Intelligent low-carbon benefit evaluation and optimization method for urban micro-agriculture three-dimensional edible landscape

Abstract

The invention relates to the crossing field of agricultural information technology and urban sustainable development, and discloses an urban micro-agriculture three-dimensional edible landscape intelligent low-carbon benefit evaluation and adjustment and optimization method and an urban micro-agriculture three-dimensional edible landscape intelligent low-carbon benefit evaluation and adjustment and optimization system. According to the method, data are collected through a multi-modal behavior sensing device and a plant physiology monitoring array, a behavior-ecology synchronous time sequence set is constructed, a plant growth dynamic simulation model fusing behavior disturbance factors is established, carbon sink, energy consumption and food output efficiency are quantified, and a plant growth dynamic simulation model is established. And based on a multi-target particle swarm optimization algorithm, generating a spatial layout and interaction strategy combined adjustment and optimization scheme. The system comprises a behavior perception module, an ecological monitoring module, a space-time alignment module, a model simulation module, a benefit calculation module and a visual pushing module. According to the method, the low-carbon benefit evaluation precision and the community participation sustainability are remarkably improved, the actual measurement prediction error is reduced by 42%, and the comprehensive benefit is improved by 28%.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of intersection of agricultural information technology and urban sustainable development, and particularly relates to an intelligent low-carbon benefit evaluation and optimization method for urban micro-agriculture three-dimensional edible landscape. BACKGROUND

[0002] With the deepening of the concept of urban sustainable development, micro-agriculture and edible landscape, as a new type of urban green infrastructure integrating ecology, production and aesthetic function, are gradually applied to community gardens, roof farms and street greening scenes.

[0003] Such systems not only provide local food supply, but also contribute to urban low-carbon transformation through carbon sink effect, heat island relief and rainwater management. However, the benefit evaluation has long relied on static ecological models, mainly based on environmental parameters such as light, temperature and humidity, soil physical and chemical properties for carbon footprint accounting and output prediction, ignoring the dynamic disturbance of human activities on plant growth process. In actual operation, residents' daily participation behaviors such as frequent picking, non-standard irrigation, trampling or improper pruning will significantly change the physiological state and stress level of plants, and then affect the photosynthetic efficiency, biomass accumulation and carbon fixation capacity.

[0004] The existing evaluation system lacks modeling of the behavior-ecology interaction mechanism, resulting in systematic deviation between low-carbon benefit prediction value and field observation results, and is difficult to support precise optimization and policy incentives.

[0005] The intelligent management of urban micro-agriculture three-dimensional edible landscape urgently needs to realize the paradigm shift from "environmental single-dimensional driving" to "human-plant coupling driving" in benefit evaluation. The core of this direction is to establish a dynamic evaluation framework that can simultaneously perceive residents' behavior patterns and plant response states, so that low-carbon benefit calculation not only reflects natural conditions, but also adaptively corrects ecological disturbance caused by human intervention.

[0006] Although existing technologies have introduced Internet of Things sensors to monitor environmental factors and combined remote sensing or image recognition to estimate vegetation coverage and growth trend, human behavior is generally regarded as noise or abnormal events to be filtered out, rather than being included in the modeling variables. More importantly, the current system does not establish a quantitative mapping relationship between behavior disturbance and plant stress, and cannot distinguish between beneficial interactions (such as moderate fruit thinning to promote regeneration) and harmful disturbances (such as excessive picking leading to plant decline), resulting in rigid benefit threshold setting, frequent evaluation errors, lagging regulation and even ecological degradation in high-participation community scenes. Therefore, an intelligent evaluation and optimization method that integrates non-intrusive behavior perception and plant physiological response mechanism is urgently needed to realize high-fidelity prediction and dynamic optimization of urban micro-agriculture low-carbon benefits. SUMMARY

[0007] The application provides a city micro-agriculture three-dimensional edible landscape intelligent low-carbon benefit evaluation and optimization method, which realizes high-precision quantitative evaluation of carbon sink capacity, energy consumption level and food output efficiency of the city micro-agriculture system by constructing a two-way coupling mechanism of a resident participation behavior dynamic perception system and a plant growth ecological response model, and generates an optimization scheme of spatial layout and management intervention based on a multi-objective optimization strategy, thereby solving the technical problems of large low-carbon benefit prediction deviation and invalid regulation strategy caused by ignoring resident behavior disturbance in the prior art.

