stomatal conductance (g s,max ) of the plant can be included. The maximum stomatal conductance is a key indicator of the potential transpiration capacity of the plant. At the same time, the response function or model parameters of the plant to key environmental factors need to be obtained. These response relationships can be obtained by consulting published plant physiology literature, ecological research reports or specialized plant physiological ecology databases, or determined by experimental measurement when conditions permit. For example, mature mathematical models such as the Jarvis model or the Ball-Berry model can be used to describe the response of stomatal conductance to environmental factors. In addition, the physiological parameters should also include the average albedo (a leaf ), transmittance (τ leaf ) and absorption rate of the leaf in the visible and near-infrared wave bands, which directly affect the radiation energy balance of the canopy. Water use efficiency (WUE), as an indicator of the balance between plant carbon fixation and water consumption, should also be included.

[0066] When acquiring the morphological parameters, exemplarily, the typical plant height (H plant ) and crown width (W canopy ) of the plant at the mature stage or when reaching the target configuration specification can be included. More importantly, parameters related to the canopy structure, such as the leaf area index (LAI plant ), which is defined as the total leaf single surface area above the unit ground surface, need to be obtained. At the same time, the leaf area density (LAD plant (z′)) and its vertical distribution function within the canopy need to be understood, where z′ is the relative height within the canopy. The vertical distribution of LAD affects the attenuation of light radiation within the canopy, the wind speed profile, and the source-sink distribution of heat and water vapor. These morphological parameters can be estimated by consulting horticultural manuals, flora, or by field measurement (e.g., using stratified felling method, LAI-2200 canopy analyzer, TRAC optical instrument, etc.) or mathematical models based on canopy geometry. In addition, the main distribution depth (D root ) of the root system and the morphological characteristics of the root system are also important morphological parameters related to the water absorption capacity of the plant.

[0067] When acquiring the phenological parameters, exemplarily, the main phenological nodes of the plant can be recorded, such as the leaf expansion period when the leaves begin to sprout, the maximum leaf area period when the leaf area reaches the maximum, and the leaf fall period when the leaves begin to wither and fall, etc. The average occurrence date or the effective accumulated temperature required for these phenological nodes is crucial for accurately simulating the seasonal changes of parameters such as leaf area index (LAI) in dynamic models, thereby affecting the seasonal dynamics of plant transpiration.

[0068] When obtaining the environmental adaptability parameters, exemplary, it can include collecting the tolerance threshold of plants to various environmental stress factors (such as drought, high temperature, strong light, low temperature, soil salinization, etc.) or qualitative / quantitative adaptation level. These information helps to select plants that can grow healthily in the target area in the long term in the configuration scheme, and can be used as constraint conditions or evaluation indexes in the intelligent optimization process.

[0069] Finally, after obtaining the above various types of attribute data of all candidate plants, it is necessary to standardize and construct a structured candidate plant database. Each record in the database corresponds to a candidate plant and contains all its quantified and standardized physiological, morphological, phenological and adaptability parameters. The establishment of the database provides systematic and standardized data support for the parameterization input of the subsequent spatiotemporal synergistic transpiration potential field model and the definition of the agent action space in the multi-agent reinforcement learning algorithm.

[0070] S2, based on the environmental data and the attribute data, constructing a spatiotemporal synergistic transpiration potential field model, the spatiotemporal synergistic transpiration potential field model being used for simulating and quantifying the spatiotemporal dynamic cooling potential and humidifying potential in the target area generated by the synergistic effect of plant communities under different plant configuration schemes;

[0071] After performing step S1 and obtaining detailed target area environmental data and candidate plant attribute data, the method of the present application further performs step S2: based on the environmental data and the attribute data, constructing a spatiotemporal synergistic transpiration potential field model (STCPF model). The core purpose of this step is to establish a physical simulation platform that can comprehensively reflect the complex interactions between plant physiological and ecological characteristics, plant population spatial configuration and environmental factors. The platform is used to quantitatively simulate and evaluate the spatiotemporal dynamic cooling potential and humidifying potential in the target area due to the synergistic effect of plant populations under different plant community configuration schemes, providing a scientific evaluation basis for subsequent intelligent optimization configuration.

