Broussonetia papyrifera circular agriculture whole industry chain ecological closed-loop system

Abstract

The invention belongs to the technical field of intelligent agricultural control, and relates to a paper mulberry circular agriculture whole industry chain ecological closed-loop system which solves the technical problems of low resource allocation efficiency, unstable product quality and low waste resource utilization rate caused by independent operation of each link of a paper mulberry industry chain and lack of collaborative optimization control. A four-layer intelligent control framework is adopted and comprises a data acquisition layer, an edge calculation layer, an optimization decision-making layer and an execution control layer. A soil sensor, a meteorological sensor, a plant physiological sensor and a product quality sensor are used for collecting whole industrial chain data, and an edge computing node is used for filtering and feature extraction preprocessing; a growth-quality prediction correlation model and a multi-target collaborative optimization algorithm are constructed, and distributed control nodes drive an execution mechanism through a communication protocol to realize full-industrial-chain automatic control of paper mulberry planting, harvesting, processing, waste treatment and resource recycling.

Description

Technical Field

[0001] The present invention relates to the field of intelligent agricultural control technology, and in particular to an ecological closed-loop system for the entire industrial chain of paper mulberry circular agriculture based on a multi-layer distributed architecture and a control method thereof. Background Art

[0002] As modern agriculture develops towards intelligent and precise control, agricultural Internet of Things (IoT) technologies and intelligent control systems have been widely used. A variety of agricultural intelligent control systems have emerged in existing technologies, primarily focusing on monitoring and controlling a single link.

[0003] CN116700405A discloses an intelligent agricultural control system, comprising an intelligent greenhouse system, an intelligent fishpond system, a master control center, and a cloud platform. A data monitoring module monitors parameters such as temperature and humidity, soil moisture, ambient light, and CO2 concentration, while an automatic control module controls intelligent devices within the greenhouse. The system enables automated monitoring of both the greenhouse and the fishpond, reducing user workload and management costs.

[0004] CN103605353B discloses a cloud-based intelligent optimization and monitoring system for agriculture. This system utilizes multiple user units and a cloud computing central server architecture. By presetting multiple sets of environmental parameters, users can obtain the optimal preset environmental parameters. The system then collects environmental parameter information in real time and controls the controlled equipment to maintain optimal conditions. This system enables data sharing and optimal execution, promoting the refined and intelligent development of agriculture.

[0005] CN116339200A discloses an edge computing-based smart agricultural greenhouse system. The system utilizes sensors, edge computing modules, a cloud platform, and an actuator architecture. The cloud platform includes multiple policy models, and the edge computing module obtains control instructions based on sensor data and the optimal policy model. This system enables scientific and large-scale crop management and rapidly responds to water, fertilizer, and pesticide needs.

[0006] After searching and analyzing, the existing technology mainly has the following deficiencies:

[0007] 1. Limited scope of application: Existing systems are mostly limited to a single link such as greenhouse cultivation or fish pond farming, and lack systematic control over the entire industry chain of paper mulberry planting, harvesting, processing, waste treatment and resource recycling.

[0008] 2. Lack of cross-link coordination: Existing technologies mainly optimize a single process link and cannot achieve the correlation control between planting period parameters and subsequent processing quality, and lack the ability of global collaborative optimization.

[0009] 3. Single control strategy: Existing systems mostly adopt real-time response control or preset parameter control, lack of forward-looking control mechanism based on prediction, and cannot adapt to the time delay characteristics of biological systems.

[0010] 4. Lack of recycling: Existing technologies do not involve intelligent control of waste resource utilization and cannot form a true ecological closed-loop system. Summary of the Invention

[0011] Technical issues

[0012] The main technical problems with existing technologies are that each link in the paper mulberry circular agriculture industry chain operates independently, lacking coordinated optimization and control. This leads to inefficient resource allocation, unstable product quality, and low waste resource utilization. Specifically, resource utilization efficiency is less than 60%, the coefficient of variation in product quality is greater than 18%, and the waste resource utilization rate is less than 35%. This makes it impossible to achieve multi-objective coordinated optimization across time and across links.