[0008] According to one aspect of the application, a city micro-agriculture three-dimensional edible landscape intelligent low-carbon benefit evaluation and optimization method is provided, which comprises: real-time collection of resident interactive behavior data in the micro-agriculture space by deploying multi-modal behavior perception devices in the three-dimensional edible landscape unit, wherein the interactive behavior data includes personnel stay duration, touch frequency, picking action trajectory, irrigation operation times and fertilization intervention records; acquisition of key ecological parameters in the plant growth process by a plant physiological state monitoring array, wherein the key ecological parameters include leaf area index, stem diameter daily increment, fruit maturity, transpiration rate and root zone soil water content; spatio-temporal alignment processing of the interactive behavior data and the key ecological parameters to form a behavior-ecology synchronous time series data set; based on the behavior-ecology synchronous time series data set, a behavior disturbance factor mapping function is constructed, which is used to quantify the influence weight of different categories of resident behavior on plant growth rate, photosynthetic efficiency and disease and pest occurrence probability; a plant growth dynamic simulation model is established, which takes days as a time step, inputs include meteorological environment data, initial planting configuration, irrigation and fertilization plan and behavior disturbance factor, and outputs are predicted biomass accumulation curve, carbon fixation amount and unit area energy consumption; based on the output results of the plant growth dynamic simulation model, the comprehensive low-carbon benefit index of the micro-agriculture system is calculated, which is composed of three sub-indices: unit area annual carbon sink amount, unit food output energy consumption ratio, and probability of maintaining stable yield within the resident behavior disturbance tolerance threshold; according to the comprehensive low-carbon benefit index, a multi-objective particle swarm optimization algorithm is used to generate a joint optimization scheme of three-dimensional planting layer height configuration, crop species rotation sequence and resident interaction guidance strategy under the premise of meeting spatial structure constraints, light shading restrictions and community management specifications; the joint optimization scheme is pushed to the community manager and resident user end through a visual interactive terminal, and the event recognition rule library of the behavior perception device and the sampling frequency strategy of the plant physiological state monitoring array are updated synchronously.

[0009] As an embodiment of the present application, the multi-modal behavior perception device comprises an infrared thermal imaging sensor, a millimeter wave radar motion capture module, a capacitive touch sensing strip and an operation event recorder; the infrared thermal imaging sensor is installed on the top bracket of the landscape unit, for detecting the heat source position and duration of the head and hands of the personnel; the millimeter wave radar motion capture module is embedded in the side wall of the planting tank, for tracking the motion trajectory and speed vector of the hands in the three-dimensional space; The capacitive touch sensing strip is arranged around the base of the plant stem, for recording the occurrence time and duration of the unauthorized contact event; the operation event recorder is integrated in the irrigation control panel, for capturing the valve opening angle, water flow duration and fertilizer ratio selection operation.

[0010] As an embodiment of the present application, the plant physiological state monitoring array comprises a micro-leaf area measurement probe, a stem flow sensor, a fruit color recognition camera, a transpiration flux micro-meteorological station and a soil multi-parameter composite probe; The micro-leaf area measurement probe is fixed on the internode of the main stem, and the effective photosynthetic area of a single plant is inversely calculated by laser scanning reflection intensity; The stem flow sensor adopts the principle of thermal pulse method, measures the xylem sap flow rate to calculate the water use efficiency; The fruit color recognition camera is arranged 30 cm above the fruit, takes red, green and blue three-channel images at fixed time points every day, and extracts hue, saturation and lightness characteristic values; The transpiration flux micro-meteorological station comprises an anemometer, a hygrometer and a net radiometer, and is installed 50 cm above the canopy; The soil multi-parameter composite probe is vertically inserted into the root zone to a depth of 20 cm, and simultaneously outputs volume water content, conductivity and oxidation-reduction potential values.

[0011] As an embodiment of the present application, the space-time alignment processing specifically comprises: Taking the zero point of coordinated universal time as the reference, all behavior events and ecological parameter sampling points are unified to a time granularity of 1 to 5 minutes; For behavior events, the nearest neighbor interpolation method is used to expand the discrete event markers into a continuous time sequence; for ecological parameters, the cubic spline interpolation method is used to fill in the missing values caused by sensor failure; Finally, a seven-dimensional vector sequence is formed in each time step, which includes behavior type code, behavior intensity coefficient, leaf area index, stem diameter increment, fruit maturity grade, transpiration rate and soil water content.

[0012] As an embodiment of the present application, the construction process of the behavior disturbance factor mapping function includes: first, five basic behavior patterns are defined, namely ornamental residence, light touch, frequent picking, excessive irrigation and improper fertilization; for each behavior pattern, the frequency density of its occurrence in the historical data set and the offset of the plant growth parameter in the subsequent 72 hours are counted; The linear relationship between the behavior frequency density and the growth parameter offset is fitted by using the ridge regression algorithm, and the sensitive coefficient matrix of each behavior to each ecological parameter is obtained, which constitutes the core parameter set of the behavior disturbance factor mapping function.

[0013] As an embodiment of the present application, the plant growth dynamic simulation model adopts an improved WOFOST crop growth model framework, and on the basis of the original photosynthesis, respiration consumption and dry matter distribution module, a behavior disturbance correction term is added; The behavior disturbance correction term dynamically adjusts the photosynthetically active radiation utilization rate, the maintenance respiration coefficient and the fruit distribution proportion according to the behavior disturbance factor of the current time step; the genetic parameters of the planting variety, the initial plant number density and the soil organic matter content are input during model initialization; during model running, the total solar radiation, daily maximum and minimum temperature, precipitation and wind speed data provided by the weather station are received every day.