[0072] To achieve the above-mentioned objectives, in the step of constructing the spatiotemporal synergistic transpiration potential field model, the target area needs to be discretized in three-dimensional space first, which is divided into multiple grid units (or voxels) with determined geometric size and spatial position. This discretization process is the basis for subsequent numerical calculation and spatial analysis, and the size of the grid unit (for example, Δx, Δy, Δz) is determined according to the simulation accuracy requirement, the calculation resource limitation and the characteristic scale of the studied physical phenomenon.

[0073] Subsequently, for each discrete grid cell, at different simulation time steps, the STCPF model couples multiple key physical process sub-models. These sub-models work together to simulate energy exchange and water cycling between the plant community and the environment. Exemplarily, these sub-models may include: a canopy radiative transport model, a plant transpiration and evaporation model, a canopy aerodynamics model, and a three-dimensional heat and water vapor transport model.

[0074] The canopy radiative transfer model aims to accurately calculate the shortwave solar radiation and longwave environmental radiation flux reaching different depths of the plant canopy, the ground, and other related surfaces, and ultimately determine the net radiation for each surface involved in energy exchange. Net radiation is the primary energy source driving plant transpiration and surface evaporation.

[0075] In this canopy radiative transfer model, the transmission of shortwave radiation needs to consider the interception, reflection, transmission, and absorption processes of direct solar radiation and sky-scattered radiation in a complex three-dimensional environment (including vegetation canopy, buildings, ground, etc.). For example, the shortwave radiation intensity I at any height z within the canopy... z It can be estimated using an extended form of Beer-Lambert's law, and the calculation formula is as follows:

[0076] I z =I top exp(-K ext ·LAI z ·G L / cos(θ sun ));

[0077] Among them: I top The intensity of incident shortwave radiation at the top of the canopy (W / m²) 2 Its value can be determined by the local meteorological data (such as total solar radiation) and solar geometric position parameters (calculated by date, time and geographical location) obtained in step 51;

[0078] K ext The extinction coefficient (dimensionless) represents the canopy, which is related to factors such as the spatial orientation distribution of the leaves, the optical properties of the leaves (reflectivity, transmittance), and the incident angle of solar radiation.

[0079] LAI z The cumulative leaf area index (m²) represents the area from the top of the canopy (usually set to z = 0) to the current calculation depth z. 2 Leaf area / m 2 The area of ​​the plant (G) reflects the total amount of shading by leaves along the light path, and its vertical distribution information is used to determine plant morphological parameters; LThe leaf tilt angle distribution function (dimensionless) describes the statistical distribution characteristics of the angle between the normal direction of the leaves and the zenith direction within the canopy, which affects the canopy's interception efficiency of incident light from different directions.

[0080] θ sun It represents the solar zenith angle (in degrees or radians), which is the angle between the sun's rays and the Earth's surface normal.

[0081] For longwave radiation transfer, the model needs to calculate the longwave radiation exchange from downward atmospheric radiation, upward ground radiation, and the emission and reception from adjacent vegetation and building surfaces. Ultimately, the model will calculate the net radiation Ri for each surface unit i involved in energy exchange (such as vegetation leaf surfaces, bare soil surfaces, hard paved surfaces, etc.). net,i (W / m 2 The net radiation consists of the sum of the short-wave radiation absorbed by the surface and the net long-wave radiation absorbed.

[0082] The core function of plant transpiration and evaporation models is to calculate the transpiration rate of plants and the evaporation rate of exposed soil or water surfaces in each grid cell containing vegetation. These processes are the direct physical mechanisms by which plant communities produce cooling effects (through the consumption of latent heat) and humidifying effects (through the release of water vapor).

