Forestry seedling-raising complete-cycle carbon footprint tracking and energy-saving type circulating irrigation system

Abstract

The invention discloses a forestry seedling-raising complete-cycle carbon footprint tracking and energy-saving type circulating irrigation system, and the system comprises a carbon footprint tracking subsystem which is provided with a multi-source sensing network module, an edge calculation node module and a block chain evidence storage module, and is used for collecting carbon emission data of a seedling-raising complete cycle in real time, the multi-source sensing network module realizes three-dimensional acquisition of seedling-raising complete-cycle carbon footprint data through a heterogeneous sensor cluster; the gradient water circulation subsystem is provided with a rainwater collection unit, a water treatment unit and a precise irrigation unit, and water resource stage treatment is achieved; the carbon-water collaborative optimization subsystem is provided with a reinforcement learning model and a digital twinborn platform, the reinforcement learning model balances carbon emission, saves water and nursery stock growth rate and dynamically generates an optimal irrigation strategy, and the digital twinborn platform realizes real-time interaction and visualization and assists the reinforcement learning model in decision making. The method at least has the following beneficial effects: full-period dynamic carbon footprint tracking is realized; gradient water circulation and energy-saving irrigation; and carbon-water intelligent collaborative optimization is carried out.

Description

Technical Field

[0001] The present application relates to the field of forestry seedling cultivation technology, and specifically to a full-cycle carbon footprint tracking and energy-saving circulating irrigation system for forestry seedling cultivation. Background Art

[0002] In the forestry seedling cultivation process, the coordinated optimization of carbon footprint management and water resource utilization is the key to improving ecological benefits. However, existing technologies have the following significant drawbacks:

[0003] Static carbon accounting methods are lagging behind: Traditional carbon footprint assessments rely on static life cycle assessments (LCA), which can only provide periodic reports and are unable to capture the dynamic fluctuations of carbon emissions at various stages of seedling cultivation (such as seed treatment, irrigation, and transportation) in real time. This leads to delayed carbon management decisions and makes it difficult to respond to sudden high-carbon emission events.

[0004] Carbon-water management disconnect: Irrigation systems are mostly based on fixed-time control, without dynamically adjusting pump operating hours based on the grid's carbon intensity, resulting in concentrated energy consumption during periods of high carbon emissions.

[0005] The water treatment process is mainly based on reverse osmosis, with energy consumption accounting for as high as 45%. The water-saving benefits have not been converted into quantitative indicators of carbon emission reduction, and the resource coordination efficiency is low.

[0006] Insufficient data credibility: Carbon emission data relies on centralized server storage, which poses a risk of tampering and lacks a traceable audit mechanism, making it difficult to meet the trusted certification requirements of the carbon trading market.

[0007] Conflict between seedling growth and resource efficiency: Existing optimization models mostly focus on a single goal (such as water saving or increasing production), and lack a multi-objective coordination mechanism for carbon emissions, water saving and seedling growth, which can easily lead to over-irrigation or uncontrolled carbon emissions.

[0008] These issues severely restrict the sustainable development of forestry seedling production. Therefore, a system integrating dynamic carbon tracking, water conservation and consumption reduction, and intelligent decision-making is urgently needed to achieve precise and coordinated management of carbon and water resources. Summary of the Invention

[0009] To this end, this application provides a full-cycle carbon footprint tracking and energy-saving circulating irrigation system for forestry seedling cultivation to solve the problems existing in the existing technology.

[0010] In order to achieve the above objectives, this application provides the following technical solutions:

[0011] The full-cycle carbon footprint tracking and energy-saving circulating irrigation system for forestry seedling cultivation includes:

[0012] The carbon footprint tracking subsystem, which includes a multi-source sensor network module, an edge computing node module, and a blockchain evidence storage module, collects carbon emission data throughout the entire seedling cultivation cycle in real time. The multi-source sensor network module uses a heterogeneous sensor cluster to achieve three-dimensional collection of carbon footprint data throughout the entire seedling cultivation cycle.

[0013] Gradient water circulation subsystem, which has rainwater collection unit, water treatment unit and precision irrigation unit to achieve graded water resource treatment;

[0014] The carbon-water collaborative optimization subsystem deploys a reinforcement learning model and a digital twin platform. The reinforcement learning model balances carbon emissions, water conservation, and seedling growth rate, dynamically generating the optimal irrigation strategy. The digital twin platform enables real-time interaction and visualization, and assists the reinforcement learning model in decision-making.

[0015] Optionally, the multi-source sensor network module includes:

[0016] 1) Energy consumption monitoring layer: smart meters are used to collect three-phase power data of each process equipment in real time, and Hall sensors are used to monitor the instantaneous fuel consumption of diesel generators;

[0017] 2) Water consumption monitoring layer: ultrasonic flowmeters are used. Time difference sensors are installed in the main irrigation pipes with a transmission frequency of 1 MHz. The flow rate is calculated by measuring the difference in ultrasonic propagation time between upstream and downstream.