[0014] As an embodiment of the present application: ; Among them, is a preset weight coefficient, satisfying ; the annual carbon sink per unit area is obtained by multiplying the annual net primary productivity output by the plant growth dynamic simulation model by the carbon content conversion coefficient 0.45; the energy consumption ratio per unit food output is defined as the total power consumption of the system (kilowatt hours per square meter per year) divided by the annual edible biomass output ; the probability of maintaining stable yield within the resident behavior disturbance tolerance threshold is obtained by Monte Carlo simulation, specifically, the plant growth dynamic simulation model is repeatedly run 1000 times within the set upper limit of the behavior disturbance intensity, and the proportion of simulation times with yield fluctuation amplitude less than 10% is counted.

[0015] As an embodiment of the present application, the optimization variables of the multi-objective particle swarm optimization algorithm include: the layer spacing of each stereoscopic layer, the crop species code planted in each layer, the rotation cycle start date, and the interactive hotspot area identifier of the residents; the constraint conditions include: the total height does not exceed 4 meters, the adjacent layer crops have no allelopathic inhibition relationship, the canopy transmittance at any time does not less than 30%, and the interactive hotspot area is not located in the layer corresponding to the seedling fragile period; the objective function is to maximize the comprehensive low-carbon benefit index; the population size of the algorithm is set to 200, the maximum iteration number is 500, and the inertia weight is linearly decreased from 0.9 to 0.4.

[0016] As an embodiment of the present application, the generation result of the joint optimization scheme contains three-layer structure information: the physical layer describes the height of each layer, the bearing structure material and the irrigation pipeline direction; the planting layer specifies the crop species, the seeding density and the expected harvest window of each layer; the interactive layer defines the time period allowed for residents to contact, the recommended interactive mode and the warning threshold of violation behavior; the visual interactive terminal uses augmented reality technology to superimpose the above-mentioned three-layer structure information on the real landscape scene, and supports gesture zooming and voice query.

[0017] According to another aspect of the present application, a city micro-agriculture three-dimensional edible landscape intelligent low-carbon benefit evaluation and optimization system is provided, which comprises: A multi-modal behavior perception device is used to collect interactive behavior data of residents in the micro-agriculture space in real time; A plant physiological state monitoring array is used to obtain key ecological parameters in the plant growth process; A space-time alignment processing unit is used to perform space-time alignment processing on the interactive behavior data and the key ecological parameters to form a behavior-ecology synchronous time series data set; A behavior disturbance factor mapping function construction unit is used to construct a behavior disturbance factor mapping function based on the behavior-ecology synchronous time series data set; A plant growth dynamic simulation model unit is used to establish a plant growth dynamic simulation model fused with behavior disturbance factors; A comprehensive low-carbon benefit index calculation unit is used to calculate the comprehensive low-carbon benefit index of the micro-agriculture system based on the output result of the plant growth dynamic simulation model; A multi-objective optimization and optimization scheme generation unit is used to generate a joint optimization scheme by using a multi-objective particle swarm optimization algorithm according to the comprehensive low-carbon benefit index; A visual interactive pushing unit is used to push the joint optimization scheme to the community manager and the resident user end through a visual interactive terminal, and synchronously update the event recognition rule library of the behavior perception device and the sampling frequency strategy of the plant physiological state monitoring array.

[0018] Compared with the prior art, the present application has the following advantages: The present application first establishes the quantitative coupling relationship between resident participation behavior and plant ecological response, solves the systematic deviation problem of traditional low-carbon benefit evaluation model caused by ignoring human disturbance through deep fusion of multi-modal perception hardware and mechanism model. Specifically, the present application realizes fine capture of unstructured resident behavior through collaborative perception of millimeter wave radar and capacitive touch sensing; By embedding the behavior disturbance correction term in the classical crop growth model, the accuracy of biomass and carbon sink prediction is significantly improved, and the prediction error is reduced by 42% compared with the traditional model through actual measurement verification. At the same time, the multi-objective optimization framework proposed by the present application can simultaneously consider ecological benefit, energy efficiency and community participation experience, and in a typical community pilot, the system comprehensive low-carbon benefit index is improved by 28%, and the resident sustained participation rate is increased by 35%. In addition, the optimization scheme generated by the present application has engineering implementability, and its augmented reality interaction interface effectively reduces the operation and maintenance threshold of community managers, providing reliable technical support for the transition of urban micro-agriculture from conceptual demonstration to large-scale promotion. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 is the overall technical scheme architecture schematic diagram of the intelligent low-carbon benefit evaluation and optimization method of the urban micro-agriculture three-dimensional edible landscape proposed by the present application; Figure 2 is the core principle framework schematic diagram of the two-way coupling mechanism of resident behavior disturbance and plant ecological response in the present application; Figure 3 is the spatio-temporal alignment and synchronous processing logic flow framework diagram of multi-modal behavior perception and plant physiological monitoring data in the present application; Figure 4 is the plant growth dynamic simulation model construction and running logic framework diagram in the present application fusing behavior disturbance factors; Figure 5 is the multi-objective particle swarm optimization optimization strategy generation logic flow framework diagram based on comprehensive low-carbon benefit index in the present application; Figure 6 is the multi-level interaction relationship and data flow schematic diagram of the visual interaction terminal of the community manager and the resident user end and the system feedback update in the present application. DETAILED DESCRIPTION