[0013] Technical Solution

[0014] In order to solve the above technical problems, the present invention provides an ecological closed-loop system for the entire industrial chain of paper mulberry circular agriculture, comprising:

[0015] The data collection layer includes soil sensors, meteorological sensors, plant physiological sensors and product quality sensors. The soil sensors include soil temperature sensors, soil moisture sensors and soil pH sensors, with detection accuracy of

[0016] ±0.5℃, ±3%RH, ±0.2pH; meteorological sensors include light intensity sensor and temperature and humidity sensor, with detection accuracy of ±10μmol / (m 2 ·s), ±0.5℃, ±5%RH; the plant physiological sensor detects the protein content and fiber content of paper mulberry leaves with detection accuracy of ±0.5% and ±1.0%, respectively; the product quality sensor detects the moisture content and crude protein content of processed products with detection accuracy of ±1.0% and ±0.5%, respectively.

[0017] The edge computing layer, including multiple edge computing nodes, uses filtering algorithms and feature extraction algorithms to preprocess the data collected by the data acquisition layer. The edge computing nodes use a moving average filtering algorithm for noise filtering with a window length of 10 sampling points. The principal component analysis method is used for feature extraction, and the principal components with a cumulative contribution rate greater than 85% are retained.

[0018] The optimization decision layer includes a paper mulberry growth-quality prediction association model module and a paper mulberry full-chain collaborative optimization objective function module. The paper mulberry growth-quality prediction association model module establishes a machine learning regression relationship based on the protein content and fiber content of paper mulberry leaves and the nitrogen content, phosphorus content, potassium content, organic matter content, light intensity, temperature, humidity, and pH value of the soil during the planting period. The regression equation is:

[0019] Y=RF(X1,X2,X3,X4,X5,X6,X7,X8),

[0020] Where Y is the protein content, X1 is the soil nitrogen content, X2 is the soil phosphorus content, X3 is the soil potassium content, X4 is the soil organic matter content, X5 is the light intensity, X6 is the temperature, X7 is the humidity, X8 is the soil pH value, and the model determination coefficient R 2 Greater than 0.90; the paper mulberry growth-quality prediction correlation model module adopts multi-time scale prediction control, including a short-term control module and a long-term optimization module. The control cycle of the short-term control module is 30 to 60 minutes, preferably 45 minutes. The Kalman filter algorithm is used to process the sensor data to improve the control accuracy and system stability. The prediction cycle of the long-term optimization module is 7 to 14 days, preferably 10 days; the paper mulberry full chain collaborative optimization objective function module adopts a multi-objective collaborative optimization function

[0021] F(x)=w1·G(x)+w2·P(x)+w3·R(x)+w4·E(x)

[0022] Perform collaborative optimization, where G(x) is the growth quality function, P(x) is the processing efficiency function, R(x) is the resource utilization function, E(x) is the environmental impact function, and the constraint condition is the resource consumption upper limit R max ≤120kg / mu, environmental impact threshold E max ≤0.3kgCO2 equivalent / kg product, solved by multi-objective particle swarm optimization algorithm, with a population size of 100 and a maximum number of iterations of 300 times. The paper mulberry full-chain collaborative optimization objective function module includes an adaptive weight adjustment mechanism, which dynamically adjusts the weight coefficients w1, w2, w3, and w4 according to the paper mulberry growth stage and environmental conditions, and triggers weight adjustment when the growth stage changes or the product quality variation coefficient is greater than 15%.

[0023] The execution control layer includes distributed control nodes and executive agencies. The distributed control nodes transmit control signals through communication protocols (such as CAN bus protocols) according to the control instructions of the optimization decision-making layer, and drive the executive agencies to realize automatic control of the entire industry chain of paper mulberry planting, harvesting, processing, waste treatment and resource recycling. The execution control layer includes redundant control nodes. When the main control node fails, the redundant control node can automatically take over the control task within a preset time (such as within 10 minutes) to ensure continuous and stable operation of the system. The fault detection adopts a heartbeat detection mechanism. The heartbeat detection mechanism adopts a periodic status signal sending method with a detection period of 60 seconds. When the heartbeat detection fails for three consecutive times, the node is determined to be faulty and the redundant node is automatically triggered to take over the control task.