[0083] In this plant transpiration and evaporation model, the latent heat flux (i.e., transpiration rate, expressed in energy units as ET) for a plant canopy unit is... L Preferably, the Penman-Monteith equation is used for calculation. This equation comprehensively considers the effects of energy supply and water vapor transport resistance on the transpiration process, and its calculation formula is as follows:

[0084]

[0085] Where: Δ svp The slope of the saturated vapor pressure-temperature curve at the current air temperature represents the degree of influence of temperature change on saturated vapor pressure.

[0086] R net Represents the net radiation received by the leaf or canopy unit (W / m²) 2 The value of G is calculated using the aforementioned canopy radiative transfer model. soil Represents soil heat flux (W / m 2 Heat flux refers to the heat entering or leaving the soil, which can be estimated based on the surface energy balance or empirical formulas. When there is vegetation cover, it usually refers to the soil heat flux below the canopy.

[0087] ρ air It represents air density and is affected by temperature and air pressure.

[0088] c p,airThe specific heat at constant pressure of air;

[0089] e sat (T surf () represents the temperature on the blade surface;

[0090] T surf The saturated vapor pressure below;

[0091] e air The actual water vapor pressure of the air surrounding the blade; r aero Represents aerodynamic drag, which is the resistance to water vapor diffusion from the blade surface to the surrounding air, and is affected by wind speed and canopy structure;

[0092] γ psych This represents the hygrometer constant, which is related to atmospheric pressure and the latent heat of vaporization of water;

[0093] r canopy Representing canopy drag, it is the total resistance to water vapor diffusion from the intercellular spaces of leaf mesophyll cells through stomata to the leaf surface, mainly composed of stomatal drag (r). stomata The stomatal resistance of leaves is determined by and influenced by the leaf area index. Stomatal resistance is regulated by various environmental and physiological factors such as light intensity, VPD, temperature, CO2 concentration, and leaf water potential. These regulatory mechanisms are quantified using plant physiological parameters (such as maximum stomatal conductance and its response function) obtained in step S1.

[0094] F stress The water stress factor (dimensionless, value range 0-1) is used to quantify the inhibitory effect of insufficient soil moisture on plant transpiration. Its value is determined based on the relationship between the current effective soil water content (determined by soil data in step S1, field capacity, and wilting point parameters) and the range of water available to the plant.

[0095] F group The representative factor is a dimensionless synergistic transpiration factor. This factor is an important feature in this invention used to reflect the influence of plant community-level interactions on transpiration. It aims to quantify the net synergistic promoting or inhibiting effect on the transpiration rate of individual plants or canopy units caused by changes in local microclimates (such as increased humidity within the canopy, decreased VPD, and reduced wind speed) due to the dense presence of plant communities. group The specific form can be parameterized based on the density, spatial arrangement, and interaction with the surrounding environment of the plant community, through empirical functions or based on more refined microscale flow field simulation results.

[0096] The purpose of an aerodynamic model within the canopy is to simulate the obstruction and disturbance effects of the plant canopy on near-surface airflow, thereby determining the wind speed profile and turbulence characteristic parameters inside and above the canopy, and to calculate the aerodynamic drag r of heat and water vapor exchange.aero The wind speed profile above the canopy can be described by a logarithmic law, while the wind speed inside the canopy can be significantly attenuated and can be estimated by an exponential decay model, whose decay coefficient is related to the leaf area density (LAD) and its vertical distribution of the canopy. The aerodynamic resistance r aero , which is closely related to the wind speed, canopy structure (e.g. height, density, roughness element size, zero plane displacement height), and atmospheric stability, can be calculated by a formula based on the Monin-Obukhov similarity theory or other empirical formulas.