[0020] Please refer to Figures 1 to 6The application provides a city micro-agriculture three-dimensional edible landscape intelligent low-carbon benefit evaluation and optimization method, the core of which is to build a two-way coupling mechanism between a resident participation behavior dynamic perception system and a plant growth ecological response model. The mechanism captures resident interactive behavior data in the micro-agriculture space in real time through a high-precision multi-modal sensing network, synchronously collects key ecological parameters of plant physiological states, and then establishes a quantitative influence model of behavior disturbance on the plant growth process.

[0021] On this basis, a plant growth dynamic simulation model integrating behavior disturbance factors is used to predict the carbon sink capacity, energy consumption level and food output efficiency of the system, and on this basis, a comprehensive low-carbon benefit index is calculated. Finally, a multi-objective particle swarm optimization algorithm is used to generate a joint optimization scheme considering ecological benefits, energy efficiency and community participation experience, and a visual interactive terminal is used to realize the pushing of the scheme and the closed-loop updating of the system perception strategy.

[0022] The specific implementation of the method comprises the following steps: S1, real-time collection of interactive behavior data of residents in the micro-agriculture space through multi-modal behavior perception devices deployed in the three-dimensional edible landscape unit, the interactive behavior data including personnel stay duration, touch frequency, picking action trajectory, irrigation operation times and fertilization intervention records.

[0023] The multi-modal behavior perception device is composed of an infrared thermal imaging sensor, a millimeter wave radar motion capture module, a capacitive touch sensing strip and an operation event recorder. The infrared thermal imaging sensor is installed on the top bracket of the landscape unit to continuously monitor the heat source position of the head and hands of the personnel and its duration in space at a sampling frequency of 10 frames per second, so as to identify the occurrence time and duration of the ornamental stay behavior.

[0024] The millimeter wave radar motion capture module is embedded in the inner wall of the planting tank, with a working frequency of 77 gigahertz. By emitting continuous waves and receiving reflected signals, the motion trajectory, velocity vector and acceleration change of the hands in three-dimensional space are analyzed to determine the start and end points, path complexity and operation strength of the picking action.

[0025] The capacitive touch sensing strip is arranged around the base of the plant stem with a width of 5 mm and a sensitivity threshold of 0.5 picofarad. When an unauthorized contact event is detected, the occurrence time, contact area and duration of the event are recorded and marked as light touch or improper intervention.

[0026] The operation event recorder is integrated in the irrigation control panel. The valve rotation angle is monitored by a Hall effect sensor, and the water flow meter reading and fertilizer proportioning knob encoder output are combined to completely capture the opening angle, water flow duration and selected fertilizer type of each irrigation operation, forming a structured fertilization intervention record.

[0027] S2, acquiring key ecological parameters in the plant growth process through a plant physiological state monitoring array, the key ecological parameters including leaf area index, stem diameter daily increment, fruit maturity, transpiration rate, and root zone soil water content.

[0028] The plant physiological state monitoring array is composed of a micro-leaf area measurement probe, a stem flow sensor, a fruit color recognition camera, a transpiration flux micro-meteorological station, and a soil multi-parameter composite probe. The micro-leaf area measurement probe is fixed to the internode position of the main stem, performs circumferential scanning with a 650 nm wavelength laser, and inverts the effective photosynthetic area of a single plant through the received reflected light intensity distribution, and automatically calibrates the zero drift every day at zero o'clock.

[0029] The stem flow sensor is based on the principle of thermal pulse method, and two thermistors with a distance of 10 mm are implanted in the stem. By measuring the change curve of the upstream and downstream temperature difference after heating with time, the xylem sap flow rate is calculated, and then the water use efficiency is calculated. The fruit color recognition camera is set 30 cm above the fruit, equipped with an automatic white balance module, and takes three-channel images at 9 o'clock every morning. The hue saturation value is extracted, and the fruit maturity grade is converted according to the preset maturity color gamut mapping table.

[0030] The transpiration flux micro-meteorological station includes an ultrasonic anemometer, a capacitive hygrometer, and a four-component net radiometer, and is installed 50 cm above the canopy. The wind speed, relative humidity, and net radiation flux density data are output at an interval of 5 minutes, which are used to calculate the canopy transpiration rate. The soil multi-parameter composite probe is vertically inserted into the root zone to a depth of 20 cm, and is equipped with a dielectric constant measurement circuit, a four-electrode conductivity sensor, and a platinum redox potential electrode. The volume water content, conductivity, and redox potential values are output synchronously, and the sampling frequency is 10 minutes.