[0024] The edge computing layer works in conjunction with the cloud computing platform. The edge computing nodes process real-time control tasks, and the cloud computing platform performs model training and parameter optimization. The system monitors the communication distance and data packet size in real time. When the communication distance is less than 3km and the data volume is less than 500KB, the LoRa network is given priority. When the communication distance is greater than 3km or the data volume is greater than 500KB, it automatically switches to the 4G network to ensure the reliability and efficiency of data transmission.

[0025] The present invention also provides a collaborative optimization control method for the entire mulberry chain based on the above-mentioned system, which realizes intelligent collaborative control of the entire mulberry industry chain through a closed-loop control process of data collection, edge computing processing, predictive modeling, multi-objective optimization and execution control.

[0026] Beneficial effects

[0027] The present invention has the following beneficial effects compared to the prior art:

[0028] 1. Through the four-layer intelligent control architecture of the entire chain of mulberry and cross-link collaborative optimization, the resource utilization efficiency has been increased from 60% of the existing technology to 71.5%, an increase of 19.2%.

[0029] 2. Through the paper mulberry growth-quality prediction correlation model and multi-time scale collaborative control strategy, the product quality variation coefficient was reduced from 18% of the existing technology to 11.8%, and the stability was improved by 34.4%.

[0030] 3. Through the intelligent control of the intelligent resource processing system for paper mulberry waste, the waste utilization rate has been increased from 35% of the existing technology to 66.2%, an increase of 89.1%.

[0031] 4. Through redundant control nodes and adaptive weight adjustment mechanism, the system availability reached 91.8%, the prediction accuracy reached 84%, and the system reliability and intelligence level were significantly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 This is a schematic diagram of the structure of the four-layer intelligent control architecture system of the paper mulberry full chain according to an embodiment of the present invention;

[0033] Figure 2 This is a flow chart of the paper mulberry dual-time domain collaborative control method according to an embodiment of the present invention;

[0034] Figure 3 It is a structural diagram of the intelligent resource processing system for paper mulberry waste according to an embodiment of the present invention. DETAILED DESCRIPTION

[0035] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0036] Example 1

[0037] The purpose of this example is to verify the overall technical solution and collaborative optimization effect of the four-layer intelligent control architecture system for the whole chain of paper mulberry. The experiment was conducted in a 50-acre paper mulberry planting base in Nanjing, Jiangsu Province, with yellow-brown soil.

[0038] like Figure 1 As shown, the four-layer intelligent control architecture system of the whole chain of mulberry tree includes data collection layer 1, edge computing layer 2, optimization decision layer 3 and execution control layer 4.

[0039] The data acquisition layer 1 includes a soil sensor 11, a meteorological sensor 12, a plant physiological sensor 13, and a product quality sensor 14. The soil sensor 11 includes a soil temperature sensor, a soil moisture sensor, and a soil pH sensor, which respectively use a DS18B20 temperature sensor (detection accuracy ±0.5°C, measurement range -55 to +125°C), an SHT30 humidity sensor (detection accuracy ±3%RH, measurement range 0 to 100%RH), and a PH-4502C pH sensor (detection accuracy ±0.2pH, measurement range 0 to 14pH). The meteorological sensor 12 includes a light intensity sensor BH1750 (detection accuracy ±10μmol / (m 2 ·s), measuring range 1~65535lx) and temperature and humidity sensor SHT35 (temperature detection accuracy ±0.5℃, humidity detection accuracy ±5%RH). Plant physiological sensor 13 uses portable chlorophyll meter SPAD-502Plus combined with near-infrared spectroscopy technology to detect the protein content and fiber content of paper mulberry leaves. The quantitative relationship between the spectrum and the component content is established through partial least squares regression, with detection accuracy of ±0.5% and ±1.0%, respectively. Product quality sensor 14 uses moisture meter MA35 (detection accuracy ±1.0%, measuring range 0.01~100%) and protein analyzer Kjeltec 8400 (detection accuracy ±0.5%, based on Kjeldahl nitrogen determination method) to detect the moisture content and crude protein content of processed products.

[0040] Edge computing layer 2 includes multiple edge computing nodes 21, each equipped with an ARM Cortex-A53 quad-core processor (1.2GHz), 2GB of LPDDR3 memory, and 16GB of eMMC storage. Edge computing node 21 uses a moving average filter algorithm for noise filtering, with a window length of 10 sampling points. The filtering formula is:

[0041]

[0042] Where N = 10. The principal component analysis method is used to extract features, and the covariance matrix is ​​calculated.