[0097] A three-dimensional heat and water vapor transport model, the core of which is to simulate the spatiotemporal dynamic changes of air temperature and air humidity in the target region caused by plant transpiration and evaporation, heat exchange between the ground and the air, and the advection transport and turbulent diffusion of airflow by solving fluid mechanics and thermodynamics control equations. Exemplarily, the following form of unsteady three-dimensional convection-diffusion equation can be solved:

[0098] For the transport of air temperature (T air ):

[0099]

[0100] For the transport of air specific humidity (q spechum ) or water vapor concentration (C vapor ):

[0101]

[0102] Wherein:

[0103] t represents time;

[0104] V wind =(u, v, w) represents a three-dimensional wind speed vector field, the distribution of which can be provided by an externally coupled computational fluid dynamics (CFD) model, or generated by a simplified parameterized wind field model (such as a model based on the law of conservation of mass) combined with the background wind field data obtained in step S1 and the influence of terrain and buildings;

[0105] represents the Hamiltonian operator, indicating the spatial gradient;

[0106] K heat and K vapor represent the turbulent eddy diffusion coefficients of heat and water vapor, respectively, which are related to atmospheric stability and turbulent intensity, and can be calculated by different parameterization schemes;

[0107] S TThe source-sink term of the air temperature equation mainly includes the temperature change caused by the sensible heat flux exchange between the plant canopy and the ground and the air, and the heating / cooling effect caused by the direct absorption of solar radiation or the release of long-wave radiation by the air.

[0108] S q The source-sink term of the air specific humidity equation mainly includes the water vapor released into the atmosphere by plant transpiration and surface evaporation.

[0109] Latent heat flux ET L is a main positive source term of S q , and is also an important negative source term of S T (in the form of -ET L / (ρ air c p,air ), which represents evaporative cooling). Sensible heat flux H sensible (usually calculated by ρ air c p,air (T surf -T air ) / r aero ) is another main source-sink term of S T .

[0110] After constructing and parameterizing the above-mentioned sub-models, the operation of the STCPF model involves numerically solving these coupled physical processes.

[0111] The core output of the STCPF model is that for a given plant community configuration scheme (i.e., which type of plant is planted in which grid cell and its density), under specific external meteorological conditions, the model can predict and quantify the temperature reduction ΔT cell,i,k and humidity increase (for example, specific humidity increase Δq cell,i,k or relative humidity increase ΔRH cell,i,k ) of each three-dimensional grid cell i in the target area at each simulation time step k relative to a baseline scenario without vegetation. These dynamically distributed temperature reduction and humidity increase values in three-dimensional space and time dimensions together constitute the so-called "spatiotemporal synergistic transpiration potential field".

[0112] S3, using a multi-agent reinforcement learning algorithm, taking the temperature reduction and humidity increase potential predicted by the spatiotemporal synergistic transpiration potential field model as the optimization target, to generate a temperature reduction and humidity increase type plant community configuration scheme for the target area.

[0113] To implement this step, a framework for multi-agent reinforcement learning must first be constructed. This involves defining units within the target area that can be used for plant configuration as multiple independent agents. For example, these configuration units may correspond to the grid cells obtained during spatial discretization in step S1, or more practically meaningful planting plots composed of several adjacent grid cells. The system will contain N... agent These are intelligent agents that work together to complete the task of plant configuration throughout the region.

[0114] Furthermore, it is necessary to define an agent for each intelligent agent. i (where i is the index of the agent, from 1 to N) agent At each decision time k, define the observable state space S. i,k This state information serves as the basis for the agent's decision-making. The state space can, for example, include the following types of information:

[0115] Firstly, the microenvironment information of the intelligent agent's own unit, such as the unit's geographical coordinates, terrain features (e.g., slope, aspect), actual area available for planting, average light conditions at the current moment or during a representative period (which can be estimated or simplified by the radiation module of the STCPF model, such as light intensity level), local wind environment features (e.g., wind speed level), soil type, and main physicochemical properties (from data in step S1). If the system allows adjustments based on existing plant configurations, the status should also include whether the unit is currently planted with plants, the types, sizes, and growth status of the planted plants, etc.