[0018] 3) Biomass monitoring layer: a multispectral camera equipped with a 5-channel filter is used to obtain the NDVI index of the seedling canopy through drone inspections, and the biomass carbon sequestration is inverted using the random forest algorithm;

[0019] 4) Environmental monitoring layer, using NDIR principle CO2 sensor, dual-wavelength infrared light source and thermopile detector, calculates the concentration according to Beer-Lambert law.

[0020] Optionally, the edge computing node module is deployed with a real-time calculation model for carbon emission intensity, and the formula is:

[0021]

[0022] Among them, C intensity Real-time carbon emission intensity, unit is kgCO2e / h or kgCO2e / m 3 , α is the dynamic coefficient of carbon intensity of the power grid, the unit is kgCO2e / kWh, E grid is the power consumption of the grid, in kWh, β is the carbon emission coefficient of the diesel generator, in kgCO2e / L, E diesel is the diesel consumption, in L; γ is the carbon emission coefficient of water treatment, in kgCO2e / m 3 , Q water is the irrigation water consumption, in m3 , kCO2 is the water-carbon coupling correction factor;

[0023] Edge nodes cache the data of the last 24 hours and enable sliding average prediction when the network is disconnected.

[0024] Optionally, the blockchain evidence storage module adopts a dual-chain architecture of "local alliance chain + public chain";

[0025] The local chain uses Hyperledger Fabric 2.5 to store complete carbon footprint data and business logic, and designs a cross-chain bridge layer. It uses the Chainlink oracle network to achieve cross-chain anchoring of key event hash values. The public chain uses the Ethereum mainnet to store tamper-proof hashes and support the minting of carbon trading certificates.

[0026] Optionally, the rainwater collection unit includes two-stage filtration, a cyclone grit chamber to remove large particles, and a float filter to intercept suspended solids;

[0027] The water treatment unit includes a biofilm reactor and a capacitive deionization module;

[0028] The precision irrigation unit uses a magnetic levitation water pump connected to a pressure-compensating dripper and adopts a pulse irrigation strategy.

[0029] Optionally, the reinforcement learning model is constructed based on a proximal policy optimization algorithm, including the following steps:

[0030] S1: Design state space: S={W soil ,C grid ,P pv ,T air ,RH,W wind ,EC water}, where W soil is the volumetric moisture content of soil, C grid is the real-time carbon intensity of the grid, in kgCO2e / kWh, P pv is the photovoltaic array output, in kW, T air is the air temperature in °C, RH is the relative humidity in W wind is the wind speed in m / s, EC water is the electrical conductivity of irrigation water, in μS / cm;

[0031] The data fusion method used is spatiotemporal alignment, with a sliding window-based Kalman filter. The 3σ principle is used to remove outliers, and missing data is filled in by LSTM.

[0032] S2: Action Space and Execution Control:

[0033] A={t irrigate ,Q irrigate,μ energy}

[0034] where t irrigate The duration of a single irrigation is in the range of [1,30], in minutes. irrigate is the irrigation flow rate, the control range is [0.1, 2.0], and the unit is L / m 2 / min, μ energy is the energy mix ratio, the control range is [0,100], and the unit is the percentage of photovoltaic power;

[0035] The action constraints are:

[0036]

[0037] S3: Performs training strategy for proximal policy optimization.

[0038] Optionally, the training strategy for proximal strategy optimization includes:

[0039] S3.1: Course learning: Conduct phased training;

[0040] S3.2: Parallel simulation: Using NVIDIA Omniverse to run 128 virtual nursery environments simultaneously;

[0041] S3.3: Design the reward function, the formula is:

[0042] R=-0.6C CO2 +0.3G growth -0.1W water +10I constraint

[0043] Among them, CCO2 is the carbon emission penalty item, which represents the carbon dioxide equivalent emissions generated during system operation, with the unit of kgCO2e. Its data source includes electricity consumption: E grid ×α, diesel consumption: E diesel ×β, water treatment: Q water ×γ;

[0044] G growth This is a plant growth reward item, which represents a quantitative indicator of the seedling growth status. It requires a normalized value. The data sources include multispectral vegetation index, stem diameter increment, and leaf area index.

[0045] W water is the water resource consumption penalty item, which represents the amount of irrigation water, in m 3 ,The data source is the accumulated value of ultrasonic flow meter;

[0046] I constraintIconstraintIt is a constraint violation penalty item, indicating whether the system violates the preset constraints. It is triggered according to the following constraints: soil moisture exceeds the safe range, water pump is overloaded, and water quality exceeds the standard.