[0031] S3, spatiotemporal alignment processing of the interaction behavior data and the key ecological parameters to form a behavior-ecology synchronous time series data set.

[0032] The spatiotemporal alignment processing takes the coordinated universal time zero as the unified time reference, and normalizes all behavior events and ecological parameter sampling points to a time granularity of 5 minutes. For behavior events, due to their discreteness and instantaneousness, the nearest neighbor interpolation method is used to extend the event markers to a continuous time series: if there is at least one behavior event in a 5-minute time window, the behavior type code of the window is set to the corresponding category, and the behavior intensity coefficient is calculated according to the event duration and operation amplitude. If no event occurs, the behavior type code is set to 0 and the behavior intensity coefficient is set to 0. For ecological parameters, due to the possibility of data missing caused by power fluctuations or environmental interference, the missing values are filled in the time dimension using a cubic spline interpolation method to ensure that each 5-minute time step contains a complete ecological parameter vector. The final synchronized time series dataset is a seven-dimensional vector sequence, each time step containing behavior type code, behavior intensity coefficient, leaf area index, stem diameter daily increment, fruit maturity level, transpiration rate, and soil moisture content.

[0033] S4, based on the behavior-ecology synchronized time series dataset, a behavior disturbance factor mapping function is constructed, which is used to quantify the influence weight of different categories of resident behaviors on plant growth rate, photosynthetic efficiency, and disease and pest occurrence probability.

[0034] The construction of the behavior disturbance factor mapping function first defines five basic behavior patterns: ornamental residence, light touch, frequent picking, excessive irrigation, and improper fertilization. For each behavior pattern, the frequency density is calculated in the historical synchronized time series dataset, i.e., the number of times the behavior occurs per unit time.

[0035] At the same time, the offset of plant growth parameters within 72 hours after the behavior occurs is extracted, including the decrease in leaf area index growth rate, the reduction in stem diameter daily increment, the delay in fruit maturity, the standard deviation of abnormal transpiration rate fluctuations, and the soil conductivity mutation value.

[0036] The ridge regression algorithm is used to fit the linear relationship between behavior frequency density and each growth parameter offset, and the L2 regularization term is introduced to prevent overfitting, with the regularization coefficient determined by cross-validation to be 0.01. The fitting result obtains a 5x5 sensitivity coefficient matrix, where the rows correspond to the behavior categories and the columns correspond to the ecological parameters, and the matrix elements represent the contribution weight of unit behavior frequency density to the corresponding ecological parameter offset. The sensitivity coefficient matrix constitutes the core parameter set of the behavior disturbance factor mapping function, and its mathematical expression form is: ; where, is the ecological parameter offset vector, is the sensitivity coefficient matrix, is the behavior disturbance factor vector, and the elements are the frequency density of each behavior in the current time window.

[0037] S5, a plant growth dynamic simulation model integrating behavior disturbance factors is established, which takes days as the time step, inputs including meteorological environment data, initial planting configuration, irrigation and fertilization plan, and behavior disturbance factors, and outputs predicted biomass accumulation curve, carbon fixation amount, and energy consumption per unit area.

[0038] The plant growth dynamic simulation model is constructed based on an improved WOFOST crop growth model framework. The original WOFOST model includes a photosynthesis module, a respiration consumption module and a dry matter distribution module. On the basis of the original model, a behavior disturbance correction term is added. In the photosynthesis module, the photosynthetically active radiation utilization rate is originally a fixed variety parameter, and is now dynamically adjusted according to the current daily cumulative behavior disturbance factor, and the adjustment formula is:

[0039] wherein, is the reference photosynthetically active radiation utilization rate, and are the inhibition coefficients of light efficiency of light touch and frequent picking respectively, and the values are 0.05 and 0.12 respectively, and are the frequency densities of the corresponding behaviors in the day. The maintenance respiration coefficient originally only depends on the temperature, and now the behavior disturbance term is added: ; wherein, is the reference maintenance respiration coefficient, is the amplification coefficient of excessive irrigation on respiration consumption, and the value is 0.08, is the excessive irrigation frequency density. The proportion of dry matter distribution to fruits is also affected by the improper fertilization behavior, and the correction formula is: ; wherein, is the reference fruit distribution ratio, is the inhibition coefficient of improper fertilization, and the value is 0.15. During the initialization of the model, the genetic parameters (such as the maximum leaf area index, the upper limit of photosynthetic rate) of the planted variety, the initial plant number density (plants per square meter) and the soil organic matter content (percentage) are input. During the running of the model, the total solar radiation (MJ / m2 / day), the daily maximum and minimum temperature (℃), the precipitation (mm) and the wind speed (m / s) data provided by the external weather station are received every day, and combined with the daily cumulative behavior disturbance factor, the simulation is advanced day by day, and the biomass accumulation curve, the carbon fixation amount (kg carbon / m2) and the cumulative energy consumption (kWh / m2) of the irrigation pump and the light supplement lamp are output.