[0116] Secondly, current plant configuration information, which specifically refers to the plant information that the agent itself has selected during the multi-step decision-making process.

[0117] Third, plant configuration information of neighboring units, such as information related to the agent. i The current plant planting status of neighboring units includes the plant species selected by neighboring units, the current or expected height of the plants, canopy width, and leaf blood trifoliate index. This proximity information is crucial for the agent to learn the synergistic effects (such as shading, wind protection, and humidity sharing) or competitive relationships (such as light competition and water competition) among plants, helping to form a more optimized group configuration. Optionally, the state space can also include some global information, such as the total cost of the currently configured scheme, the resources used (such as budget), the achieved plant species diversity index, the gap from the overall configuration goal, or the current global time step of the simulation.

[0118] Next, for each agent i Define it in a specific state s i,k ∈S i,k The action space that can be taken is A.i,k The action of the agent directly determines the plant planting scheme of the configuration unit where it is located. The action space is usually designed as a set of discrete options, which can include, for example:

[0119] First, a specific plant type (PType m ) is selected from the candidate plant database established in step S1 for planting, where m is the index of the plant type.

[0120] Second, a suitable planting density or quantity (Density n ) is determined for the selected plant type, which can be defined as low, medium, and high density levels, or corresponding to a specific number of plants per unit area or a percentage of coverage, where n is the index of the density level.

[0121] Third, it includes a special action, i.e., selecting "not planting any plants" or "removing the existing plants in the current unit". Therefore, a specific action a i taken by the agent at time k i,k can be a tuple containing the above information.

[0122] All agents together form a complete plant community configuration scheme P(a k ) formed at time k.

[0123] .

[0124] In order to guide the agent to learn a configuration strategy that can produce good cooling and humidifying effects and meet other requirements, a reward function R k needs to be carefully designed. The reward function is used to evaluate the overall benefits of the plant configuration scheme P(a k ) formed after all agents take joint action a k at decision time k. After each decision step, the plant configuration scheme P(a k ) will be input into the constructed STCPF model for simulation evaluation, and the output of the STCPF model (i.e., cooling potential and humidifying potential) will be the core basis for calculating the reward. The calculation formula of the reward function can be exemplarily as follows:

[0125]

[0126] where: ΔT cell,j,k+1 (P(a k )) represents the average or cumulative temperature reduction that the grid unit j in the target area can achieve in the next evaluation period after executing the joint action a k based on the STCPF model prediction;

[0127] Δq cell,j,k+1 (P(a k )) represents the average or cumulative amount of increase in relative humidity or other humidity indicators that grid cell j in the target area can achieve in the next evaluation period after performing joint action a k ;

[0128] Cells represents the set of all grid cells in the target area that participate in the evaluation of the cooling and humidifying benefits, which can be all the configured cells or a pre-selected set of key sensitive areas (e.g. pedestrian walkways, resting areas, near building windows, etc.); temp and w hum represent the weight coefficients (dimensionless) for the cooling and humidifying benefits, respectively, used to adjust the relative importance of the two in the total reward, whose specific values can be set according to the specific needs of the application scenario;

[0129] Cost(a k ) represents the configuration cost related to performing joint action a k , including the procurement cost of selected plants, planting construction cost, and possibly the short-term maintenance cost estimate. The cost term exists to guide the algorithm to consider economic efficiency while pursuing benefits;

[0130] Penalty(a k ) represents the penalty term for violating pre-set constraints. These constraints can include: the total configuration cost exceeds the pre-set upper limit of the budget; certain characteristics of selected plants exceed the limits of specific configuration cells; plants are configured in micro-environmental conditions that are not suitable for their growth; the configuration scheme destroys necessary sight corridors or emergency passage spaces; or the configuration scheme may cause adverse ecological consequences, etc. When the constraints are violated, this term is a large negative value to prevent the agent from learning such behaviors; div Bonus k (a ) represents the reward term for configuration schemes that can enhance the species diversity of plant communities, meet specific aesthetic indicators (such as color matching, seasonal changes), or enhance regional ecological connectivity, etc. The existence of this term can guide the algorithm to generate configuration schemes that are not only functionally superior, but also perform well in other aspects.