[0047] Optionally, the digital twin platform couples the WRF meteorological model with the HYDRUS-2D model and enables three-dimensional visualization;

[0048] The WRF meteorological model is downscaled and incorporated into the digital twin platform.

[0049] Optionally, the digital twin platform accesses WRF forecast data to simulate waterlogging risks of different drainage schemes.

[0050] Compared with the prior art, this application has at least the following beneficial effects:

[0051] 1. Full-cycle dynamic carbon footprint tracking: Based on real-time data collection from a multi-source sensor network, edge computing nodes dynamically calculate carbon emission intensity in a cycle of ≤50ms. The monitoring frequency is increased from once a day (traditional manual recording) to once a minute, improving the accuracy of identifying carbon emission fluctuations.

[0052] The blockchain dual-chain architecture enables tamper-proof data storage, cross-chain anchoring of key event hashes, improves data credibility, and supports third-party audits and carbon offset trading.

[0053] 2. Gradient water circulation and energy-saving irrigation: The three-stage water treatment makes the water utilization rate reach 92%, saving 58% of water compared with traditional flood irrigation. The energy consumption of CDI module is ≤0.8kWh / m 3 (65% lower than reverse osmosis);

[0054] The magnetic levitation water pump combined with the pulse irrigation strategy can reduce energy consumption per unit water volume and reduce evaporation losses.

[0055] 3. Intelligent Carbon-Water Co-Optimization: The reinforcement learning model (PPO algorithm) dynamically balances multiple objectives in state space and action space. The reward function ensures that emission reduction is prioritized, reducing the carbon footprint per seedling by 35% and improving seedling growth rate.

[0056] The digital twin platform couples the WRF meteorological model with the HYDRUS-2D soil model to simulate the carbon-water impact of different strategies within 48 hours and reduce the risk of waterlogging. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] To more intuitively illustrate the prior art and the present application, exemplary drawings are provided below. It should be understood that the specific shapes and structures shown in the drawings should not generally be considered as limiting conditions for implementing the present application; for example, based on the technical concepts disclosed in this application and the exemplary drawings, those skilled in the art are capable of easily making routine adjustments or further optimizations to the addition / reduction / attribution division, specific shapes, positional relationships, connection methods, dimensional ratios, etc. of certain units (components).

[0058] Figure 1 Schematic diagram of the module connection of the forestry seedling full-cycle carbon footprint tracking and energy-saving circulating irrigation system provided for this application;

[0059] Figure 2 Schematic diagram of the module connections of the data twin platform for full-cycle carbon footprint tracking of forestry seedling cultivation and energy-saving circulating irrigation system provided for this application. DETAILED DESCRIPTION

[0060] The present application will be further described below in detail through specific embodiments in conjunction with the accompanying drawings.

[0061] The present invention provides a forestry seedling cultivation full-cycle carbon footprint tracking and energy-saving circulating irrigation system, comprising:

[0062] Carbon footprint tracking subsystem, gradient water cycle subsystem, carbon-water synergistic optimization subsystem;

[0063] The carbon footprint tracking subsystem, which includes a multi-source sensor network module, an edge computing node module, and a blockchain evidence storage module, collects carbon emission data for the entire seedling cultivation cycle (seed treatment, substrate preparation, irrigation, and transportation) in real time. The multi-source sensor network module uses a heterogeneous sensor cluster to achieve three-dimensional collection of carbon footprint data for the entire seedling cultivation cycle.

[0064] Specifically, the multi-source sensor network module includes:

[0065] 1) Energy consumption monitoring layer: smart meters (RS485 interface) are used to collect three-phase power data (voltage, current, power factor) of each process equipment in real time using the Modbus-RTU protocol. Hall sensors are used to monitor the instantaneous fuel consumption of diesel generators. The conversion formula is:

[0066]

[0067] Among them, Q dieselis the total diesel consumption in liters (L), indicating the total fuel consumption of the diesel generator in the time period [t1, t2]. t1 and t2 are the start and end points of the time range in hours (h). k is the fuel consumption rate of the diesel generator in liters / (kilowatt-hour). k = 0.246 L / (kW-h), which is applicable to 50kW diesel generator sets. load (t) is the real-time load power in kilowatts (kW), which represents the output power of the diesel generator at time t. It changes dynamically over time and is usually collected in real time by an electric meter or sensor. To integrate over time, the instantaneous power Pload(t) is accumulated in the time period [t1, t2] to obtain the total energy (in kW·h), which is then multiplied by k to convert it into diesel consumption;

[0068] This formula is used to calculate diesel engine carbon emissions in the carbon footprint tracking subsystem of the forestry seedling system:

[0069] Combined with Q diesel And the diesel carbon emission factor (β=2.68kgCO2 / L), the carbon emissions can be calculated:

[0070] E diesel =Q diesel β

[0071] Ultimately, it will incorporate dynamic tracking of carbon footprint throughout the entire cycle to support carbon-water coordinated optimization decision-making.