[0043] The eigenvalue λi and the eigenvector v i , retain the cumulative contribution rate ∑λ i / ∑λ total Greater than 85% of the principal components achieve data dimensionality reduction and feature extraction.

[0044] The optimization decision layer 3 includes a paper mulberry growth-quality prediction association model module 31 and a paper mulberry full-chain collaborative optimization objective function module 32. The paper mulberry growth-quality prediction association model module 31 establishes a random forest regression relationship based on the protein content Y of paper mulberry leaves and the soil nitrogen content X1, phosphorus content X2, potassium content X3, organic matter content X4, light intensity X5, temperature X6, humidity X7, and pH value X8:

[0045] Y=RF(X1,X2,X3,X4,X5,X6,X7,X8),

[0046] 100 decision trees are used with a maximum depth of 10 and a model determination coefficient R 2 =0.92, and the model parameters were determined by 10-fold cross validation. The paper mulberry full chain collaborative optimization objective function module 32 adopts the multi-objective collaborative optimization function

[0047] F(x)=w1·G(x)+w2·P(x)+w3·R(x)+w4·E(x),

[0048] The growth quality function is:

[0049] G(x) = α1·protein content + α2·fiber content (α1 = 0.6, α2 = 0.4),

[0050] Processing efficiency function:

[0051] P(x)=β1·Processing time -1 +β2·Product qualification rate (β1=0.3,β2=0.7),

[0052] Resource utilization function:

[0053] R(x) = γ1·water use efficiency + γ2·nutrient use efficiency (γ1 = 0.5, γ2 = 0.5),

[0054] Environmental impact function:

[0055] E(x)=δ1·Carbon emissions -1 +δ2·waste utilization rate (δ1=0.4, δ2=0.6).

[0056] The execution control layer 4 includes a distributed control node 41 and an actuator 42. The distributed control node 41 uses a Siemens S7-1200 PLC controller, equipped with a CPU 1214C processor (processing speed 0.1ms / 1000 instructions), 14 digital input / output ports and 2 analog input ports (resolution 12 bits). The control node transmits control signals via the CAN bus protocol (baud rate 125kbps, data frame format compliant with ISO 11898 standard). The actuator 42 includes an irrigation system (solenoid valve control, flow range 0.5~3m 3 / h), fertilization system (peristaltic pump driven, flow accuracy ±5%), temperature regulation system (heating power 3kW, cooling power 2kW), processing equipment control system and waste treatment equipment.

[0057] During system operation, the data acquisition layer 1 collects various parameter data every 5 minutes. This data is uploaded to the edge computing layer 2 via a LoRa wireless transmission module (frequency 433MHz, transmission range 3km, power consumption <80mA). The edge computing layer 2 performs real-time preprocessing and feature extraction on the collected data, with a single data processing time of less than 15 seconds. The optimization decision layer 3 performs predictive modeling and multi-objective optimization based on the preprocessed data. It uses a multi-objective particle swarm optimization algorithm with a population size of 100, 300 iterations, an inertia weight of 0.9, a learning factor c1=c2=2.0, a convergence accuracy of 0.01, and an optimization calculation time of less than 180 seconds. The execution control layer 4 generates control instructions based on the optimization results. The control signal transmission delay is less than 200ms, driving the actuators for automatic control.

[0058] Verified by this example, the system resource utilization efficiency reached 71.5%, the product quality variation coefficient dropped to 11.8%, the waste utilization rate reached 66.2%, the system response time was less than 5 minutes, and the prediction accuracy reached 82%, which proved the effectiveness of the four-layer intelligent control architecture of the entire chain of paper mulberry and the superiority of collaborative optimization.

[0059] Example 2

[0060] The purpose of this example is to verify the technical effect of the multi-time scale coordinated control strategy and the optimization performance of the adaptive weight adjustment mechanism. 800 data points are used for model training and verification.