[0131] After defining the MARL framework (agent, state, action) and reward function, a specific multi-agent reinforcement learning algorithm needs to be selected and implemented, and the agent needs to be trained. According to the scale of the target problem, the complexity of the collaboration relationship between agents, and the availability of computing resources, a suitable MARL algorithm can be selected.

[0132] Typically, a training paradigm of centralized training and decentralized execution can be adopted. In this paradigm, the training phase can utilize global information (e.g., the joint state and joint actions of all agents) or a centralized network of critics to guide the learning process of policies for all agents, which helps to address the problems of credit allocation and non-stationarity in multi-agent learning. In the execution phase after training (i.e., when actually generating configuration schemes), each agent makes decisions independently based solely on its own local observations, which better reflects the situation in real-world applications where agents may not have access to global information.

[0133] Multiple agents interact and train in a simulated environment that incorporates the STCPF model constructed in step S2. The specific training process is as follows:

[0134] First, at the start of each training round, the simulation environment is initialized, for example, the target configuration area is treated as blank, or it has a certain preset initial plant configuration state.

[0135] Then, at each decision time step k:

[0136] All agents i Based on its current policy network π i (s i,k ) and its local state s observed at the current moment. i,k Select their respective actions a i,k .

[0137] The joint action of all agents a k The current plant community configuration scheme P(a) constitutes k It was submitted to the STCPF model.

[0138] The STCPF model runs a complete simulation (e.g., simulating the microclimate effects of a typical day or a specific key period) and outputs the configuration scheme P(a k The spatiotemporal distribution data of cooling potential ΔT and humidification potential Δq of each grid cell within the target area.

[0139] Based on the output of the STCPF model and the predefined reward function, calculate the current joint action a. k The global reward R obtained k At the same time, the system transitions to the next state s. k+1 (where s) k+1 By s k and a k This is determined by the dynamic evolution of the environment; for example, the configuration status of neighboring units may have been updated.

[0140] The selected MARL algorithm utilizes the collected experience samples to update the policy network parameters of each agent. The update process usually involves minimizing an algorithm-dependent loss function through a gradient descent (or its variants) algorithm, with the goal of enabling the agent to learn a policy that maximizes the cumulative expected reward. This interactive training process is repeated for a large number of training rounds, or until the agent's policy performance converges, or a pre-set maximum number of training steps / time is reached.

[0141] After the MARL model is fully trained and converges, the trained agent policies are utilized to generate the final cooling and humidifying plant community configuration scheme. Specifically, for a given target area (whose environmental data and constraints have been obtained in step S1), the trained MARL decision-making process is run once. During this process, each agent agent i According to its final learned optimal or near-optimal policy and the information it observes in the current (usually initial) state, outputs the action it considers optimal (i.e., selects which plant species, adopts what planting density, etc.). The collection of optimal actions output by all agents collectively constitutes the final cooling and humidifying plant community configuration scheme for the target area.

[0142] The generated configuration scheme should specify the recommended plant species to be planted in each configuration unit (e.g., each grid or plot) within the target area, its specifications (e.g., recommended nursery diameter / ground diameter level, crown width level or height level), specific planting quantity or planting density, and specific spatial layout information of these plants.