[0072] Through this formula, the system can accurately quantify diesel engine carbon emissions, providing a data basis for energy-saving irrigation and blockchain evidence storage.

[0073] 2) Water consumption monitoring layer, using ultrasonic flowmeter (±1% accuracy), install time difference sensor in irrigation main pipe, with transmission frequency of 1MHz, and calculate flow rate by measuring the difference in ultrasonic propagation time between upstream and downstream:

[0074]

[0075] v represents the fluid velocity in meters per second (m / s), L is the length of the ultrasonic propagation path in meters (m), θ is the installation angle of the ultrasonic sensor in degrees (°) or radians (rad). The angle between the ultrasonic probe and the pipeline axis directly affects the measurement sensitivity. θ = 45° is usually selected (to balance sensitivity and signal attenuation). The angle error must be controlled within ±1°, otherwise the flow velocity measurement error will exceed 2%. down is the ultrasonic wave propagation time in the downstream direction, in seconds (s), t up The ultrasonic backflow propagation time is in seconds (s). It is used to convert the water flow rate in real time based on the pipe diameter parameter (DN50). The detection range is 0.3-5m. 3 / h.

[0076] 3) Biomass monitoring layer: A multispectral camera (400-1000nm band) equipped with a 5-channel filter (blue 450nm, green 560nm, red 650nm, red edge 720nm, near infrared 840nm) was used to obtain the NDVI index of the seedling canopy through drone inspections (flight altitude 30m, resolution 3cm / px):

[0077]

[0078] Among them, NDVI is called Normalized Difference Vegetation Index, which is an indicator that quantifies vegetation coverage and growth status, with a value range of -1 to 1;

[0079] >0.6: Dense and healthy vegetation

[0080] 0.2-0.5: Medium vegetation cover

[0081] <0: water or bare land

[0082] NIR stands for Near-Infrared Reflectance, with a wavelength range of 780-1300nm (840nm is commonly used for multispectral cameras). Its physical meaning is that healthy vegetation cell structures result in high near-infrared reflectance (reflectivity can reach 40-60%).

[0083] Red stands for Red Reflectance, with a wavelength range of 630-690nm (650nm is commonly used for multispectral cameras). Its physical meaning is that chlorophyll in vegetation strongly absorbs red light (reflectance is usually 5-15%).

[0084] The biomass carbon sink is inverted by combining a machine learning model (random forest algorithm) with an accuracy of ±0.2kgC / m.

[0085] 4) Environmental monitoring layer, using NDIR principle CO2 sensor, using dual-wavelength infrared light source (3.9μm / 4.2μm) and thermopile detector, calculates the concentration according to the Beer-Lambert law, the formula is:

[0086]

[0087] Where C is the gas concentration (such as CO2) in ppm (parts per million) or % (percentage), which represents the volume concentration of the target gas in the detection environment, and σ is the gas absorption cross section (or molar absorption coefficient), cm 2 / molecule (absorption cross section) or cm -1 (mol / L)-1 (molar absorption coefficient), L is the optical path length, in centimeters (cm), which represents the path length of infrared light through the measured gas, V ref The reference channel signal voltage is expressed in volts (V), which represents the voltage measured at a wavelength without target gas absorption (such as 3.9 μm). It is used to compensate for light source fluctuations and temperature drift. meas To measure the channel signal voltage, the unit is volt (V), which represents the voltage measured at the target gas absorption wavelength (such as 4.2μm), reflecting the degree of light intensity attenuation. ln is the natural logarithm (with e as the base), which means converting the exponential relationship of light intensity attenuation into a linear concentration value. The range is 0-5000ppm, and the temperature compensation accuracy is ±0.5℃.

[0088] The edge computing node module deploys a real-time carbon emission intensity calculation model (executed by the edge computing node), and the formula is:

[0089]

[0090] Among them, C intensity The real-time carbon emission intensity (CO2 equivalent emissions per unit time or unit output), in kgCO2e / h or kgCO2e / m 3 , α is the dynamic coefficient of grid carbon intensity (reflecting the carbon emission factor of the real-time power generation structure of the grid, obtained from the regional grid operator API, and needs to be updated every minute), the unit is kgCO2e / kWh, E grid is the power consumption of the grid, in kWh, which is derived from real-time monitoring of smart meters (RS485 / MODBUS protocol), β is the carbon emission coefficient of the diesel generator, in kgCO2e / L, E diesel is the diesel consumption, in L; γ is the carbon emission coefficient of water treatment, in kgCO2e / m 3 , dynamically adjusted according to the water cycle level (edge ​​computing calibration), Q water is the irrigation water consumption, in m 3 , kCO2 is the water-carbon coupling correction factor;