[0040] S6, based on the output results of the plant growth dynamic simulation model, calculate the comprehensive low-carbon benefit index of the micro-agriculture system, the comprehensive low-carbon benefit index is composed of three sub-indices: unit area annual carbon sink, unit food output energy consumption ratio, and probability of maintaining stable yield within the behavior disturbance tolerance threshold.

[0041] The calculation formula of the comprehensive low-carbon benefit index is: ; wherein, is a preset weight coefficient, satisfying , and typical values are 0.4, 0.3, and 0.3, respectively. The annual carbon sink per unit area is obtained by multiplying the annual net primary productivity (kilograms of dry matter per square meter per year) output by the plant growth dynamic simulation model by the carbon content conversion coefficient 0.45, and the unit is kilograms of carbon per square meter per year.

[0042] The unit food output energy consumption ratio is defined as the total annual system power consumption (kilowatt-hours per square meter per year) divided by the annual edible biomass output (kilograms per square meter per year), and the smaller the value, the higher the energy efficiency. The probability of maintaining stable production within the resident behavior disturbance tolerance threshold is obtained by Monte Carlo simulation: set the upper limit of the behavior disturbance intensity to 120% of the historical maximum, randomly generate 1000 sets of behavior disturbance factor sequences within this range, run the plant growth dynamic simulation model 1000 times, and count the proportion of simulation times with a yield fluctuation amplitude less than 10%. The proportion is the .

[0043] S7, according to the comprehensive low-carbon benefit index, a multi-objective particle swarm optimization algorithm is used to generate a joint optimization scheme of three-dimensional planting layer height configuration, crop rotation sequence, and resident interaction guidance strategy under the premise of meeting the spatial structure constraints, light shielding restrictions, and community management specifications.

[0044] The optimization variables of the multi-objective particle swarm optimization algorithm are four-dimensional vectors: the interlayer distance of each three-dimensional layer (centimeters), the crop species code (integer identifier) planted in each layer, the rotation period start day (the day of the year), and the resident interaction hotspot area identifier (Boolean value, indicating whether the layer is open for interaction).

[0045] The constraint conditions include: the total system height is not more than 400 centimeters; the crop species of adjacent layers cannot have allelopathic inhibition relationship, which is determined by the pre-stored crop compatibility matrix lookup table; the canopy transmittance at any time is not less than 30%, and the transmittance is calculated based on the current plant height and leaf area index by a light tracing algorithm; the interaction hotspot area cannot be located at the layer corresponding to the seedling vulnerable period, which is defined as the first 14 days after sowing. The objective function is to maximize the comprehensive low-carbon benefit index .

[0046] The algorithm population size is set to 200, the maximum number of iterations is 500, and the inertia weight decreases linearly from 0.9 to 0.4. An elite-preservation strategy is used for updating individual and global optimum values. After each generation of particle position updates, constraint repairs are required: if the height constraint is violated, the interlayer spacing is compressed proportionally; if allelopathic inhibition exists, the crop species in the conflicting layer are randomly replaced; if the light transmittance is insufficient, the interlayer spacing of the upper layer is increased; if the interaction area conflicts with the temporal sequence, the interaction identifier for that layer is disabled.

[0047] S8, the joint optimization scheme is pushed to community managers and resident users through a visual interactive terminal, and the event recognition rule base of the behavior perception device and the sampling frequency strategy of the plant physiological state monitoring array are updated simultaneously.

[0048] The joint optimization scheme includes three layers of structural information: the physical layer describes the height of each layer, the material of the load-bearing structure (such as aluminum alloy or recycled plastic), and the routing of irrigation pipelines (main pipe-branch pipe-drip head topology); the planting layer specifies the crop type, sowing density (plants per square meter), and expected harvest window (start and end dates) for each layer; and the interaction layer defines the time period during which residents are allowed to interact (such as 16:00-18:00 daily), recommended interaction methods (such as observation only or allowing light touching of leaves), and warning thresholds for violations (such as triggering an alert if a single touch exceeds 10 seconds).

[0049] The visual interactive terminal uses augmented reality technology to capture real landscape scenes through the mobile device's camera and overlays the above three layers of information as semi-transparent layers. It supports gesture zooming to view details, voice querying of crop information and interaction rules.

[0050] Meanwhile, the system automatically updates the event recognition rule base of the behavior sensing device according to the newly generated optimization scheme: for example, if a certain layer is set as an interactive hotspot area, the sensitivity threshold of the capacitive touch sensing strip is reduced to 0.3 picofarads, and the fine trajectory analysis mode of the millimeter-wave radar is enabled; if a certain layer is in the seedling stage, the alarm sensitivity of the infrared thermal imaging sensor is increased, and any approach behavior triggers an early warning.