[0061] like Figure 2As shown, the paper mulberry dual-time domain collaborative control method includes collaborative control processes of two time scales: short-term control and long-term optimization. On the basis of Example 1, the paper mulberry growth-quality prediction association model module 31 adopts multi-time scale predictive control, including a short-term control module 311 and a long-term optimization module 312. The control period of the short-term control module 311 is set to 45 minutes, and Kalman filtering is used for processing. The process noise variance Q is set to 0.05 (based on the statistical variance of soil temperature and humidity changes), and the observation noise variance R is set to 0.25 (based on sensor accuracy specifications). The Kalman filter state equation is: x(k+1)=Ax(k)+Bu(k)+w(k), and the observation equation is: z(k)=Hx(k)+v(k), where the state transfer matrix Control Matrix The observation matrix H = [1 0], w(k) and v(k) are process noise and observation noise respectively.

[0062] The prediction period of the long-term optimization module 312 is set to 10 days, and predictions are made based on a random forest model trained on historical data. The random forest model uses 80 decision trees with a maximum depth of 8 and a minimum number of split samples of 5. The input features include 12 environmental parameters such as soil temperature, humidity, pH, nitrogen, phosphorus and potassium content, organic matter content, light intensity, air temperature and humidity, and rainfall. The output prediction indicators include four indicators such as protein content, fiber content, biomass, and quality index. Training uses grid search to optimize hyperparameters, and 5-fold cross-validation to evaluate model performance. The final model has an R score of 0.001 on the test set. 2 Reached 0.85.

[0063] The objective function module 32 of the collaborative optimization of the entire paper mulberry chain includes an adaptive weight adjustment mechanism. In the first 45 days of the paper mulberry growth period, the weight coefficients are set to w1=0.5, w2=0.2, w3=0.15, w4=0.15, focusing on growth quality. 45 days after the maturity period, the weight coefficients are adjusted to w1=0.2, w2=0.5, w3=0.15, w4=0.15, focusing on processing efficiency. The trigger conditions for weight adjustment are: ① Changes in growth stage (determined to enter the maturity period by leaf area index LAI>2.8); ② Product quality coefficient of variation greater than 15%; ③ Environmental parameters deviate from the normal range by more than 20%. The weight adjustment algorithm adopts fuzzy logic control, with the input variables being growth stage index, quality stability index and environmental deviation index, the output variable being the weight adjustment amount, and the membership function adopting a triangular function.

[0064] The system also configures a cloud computing platform 5 to work in conjunction with the edge computing layer 2. The edge computing node 21 handles real-time control tasks with a response time of less than 5 minutes. The cloud computing platform 5 performs model training and parameter optimization and is equipped with an 8-core CPU (Intel Xeon E5-2640 v4), 32GB DDR4 memory, and 500GB SSD storage space. The data transmission strategy is: when the communication distance is less than 3km and the data volume is less than 500KB, the LoRa network (frequency 433MHz, transmission rate 5.47kbps, power consumption 200kbps) is preferred.

[0065] <120mW). When the communication distance is greater than 3km or the data volume is greater than 500KB, the 4G network (LTE Cat.4, downlink rate 100Mbps, uplink rate 50Mbps) is selected for data transmission. The data transmission frequency is set to 1 time every 5 minutes, and the data compression algorithm uses the LZ77 algorithm, with a compression ratio of 50%.

[0066] Verified by this example, the multi-time-scale collaborative control strategy improved the prediction accuracy to 84%, and the adaptive weight adjustment mechanism made the control effect of the system more precise at different growth stages. The protein content during the growth period increased to 17.8%, and the processing efficiency during the maturity period increased by 12%, proving the technical advantages of dual-time-domain control and adaptive weight adjustment.

[0067] Example 3

[0068] The purpose of this embodiment is to verify the intelligent control effect and system fault tolerance of the paper mulberry waste intelligent resource processing system.

[0069] like Figure 3 As shown, the system also includes an intelligent resource processing system for paper mulberry waste 6, including a fermentation monitoring sensor 61, a fermentation controller 62 and a waste processing device 63. The fermentation monitoring sensor 61 includes a temperature sensor PT100 (monitoring range 35-45°C, accuracy ±0.5°C, response time <60s), a pH sensor PH-8414 (monitoring range 6.0-7.5, accuracy

[0070] ±0.2 pH, temperature compensation range 0-60°C) and a DO-958 dissolved oxygen sensor (monitoring range 0.5-2.5%, accuracy ±0.2%, membrane material polytetrafluoroethylene). The fermentation controller 62 uses a Schneider Modicon M221 PLC controller with 16 digital input / output ports and 4 analog input ports (resolution 12 bits, accuracy 0.2%).