[0091] Specifically, the dynamic coefficient α is realized by obtaining real-time data from the State Grid carbon emission factor publishing platform (such as the China Carbon Market Network) every 5 minutes through the HTTP protocol. The calculation formula is:

[0092] α=0.3·Coal-fired power share + 0.1·Gas-fired power share + 0.001·Green power shareα=0.3·Coal-fired power share + 0.1·Gas-fired power share + 0.001·Green power share

[0093] Edge nodes cache the data of the last 24 hours and enable sliding average prediction when the network is disconnected.

[0094] Furthermore, edge computing is optimized and quantified for deployment: floating-point operations are converted to INT8 fixed-point numbers, and memory usage is compressed from 32KB to 8KB; real-time performance is guaranteed: the measured calculation cycle on STM32L4 is ≤50ms (main frequency 80MHz).

[0095] The above formula incorporates the carbon intensity of the power grid (α), diesel backup power (β), and water treatment process (γ·kCO2) into a unified calculation framework to achieve dynamic coupling. Each parameter corresponds to physical entity monitoring data, supporting carbon footprint traceability audits.

[0096] The formula is designed to match the computing power characteristics of the STM32L4 chip (single-precision floating-point unit + hardware multiplier).

[0097] The blockchain evidence storage module adopts a dual-chain architecture of "local consortium chain + public chain", taking into account data privacy, audit transparency, and cross-chain trusted verification. Specifically, the local chain uses Hyperledger Fabric 2.5 to store complete carbon footprint data and business logic, and designs a cross-chain bridge layer using the Chainlink oracle network to achieve cross-chain anchoring of key event hash values. The public chain uses the Ethereum mainnet to store tamper-proof hashes and support the minting of carbon trading certificates.

[0098] Dual-chain division of labor architecture:

[0099] Hyperledger Fabric: Ensuring enterprise data privacy and business agility.

[0100] Ethereum: Leveraging its immutability to achieve trusted transfer of carbon assets.

[0101] Decentralized cross-chain: Replacing traditional centralized bridges with the Chainlink oracle network eliminates single point of failure risks.

[0102] Gradient water circulation subsystem, which has a rainwater collection unit with two-stage filtration, a cyclone grit chamber (1.2m diameter) to remove large particles, and a float filter (200 mesh) to intercept suspended solids;

[0103] Water treatment unit: including biofilm reactor (filler specific surface area 800m 2 / m 3 ) and capacitive deionization (CDI) modules (operating voltage 1.2V, salt rejection rate >85%);

[0104] Precision irrigation unit: A magnetic levitation water pump (rated power 5.5kW) connected to a pressure-compensating dripper (working pressure 0.1-0.3MPa), using a pulsed irrigation strategy;

[0105] Specifically, the main body of the cyclone sand settling tank is cylindrical with a bottom cone angle of 60°. It is made of fiberglass reinforced plastic (FRP) and needs to be designed for corrosion resistance. The working principle is: the water inlet is tangentially introduced to the rainwater, and the centrifugal force is used to separate the density >2.65g / cm 3 Sand particles (particle size> 200μm). Design flow rate 15m 3 / h, sedimentation efficiency>90%, the top overflow port discharges light suspended solids, which can reduce subsequent pumping energy consumption (18% energy saving compared to traditional sand filter), and no chemical agents are added;

[0106] The float filter has a filter assembly including an outer layer, which uses a 200-mesh stainless steel filter (pore size 75μm) to intercept suspended solids (SS≤50mg / L);

[0107] The inner layer adopts a floating ball valve group (density 0.95g / cm 3 ), automatically adjust the water flow as the water level rises and falls;

[0108] The float filter has a backwash mechanism, which uses a pressure differential sensor (range 0-0.6MPa) to monitor filter blockage and triggers a combined air-water backwash (air pressure 0.3MPa, water intensity 8L / (m 2 ·s));

[0109] In the water treatment unit, the filler structure of the biofilm reactor uses polyethylene (PE) porous suspended filler with a pore size of 5mm and a specific surface area of ​​800m 2 / m 3 , filling rate 70%, microbial film thickness controlled at 200-500μm, composed of nitrifying bacteria (Nitrosomonas) and denitrifying bacteria (Pseudomonas), can replace the traditional chemical coagulation process, reduce carbon emissions from chemical production (carbon emissions per ton of water treated are reduced by 0.8kgCO2e), specific operating parameters are hydraulic retention time (HRT): 6 hours, dissolved oxygen (DO) graded control (2-3mg / L in aerobic zone, 0.2-0.5mg / L in anoxic zone), denitrification efficiency: ammonia nitrogen removal rate>95%, total nitrogen (TN) removal rate>85%, carbon source dosage (sodium acetate) is reduced by 40% compared with the activated sludge method.