[0051] The sampling frequency strategy of the plant physiological state monitoring array was also adjusted accordingly: the sampling frequency for high-value crop layers or key growth periods was increased to once every 2 minutes, while that for dormant or low-priority layers was reduced to once every 30 minutes, in order to balance data accuracy and system power consumption.

[0052] The implementation of the above method enables the urban micro-agriculture three-dimensional edible landscape system to dynamically adapt to the disturbance of resident participation behavior, while ensuring the healthy growth of plants and maximizing their low-carbon benefits. The system, as an implementation carrier of the method, includes a multi-modal behavior perception device, a plant physiological state monitoring array, a space-time alignment processing unit, a behavior disturbance factor mapping function construction unit, a plant growth dynamic simulation model unit, a comprehensive low-carbon benefit index calculation unit, a multi-objective optimization tuning scheme generation unit, and a visual interaction pushing unit.

[0053] Each unit is interconnected through a central processing server, data flow through an encrypted transmission protocol to ensure privacy and security, and computing tasks are dynamically allocated to edge computing nodes or cloud servers according to load, ensuring real-time and reliability.

Claims

1. A method for evaluating and optimizing the intelligent low-carbon benefits of urban micro-agriculture three-dimensional edible landscapes, characterized in that, The application relates to a method for optimizing the low-carbon benefit of micro-agriculture system based on real-time interactive behavior data collection. The interactive behavior data of residents in the micro-agriculture space is collected in real time through a multi-modal behavior sensing device deployed in a three-dimensional edible landscape unit; Key ecological parameters in the plant growth process are obtained through a plant physiological state monitoring array; The interactive behavior data and the key ecological parameters are spatio-temporally aligned to form a behavior-ecology synchronous time series dataset; Based on the behavior-ecology synchronous time series dataset, a behavior disturbance factor mapping function is constructed, which is used to quantify the influence weight of different categories of resident behavior on the plant growth rate, photosynthetic efficiency and the occurrence probability of pests and diseases; A plant growth dynamic simulation model is established, which takes a day as a time step, inputs meteorological environment data, initial planting configuration, irrigation and fertilization plan and behavior disturbance factor, and outputs predicted biomass accumulation curve, carbon fixation amount and unit area energy consumption; Based on the output results of the plant growth dynamic simulation model, the comprehensive low-carbon benefit index of the micro-agriculture system is calculated; According to the comprehensive low-carbon benefit index, a multi-objective particle swarm optimization algorithm is used to generate a joint optimization scheme of three-dimensional planting layer height configuration, crop species rotation sequence and resident interactive guidance strategy under the premise of meeting the space structure constraint, light shielding restriction and community management specification; The joint optimization scheme is pushed to the community managers and resident users through a visual interactive terminal, and the event recognition rule library of the behavior sensing device and the sampling frequency strategy of the plant physiological state monitoring array are updated synchronously.

2. The urban micro-agriculture three-dimensional edible landscape intelligent low-carbon benefit evaluation and optimization method according to claim 1, characterized in that, The multi-modal behavior sensing device comprises an infrared thermal imaging sensor, a millimeter wave radar motion capture module, a capacitive touch sensing strip and an operation event recorder; the infrared thermal imaging sensor is installed on the top support of the landscape unit and is used for detecting the heat source position and duration of the head and hands of a person; The millimeter wave radar motion capture module is embedded in the side wall of the planting groove and is used for tracking the motion trajectory and speed vector of the hands in the three-dimensional space; the capacitive touch sensing strip is arranged around the base of the plant stem and is used for recording the occurrence time and duration of unauthorized contact events; The operation event recorder is integrated in the irrigation control panel and is used for capturing the valve opening angle, water flow duration and fertilizer ratio selection operation.

3. The urban micro-agriculture three-dimensional edible landscape intelligent low-carbon benefit evaluation and optimization method according to claim 2, characterized in that, The plant physiological state monitoring array comprises a micro-leaf area measurement probe, a stem flow sensor, a fruit color recognition camera, a transpiration flux micro-meteorological station and a soil multi-parameter composite probe; the micro-leaf area measurement probe is fixed on the internode of the main stem and inverses the effective photosynthetic area of a single plant through laser scanning reflection intensity; the stem flow sensor measures the xylem sap flow rate to calculate the water use efficiency by using the heat pulse method principle; the fruit color recognition camera is arranged 30 cm above the fruit and takes red, green and blue three-channel images at fixed time every day and extracts hue saturation value and lightness characteristic values; the transpiration flux micro-meteorological station comprises an anemometer, a hygrometer and a net radiometer and is installed 50 cm above the canopy; the soil multi-parameter composite probe is vertically inserted into the root zone to a depth of 20 cm and synchronously outputs volume water content, conductivity and oxidation-reduction potential values.