[0071] Waste treatment equipment 63 includes a crusher (power 11kW, processing capacity 1.5t / h), a fermentation tank, a stirring system, a ventilation system and a temperature control system. The fermentation tank volume is 30m 3, made of 304 stainless steel, equipped with automatic stirring device (speed 8 ~ 40rpm adjustable, power 4kW) and forced ventilation system (air volume 80 ~ 400m 3 / h adjustable, fan power 2.2kW). The temperature control system uses a combination of electric heating and water cooling, with a heating power of 15kW (using stainless steel electric heating tubes) and a cooling capacity of 12kW (using plate heat exchangers).

[0072] The fermentation control strategy adopts staged control: in the first stage (0-4 days), the temperature is controlled at 40-42°C, the pH value is controlled at 6.5-7.0, and the oxygen content is controlled at 1.5-2.0%. Aerobic fermentation is mainly carried out, and the microorganisms are mainly aerobic bacteria. The fermentation rate constant k1=0.12h-1; in the second stage (5-9 days), the temperature is controlled at 38-40°C, the pH value is controlled at 6.0-6.5, and the oxygen content is controlled at 0.8-1.2%. Anaerobic fermentation is switched to, and methanogens become active. The fermentation rate constant k2=0.06h-1; in the third stage (10-18 days), the temperature is controlled at 35-38°C, the pH value is controlled at 6.0-6.5, and the oxygen content is controlled at 0.5-1.0%. Stabilization treatment is carried out, and the organic matter degradation rate reaches more than 55%.

[0073] The execution control layer 4 also includes a redundant control node 43. When the main control node fails, the redundant control node 43 automatically takes over the control task within 10 minutes. Fault detection adopts a heartbeat detection mechanism with a detection cycle of 60 seconds. The heartbeat signal format is a UDP data packet (including node ID, timestamp, status code). A fault is determined when the heartbeat detection fails for three consecutive times. Redundant switching adopts a hot backup method. The redundant node synchronizes the control status and parameter settings of the main node in real time. Data synchronization adopts a master-slave replication mode, and the synchronization delay is <2 seconds. The fault switching process is: fault detection → status verification → authority switching → control takeover → status notification. The entire process is completed automatically.

[0074] Through the verification of this embodiment, the intelligent resource processing system for paper mulberry waste realizes the efficient conversion of paper mulberry waste into organic fertilizer and biomass energy, and the waste resource utilization rate reaches 67.5%. The output rate of organic fertilizer is 30% of the weight of the waste, the organic matter content is ≥40% (determined by potassium dichromate oxidation method), and the total content of nitrogen, phosphorus and potassium is ≥4% (nitrogen content is determined by Kjeldahl method, phosphorus content is determined by molybdenum antimony colorimetry, and potassium content is determined by flame photometry). The output rate of biomass energy is 12% of the weight of the waste, and the calorific value is ≥14MJ / kg (determined by oxygen bomb calorimeter). The system availability reaches 91.8%, and the average fault automatic switching time is 8.5 minutes, which proves the effectiveness of waste resource processing and the high reliability of the system.

[0075] Comparative Example 1

[0076] The purpose of this comparative example is to verify the impact of the lack of cross-link parameter correlation modeling on the overall performance of the system.

[0077] The same hardware configuration and data acquisition method as in Example 1 are used, but the optimization decision layer 3 only optimizes each link independently and does not establish a cross-link parameter association model. The planting link only optimizes the growth quality function G(x), and the objective function is:

[0078] G(x) = α1·protein content + α2·fiber content;

[0079] The processing link only optimizes the processing efficiency function P(x), and the objective function is:

[0080] P(x)=β1·Processing time -1 +β2·Product qualification rate;

[0081] The waste treatment process only optimizes the resource utilization function R(x), and the objective function is:

[0082] R(x) = γ1·water use efficiency + γ2·nutrient use efficiency.

[0083] Each link uses a fixed weight coefficient for single-objective optimization without considering the parameter coupling relationship between links.