[0110] The electrodes of the capacitive deionization (CDI) module are activated carbon-graphene composite electrodes (specific surface area 2100m 2 / g), coated with Nafion ion exchange membrane, with a channel spacing of 1.5mm, and a cross-flow design to improve ion adsorption efficiency. A 1.2V DC voltage was applied during the desalination process. Na + , Cl- is adsorbed to the electrode surface under the action of the electric field, the desalination rate is >85%, and the energy consumption is ≤0.8kWh / m 3(65% lower than reverse osmosis), in regeneration mode, the voltage is reversed (-0.5V) to release the adsorbed ions, and the concentrated water discharge is <5%. The overall adsorption / regeneration cycle is dynamically adjusted based on the feedback from the online conductivity meter (range 0-2000μS / cm).

[0111] The core component of the magnetic levitation water pump system in the precision irrigation unit is a permanent magnet synchronous motor (IE5 energy efficiency rating) with a rated power of 5.5kW, a continuously variable speed range of 0-3000r / min, an active magnetic bearing (AMB), a radial clearance of 0.2mm, and a displacement sensor (resolution of 1μm) to monitor the rotor position in real time.

[0112] The pressure-compensating dripper utilizes a silicone diaphragm (0.5mm thick) and a spring (elastic coefficient 2N / mm) to form a pressure-compensating mechanism. The flow channel adopts a maze design (path length 15cm), a turbulence coefficient ≤ 0.3, and an anti-clogging rating of IP68. The irrigation strategy employed is pulse control, with a 3-minute on-time (flow rate 2L / h per dripper) and 7-minute off-time cycle. This reduces evaporation losses by 23% compared to continuous irrigation. Zone management is also implemented, with four irrigation zones divided according to the seedling growth period, and independently controlled by solenoid valves (24VDC, response time <1s).

[0113] At the same time, the precision irrigation unit uses an IoT gateway to receive data from soil moisture sensors (FDR principle, accuracy ±2%) and weather stations, and issues irrigation instructions through the MQTT protocol.

[0114] The carbon-water collaborative optimization subsystem deploys a reinforcement learning model and a digital twin platform. The reinforcement learning model balances carbon emissions, water conservation, and seedling growth rates, dynamically generating the optimal irrigation strategy. The digital twin platform enables real-time interaction and visualization, and assists the reinforcement learning model in making decisions.

[0115] The reinforcement learning model is built based on the proximal policy optimization algorithm, which includes the following steps:

[0116] S1: Design state space: S={W soil ,C grid ,P pv ,T air ,RH,W wind ,EC water}, where W soil is the soil volume moisture content (%), the data comes from the time domain reflectometer (TDR), C grid is the real-time carbon intensity of the power grid, in kgCO2e / kWh, the data comes from the regional power grid API, P pv The output of the photovoltaic array in kW. The data comes from the RS485 interface of the inverter. airis the air temperature in °C, the data comes from the BME280 sensor, RH is the relative humidity (%), the data comes from the BME280 sensor, W wind is the wind speed in m / s, the data comes from the ultrasonic anemometer, EC water is the electrical conductivity of irrigation water in μS / cm, the data comes from Atlas Scientific EC sensor;

[0117] The data fusion method used is spatiotemporal alignment, with a sliding window-based Kalman filter (window length = 15 min, step length = 1 min). The 3σ principle is used to remove outliers, and missing data are filled by LSTM.

[0118] S2: Action Space and Execution Control, A={t irrigate ,Q irrigate ,μ energy}, where t irrigate is the duration of a single irrigation, the control range is [1,30], the unit is minutes, the magnetic levitation water pump inverter is selected as the actuator, Q irrigate is the irrigation flow rate, the control range is [0.1, 2.0], and the unit is L / m 2 / min, piezoelectric ceramic valve is selected as the actuator, μ energy is the energy mix ratio, the control range is [0,100], the unit is the photovoltaic ratio, and a bidirectional converter is selected as the actuator;

[0119] Action constraints:

[0120]

[0121] S3: The training strategy of Proximal Policy Optimization (PPO) includes:

[0122] S3.1: Course learning: Conduct phased training (basic scenario → extreme weather → equipment failure);

[0123] S3.2: Parallel simulation: Using NVIDIA Omniverse to run 128 virtual nursery environments simultaneously;

[0124] S3.3: Design the reward function, the formula is:

[0125] R=-0.6C CO2 +0.3G growth -0.1W water +10I constraint

[0126] Among them, CCO2 is the carbon emission penalty item, which represents the carbon dioxide equivalent emissions generated during system operation, with the unit of kgCO2e. Its data source includes electricity consumption: E grid ×α, diesel consumption: E diesel ×β, water treatment: Q water ×γ;

[0127] G growth This is a plant growth reward item, which represents a quantitative indicator of the seedling growth status. It needs to be normalized (0-1). The data sources include multispectral vegetation index (NDVI / NDRE), stem diameter increment (mm / day), and leaf area index (LAI);

[0128] W water is the water resource consumption penalty item, which represents the amount of irrigation water, in m 3 ,The data source is the accumulated value of ultrasonic flow meter;

[0129] I constraintIconstraint is the constraint violation penalty term (0 / 1 variable), indicating whether the system violates the preset constraints (0 = not violated, 1 = violated), and is triggered according to the following constraints: soil moisture content exceeds the safe range (<40% or >85%), water pump overload operation (power > rated value), water quality exceeds the standard (EC > 2000 μS / cm);

[0130] Carbon emissions are the core optimization objective, and negative weights indicate that reducing carbon emissions significantly increases rewards. The coefficient has the largest absolute value (0.6), reflecting the highest priority for emission reduction. Promoting seedling growth is the production objective, but its weight is relatively low (0.3), indicating that growth should be moderately promoted under carbon-water constraints rather than blindly pursuing high yields. Water conservation is less important than emission reduction, with a weight of 0.1. Overirrigation is prevented but necessary water use is not strictly restricted. Constraint violation penalties have a very high weight (10), forcing the model to prioritize the basic conditions of system safety and plant survival.

[0131] Digital twin platform: Coupling the WRF meteorological model with the HYDRUS-2D model, achieving a spatiotemporal resolution of 0.5 km × 0.5 km × 15 minutes and enabling 3D visualization;

[0132] The WRF meteorological model needs to be downscaled and incorporated into the digital twin platform. A dynamical-statistical hybrid downscaling method is used: 1. WRF outputs 0.5 km coarse grid data → 2. A 50 m microscale field is generated based on a convolutional neural network (U-Net) → 3. Terrain correction (slope / vegetation cover) is performed, transfer learning is introduced to optimize the soil model, and the HYDRUS-2D Richards equation solver is adapted to the GPU cluster.

[0133] The technical features of the above embodiments can be combined arbitrarily (as long as there is no contradiction in the combination of these technical features). In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described; these embodiments that are not explicitly written should also be considered to be within the scope of this specification.

Claims

1. Forestry seedling cultivation full cycle carbon footprint tracking and energy-saving circulating irrigation system, characterized by: include: The carbon footprint tracking subsystem, which includes a multi-source sensor network module, an edge computing node module, and a blockchain evidence storage module, collects carbon emission data throughout the entire seedling cultivation cycle in real time. The multi-source sensor network module uses a heterogeneous sensor cluster to achieve three-dimensional collection of carbon footprint data throughout the entire seedling cultivation cycle. Gradient water circulation subsystem, which has rainwater collection unit, water treatment unit and precision irrigation unit to achieve graded water resource treatment; The carbon-water collaborative optimization subsystem deploys a reinforcement learning model and a digital twin platform. The reinforcement learning model balances carbon emissions, water conservation, and seedling growth rate, dynamically generating the optimal irrigation strategy. The digital twin platform enables real-time interaction and visualization, and assists the reinforcement learning model in decision-making.

2. The forestry seedling cultivation full cycle carbon footprint tracking and energy-saving circulating irrigation system according to claim 1 is characterized in that: The multi-source sensor network module includes: 1) Energy consumption monitoring layer: smart meters are used to collect three-phase power data of each process equipment in real time, and Hall sensors are used to monitor the instantaneous fuel consumption of diesel generators; 2) Water consumption monitoring layer: ultrasonic flowmeters are used. Time difference sensors are installed in the main irrigation pipes with a transmission frequency of 1 MHz. The flow rate is calculated by measuring the difference in ultrasonic propagation time between upstream and downstream. 3) Biomass monitoring layer: a multispectral camera equipped with a 5-channel filter is used to obtain the NDVI index of the seedling canopy through drone inspections, and the biomass carbon sequestration is inverted using the random forest algorithm; 4) Environmental monitoring layer, using NDIR principle CO2 sensor, dual-wavelength infrared light source and thermopile detector, calculates the concentration according to Beer-Lambert law.