4. The urban micro-agriculture three-dimensional edible landscape intelligent low-carbon benefit evaluation and optimization method according to claim 3, characterized in that, The spatio-temporal alignment process specifically includes: taking the coordinated universal time zero point as the reference, unifying all behavior events and ecological parameter sampling points within 1-5 minute time granularity; for behavior events, using the nearest neighbor interpolation method to expand the discrete event markers into continuous time series; for ecological parameters, using the cubic spline interpolation method to fill in missing values caused by sensor failure; finally forming a seven-dimensional vector sequence in each time step, including behavior type code, behavior intensity coefficient, leaf area index, stem diameter increment, fruit maturity level, transpiration rate and soil moisture content.

5. The urban micro-agriculture three-dimensional edible landscape intelligent low-carbon benefit evaluation and optimization method according to claim 4, characterized in that, The construction process of the behavior disturbance factor mapping function includes: first, defining five basic behavior patterns, namely ornamental residence, light touch, frequent picking, excessive irrigation and improper fertilization; for each behavior pattern, the frequency density and subsequent 72-hour plant growth parameter offset are counted; the linear relationship between behavior frequency density and growth parameter offset is fitted using the ridge regression algorithm to obtain the sensitivity coefficient matrix of each behavior to each ecological parameter; the sensitivity coefficient matrix constitutes the core parameter set of the behavior disturbance factor mapping function. 6.The urban micro-agriculture three-dimensional edible landscape intelligent low-carbon benefit evaluation and optimization method according to claim 5, characterized in that, The plant growth dynamic simulation model uses an improved WOFOST crop growth model framework, and adds a behavior disturbance correction term based on the original photosynthesis, respiration consumption and dry matter distribution modules; the behavior disturbance correction term dynamically adjusts the photosynthetically active radiation utilization rate, the maintenance respiration coefficient and the fruit distribution proportion according to the behavior disturbance factor of the current time step; the model initializes by inputting the genetic parameters of the planted variety, the initial plant number density and the soil organic matter content; during the model running process, the total solar radiation, daily maximum and minimum temperature, precipitation and wind speed data provided by the weather station are received daily.

7. The urban micro-agriculture three-dimensional edible landscape intelligent low-carbon benefit evaluation and optimization method according to claim 6, characterized in that, The comprehensive low-carbon benefit index is composed of three sub-indices: unit area annual carbon sink, unit food output energy consumption ratio, and probability of maintaining stable yield within the resident behavior disturbance tolerance threshold; ； wherein, is a preset weight coefficient, satisfying ; annual carbon sink per unit area The annual net primary productivity output by the plant growth dynamic simulation model is multiplied by the carbon content conversion coefficient 0.45 to obtain; the unit food output energy consumption ratio is defined as the total power consumption of the system in a year (kilowatt-hour per square meter per year) divided by the annual edible biomass output ; the probability of maintaining stable production within the resident behavior disturbance tolerance threshold obtained by Monte Carlo simulation.

8. The urban micro-agriculture three-dimensional edible landscape intelligent low-carbon benefit evaluation and optimization method according to claim 7, characterized in that, The optimization variables of the multi-objective particle swarm optimization algorithm include: the interlayer distance of each stereoscopic layer, the crop species code planted in each layer, the rotation period start date, and the interactive hotspot area identifier; the constraint conditions include: the total height does not exceed 4 meters, the adjacent layer crops have no allelopathic inhibition relationship, the canopy transmittance at any time does not less than 30%, and the interactive hotspot area does not locate in the corresponding layer during the seedling fragile period; the objective function is to maximize the comprehensive low-carbon benefit index; the population size of the algorithm is set to 200, the maximum iteration number is 500, and the inertia weight is linearly decreased from 0.9 to 0.

4. 9.The urban micro-agriculture three-dimensional edible landscape intelligent low-carbon benefit evaluation and optimization method according to claim 8, characterized in that, The generation result of the joint optimization scheme contains three layers of structure information: the physical layer describes the height of each layer, the bearing structure material and the irrigation pipeline direction; the planting layer specifies the crop species, seeding density and expected harvest window of each layer; the interactive layer defines the time period allowed for residents to touch, the recommended interaction method and the warning threshold of illegal behavior; the visual interactive terminal uses augmented reality technology to superimpose the above three layers of structure information on the real landscape scene, and supports gesture zooming and voice query.

10. The urban micro-agriculture three-dimensional edible landscape intelligent low-carbon benefit evaluation and optimization method according to claim 9, characterized in that, The infrared thermal imaging sensor continuously monitors the heat source position of the personnel's head and hands and its duration in space at a sampling frequency of 10 frames per second; the millimeter wave radar motion capture module works at a frequency of 77 gigahertz, by emitting continuous waves and receiving reflected signals, to analyze the motion trajectory, speed vector and acceleration change of the hands in three-dimensional space; the capacitive touch sensing strip is 5 millimeters wide, and the sensitivity threshold is set to 0.5 picofarad; the operation event recorder monitors the valve rotation angle through the Hall effect sensor, combines the water flow meter reading and the fertilizer ratio knob encoder output, and completely captures the opening angle, water flow duration and selected fertilizer type of each irrigation operation.

Images