[0143] Finally, to facilitate user understanding and implementation, the generated plant community configuration scheme can be presented in various forms by the scheme output module. For example, it can be output as a detailed configuration report document, which includes a plant list (listing the scientific name, Chinese name, family, selected specifications, required quantity, etc. of each plant), a plant spatial distribution map, a quantitative description of the expected cooling and humidifying effect of the final configuration scheme under typical meteorological conditions based on the STCPF model, a potential distribution map, an estimated total cost of the scheme, a species diversity index evaluation result, and brief post-care management recommendations for selected plants, etc. In addition, three-dimensional effect maps or interactive virtual reality (VR) scenes can also be generated to more intuitively demonstrate the effects of the completed configuration scheme.

[0144] In combination with the accompanying Figure 2 Another embodiment of the present application provides a system suitable for cooling and humidifying plant community configuration, comprising:

[0145] a data acquisition module for acquiring environmental data of a target area and attribute data of candidate plants;

[0146] The spatiotemporal synergistic transpiration potential field modeling module is connected with the data acquisition module, and is configured to construct and run a spatiotemporal synergistic transpiration potential field model based on the environmental data and the attribute data, so as to simulate and quantify the spatiotemporal dynamic cooling potential and humidifying potential generated by the synergistic effect of the plant community in the target region under different plant configuration schemes;

[0147] The intelligent optimization configuration module is connected with the spatiotemporal synergistic transpiration potential field modeling module, and is configured to adopt a multi-agent reinforcement learning algorithm, take the cooling potential and the humidifying potential predicted by the spatiotemporal synergistic transpiration potential field model as the optimization target, and generate a cooling and humidifying type plant community configuration scheme for the target region.

[0148] The scheme output module is connected with the intelligent optimization configuration module, and is configured to output the cooling and humidifying type plant community configuration scheme.

[0149] The system of the embodiment can be used to execute the method embodiments described above, and has similar principles and technical effects, which will not be described here.

[0150] The embodiments of the specific implementation are preferred embodiments of the present application, and are not intended to limit the protection scope of the present application, wherein the same parts are denoted by the same reference numerals. Therefore, equivalent changes made according to the structure, shape, principle of the present application should be covered within the protection scope of the present application.

Claims

1. A method suitable for a temperature-lowering, humidifying type plant community arrangement, characterized by, The method comprises the following steps: S1, obtaining environmental data of a target area and attribute data of candidate plants; S2, constructing a spatiotemporal synergistic transpiration potential field model based on the environmental data and the attribute data, the spatiotemporal synergistic transpiration potential field model being used for simulating and quantifying spatiotemporal dynamic cooling and humidifying potential generated by synergistic action of plant communities in different plant configuration schemes in the target area; S3, using a multi-agent reinforcement learning algorithm to generate a cooling and humidifying plant community configuration scheme for the target area, with the cooling and humidifying potential predicted by the spatiotemporal synergistic transpiration potential field model as an optimization target, The step of constructing a spatiotemporal synergistic transpiration potential field model comprises: discretizing the target area in three-dimensional space to obtain a plurality of grid cells; for each grid cell, coupling a canopy radiation transfer model, a plant transpiration and evaporation model, an intra-canopy aerodynamic model, and a three-dimensional heat and water vapor transport model; The canopy radiation transmission model comprises the following steps of calculating the shortwave radiation intensity at any height zz in the canopy The calculation formula is: ； where: is the incident shortwave radiation intensity at the top of the canopy; is the canopy extinction coefficient; is the cumulative leaf area index from the top of the canopy to depth z; is the leaf inclination distribution function; is the solar zenith angle, The step of using a multi-agent reinforcement learning algorithm further comprises: Designing a reward function for evaluating the joint action taken by all agents at decision time k Plant configuration scenarios formed thereafter The overall benefit generated, the reward function is calculated as: ； wherein: and are the cooling and humidifying benefits, respectively, predicted by the spatiotemporal coordinated transpiration potential field model for the grid cells after performing the action at the next evaluation period; represent the set of grid cells within all target areas; and are the weight coefficients for the cooling and humidifying benefits, respectively; is the configuration cost associated with performing the joint action ; is a penalty term for violating the preset constraint conditions; is a reward term for promoting species diversity or meeting specific aesthetic indicators.