3. The forestry seedling cultivation full cycle carbon footprint tracking and energy-saving circulating irrigation system according to claim 1 is characterized in that: The edge computing node module is deployed with a real-time calculation model for carbon emission intensity, and the formula is: Among them, C intensity Real-time carbon emission intensity, unit is kgCO2e / h or kgCO2e / m 3 , α is the dynamic coefficient of carbon intensity of the power grid, the unit is kgCO2e / kWh, E grid is the power consumption of the grid, in kWh, β is the carbon emission coefficient of the diesel generator, in kgCO2e / L, E diesel is the diesel consumption, in L; γ is the carbon emission coefficient of water treatment, in kgCO2e / m 3 , Q water is the irrigation water consumption, in m 3 , kCO2 is the water-carbon coupling correction factor; Edge nodes cache the data of the last 24 hours and enable sliding average prediction when the network is disconnected.

4. The forestry seedling cultivation full cycle carbon footprint tracking and energy-saving circulating irrigation system according to claim 1 is characterized in that: The blockchain evidence storage module adopts a dual-chain architecture of "local alliance chain + public chain"; The local chain uses Hyperledger Fabric 2.5 to store complete carbon footprint data and business logic, and designs a cross-chain bridge layer. It uses the Chainlink oracle network to achieve cross-chain anchoring of key event hash values. The public chain uses the Ethereum mainnet to store tamper-proof hashes and support the minting of carbon trading certificates.

5. The forestry seedling cultivation full cycle carbon footprint tracking and energy-saving circulating irrigation system according to claim 1 is characterized in that: The rainwater collection unit includes two-stage filtration: a cyclone grit chamber removes large particles, and a float filter intercepts suspended solids; The water treatment unit includes a biofilm reactor and a capacitive deionization module; The precision irrigation unit uses a magnetic levitation water pump connected to a pressure-compensating dripper and adopts a pulse irrigation strategy.

6. The forestry seedling cultivation full cycle carbon footprint tracking and energy-saving circulating irrigation system according to claim 1 is characterized in that: The reinforcement learning model is built based on the proximal policy optimization algorithm, which includes the following steps: S1: Design state space: S={W soil ,C grid ,P pv ,T air ,RH,W wind ,EC water }, where W soil is the volumetric moisture content of soil, C grid is the real-time carbon intensity of the power grid, in kgCO2e / kWh, P pv is the photovoltaic array output, in kW, T air is the air temperature in °C, RH is the relative humidity in W wind is the wind speed in m / s, EC water is the electrical conductivity of irrigation water, in μS / cm; The data fusion method used is spatiotemporal alignment, with a sliding window-based Kalman filter. The 3σ principle is used to remove outliers, and missing data is filled in by LSTM. S2: Action Space and Execution Control: A={t irrigate ,Q irrigate ,μ energy } where t irrigate The duration of a single irrigation is in the range of [1,30], in minutes. irrigate is the irrigation flow rate, the control range is [0.1, 2.0], and the unit is L / m 2 / min, μ energy is the energy mix ratio, the control range is [0,100], and the unit is the percentage of photovoltaic power; The action constraints are: S3: Performs training strategy for proximal policy optimization.

7. The forestry seedling cultivation full cycle carbon footprint tracking and energy-saving circulating irrigation system according to claim 6 is characterized in that: The training strategies for proximal policy optimization include: S3.1: Course learning: Conduct phased training; S3.2: Parallel simulation: Using NVIDIA Omniverse to run 128 virtual nursery environments simultaneously; S3.3: Design the reward function, the formula is: R=-0.6C CO2 +0.3G growth -0.1W water +10I constraint Among them, CCO2 is the carbon emission penalty item, which represents the carbon dioxide equivalent emissions generated during system operation, with the unit of kgCO2e. Its data source includes electricity consumption: E grid ×α, diesel consumption: E diesel ×β, water treatment: Q water ×γ; G growth This is a plant growth reward item, which represents a quantitative indicator of the seedling growth status. It requires a normalized value. The data sources include multispectral vegetation index, stem diameter increment, and leaf area index. W water is the water resource consumption penalty item, which represents the amount of irrigation water, in m 3 ,The data source is the accumulated value of ultrasonic flow meter; I constraintIconstraint It is a constraint violation penalty item, indicating whether the system violates the preset constraints. It is triggered according to the following constraints: soil moisture exceeds the safe range, water pump is overloaded, and water quality exceeds the standard.

8. The forestry seedling cultivation full cycle carbon footprint tracking and energy-saving circulating irrigation system according to claim 7 is characterized in that: The digital twin platform couples the WRF meteorological model with the HYDRUS-2D model and enables three-dimensional visualization; The WRF meteorological model is downscaled and incorporated into the digital twin platform.

9. The forestry seedling cultivation full cycle carbon footprint tracking and energy-saving circulating irrigation system according to claim 8, characterized in that: The digital twin platform is connected to WRF forecast data to simulate the waterlogging risks of different drainage schemes.

Images