2. The method for configuring a temperature-lowering and humidifying type plant community according to claim 1, wherein The plant transpiration and evaporation model includes the step of calculating the latent heat flux for each grid cell containing vegetation using the Penman-Monteith equation: ； where: is the slope of the saturation vapor pressure-temperature curve; is the net radiation; is the soil heat flux; is the air density; is the air pressure-specific heat; is the surface temperature is the corresponding saturation vapor pressure; is the actual vapor pressure; is the aerodynamic resistance; is the psychrometer constant; is the canopy resistance; is the water stress factor; is the population synergistic transpiration impact factor; The plant community synergistic transpiration influencing factor is used to quantify the synergistic promotion or inhibition effect of local microenvironment changes caused by the presence of plant communities on individual plant transpiration.

3. The method for configuring a temperature-lowering and humidifying type plant community according to claim 1, wherein The step of using a multi-agent reinforcement learning algorithm comprises: defining configuration units of the target area as a plurality of agents; defining a state space for each agent, the state space comprising microenvironment information of the agent's own unit, current plant configuration information, and plant configuration information of adjacent units; defining an action space for each agent, the action space comprising a plant species to be planted and a planting density.

4. The method for configuring a temperature-lowering and humidifying type plant community according to claim 3, wherein The step of using a multi-agent reinforcement learning algorithm further comprises: The plurality of agents interact with a simulation environment in which the spatiotemporal synergistic transpiration potential field model is embedded for training, during which the agents update their strategies according to rewards fed back by the simulation environment according to plant configurations formed by the agents according to current strategies, until the strategies converge.

5. The method for configuring a temperature-lowering and humidifying type plant community according to claim 1, wherein The attribute data of the candidate plants comprises: physiological parameters, morphological parameters, phenological parameters, and environmental adaptability parameters of plants; the physiological parameters comprise maximum stomatal conductance and its response function to environmental factors, and leaf optical properties; the morphological parameters comprise leaf area index, leaf area density, and vertical distribution thereof.

6. The method for configuring a temperature-lowering and humidifying type plant community according to claim 1, wherein The generated cooling and humidifying plant community configuration scheme comprises: species, quantity or density, and spatial layout information of plants selected for each configuration unit in the target area.

7. The method for configuring a temperature-lowering and humidifying type plant community according to claim 1, wherein The method further comprises: long-term monitoring of an implemented plant community configuration scheme to obtain actual operation effect data and site environmental dynamic change data; when the actual operation effect data deviates from the expected or the site environment changes, using the monitoring data and change data as new inputs to re-run the multi-agent reinforcement learning algorithm for re-optimization to generate an adjusted plant community configuration scheme.

8. A system suitable for the configuration of a hypothermal and hygrophytic plant community according to the method of any one of claims 1 to 7, characterized in that, comprise: a data acquisition module configured to acquire environmental data of a target area and attribute data of candidate plants; The spatiotemporal synergistic transpiration potential field modeling module is connected with the data acquisition module and is configured to construct and run a spatiotemporal synergistic transpiration potential field model based on the environmental data and the attribute data, so as to simulate and quantify the spatiotemporal dynamic cooling potential and humidifying potential generated by synergistic action of plant communities in the target region under different plant configuration schemes; The intelligent optimization configuration module is connected with the spatiotemporal synergistic transpiration potential field modeling module and is configured to adopt a multi-agent reinforcement learning algorithm, take the cooling potential and humidifying potential predicted by the spatiotemporal synergistic transpiration potential field model as an optimization target, and generate a cooling and humidifying type plant community configuration scheme for the target region; The scheme output module is connected with the intelligent optimization configuration module and is configured to output the cooling and humidifying type plant community configuration scheme.

Images