[0084] The test conditions were exactly the same as those in Example 1. A six-month comparative test was conducted at the same paper mulberry planting base (located in Nanjing, Jiangsu Province, with an area of ​​50 mu and yellow-brown soil). A randomized block design was used with three replicates, and each treatment area was no less than 8 mu. The test method used the same sensor configuration and data acquisition frequency, and the same performance evaluation index. Resource utilization efficiency was calculated using the material balance method:

[0085]

[0086] The coefficient of variation of product quality is calculated using statistical analysis:

[0087]

[0088] The waste utilization rate is calculated using the mass balance method:

[0089]

[0090] The comparative test results showed that the resource utilization efficiency was 62.8%, 12.2% lower than the 71.5% in Example 1; the product quality coefficient of variation was 15.6%, 32.2% higher than the 11.8% in Example 1; the waste utilization rate was 48.5%, 26.7% lower than the 66.2% in Example 1; and the overall system efficiency was 18.5% lower than that in Example 1. Specific data comparisons are as follows: the average protein content of paper mulberry leaves was 16.8% (compared to 17.8% in Example 1), the average fiber content was 28.2% (compared to 26.5% in Example 1), the processing qualification rate was 78.2% (compared to 86.8% in Example 1), and the organic fertilizer output rate was 24% (compared to 30% in Example 1). The differences in each indicator reached a significant level (P < 0.05) through analysis of variance.

[0091] By comparison, it can be seen that the lack of cross-link parameter association modeling leads to the inability to coordinate optimization of various links, and the nutritional regulation during the planting period cannot match the subsequent processing requirements, resulting in waste of resources and unstable quality, and a significant decline in overall performance, which proves the necessity of cross-link parameter association modeling and the superiority of the technical solution of the present invention.

[0092] Comparative Example 2

[0093] The purpose of this comparative example is to verify the impact of the lack of multi-time scale collaborative control strategy on the system prediction accuracy and control effect.

[0094] The same system configuration as in Example 2 is used, but the paper mulberry growth-quality prediction correlation model module 31 only uses a single time scale of real-time control, the control cycle is fixed at 45 minutes, no long-term optimization module is set, and no Kalman filter processing is used. The control strategy uses a simple PID control, and the proportional coefficient K p =1.0, integral coefficient K i =0.6, differential coefficient K d =0.2, the control algorithm is:

[0095]

[0096] Where e(t) is the error signal and u(t) is the control output.

[0097] The test was carried out under the same environmental conditions (temperature 12-32°C, humidity 35-85% RH, light intensity 180-

[0098] 750 μmol / (m 2 ·s)), the same prediction accuracy evaluation method is used. The prediction accuracy calculation formula is:

[0099]

[0100] Forecast performance was analyzed using statistical analysis of forecast errors over a 10-day forecast period. The test sample size was 800 data points, covering different growth stages and environmental conditions. Time series cross-validation was used to evaluate model performance.

[0101] The comparative test results show that the prediction accuracy is 68.5%, which is 18.5% lower than the 84% in Example 2; the control response time is 12.8 minutes, which is 156% longer than the 5 minutes in Example 2; the system stability is poor, and the control accuracy decreases significantly under environmental disturbances. The average control error is 38% higher than that in Example 2. Specifically, the temperature control accuracy is ±2.2℃ (Example 2 is

[0102] ±1.5°C), humidity control accuracy ±12% RH (±8% RH in Example 2), pH control accuracy ±0.4 (±0.25 in Example 2), and nutrient concentration control accuracy ±18% (±12% in Example 2). In the event of a sudden environmental change (such as a sudden drop in temperature of 8°C), the system takes 65 minutes to return to a stable state (only 25 minutes in Example 2).

[0103] By comparison, it can be seen that the lack of multi-time-scale collaborative control strategy leads to insufficient system prediction accuracy, delayed control response, and inability to effectively cope with the time delay characteristics and environmental disturbances of biological systems, which proves the importance and technical advantages of dual-time-domain collaborative control.

[0104] Comparative Example 3

[0105] The purpose of this comparative example is to verify the performance difference between the traditional three-layer architecture and the four-layer intelligent control architecture of the paper mulberry full chain of the present invention.

[0106] The system utilizes a traditional three-tier architecture, consisting of a data collection layer, a cloud processing layer, and an execution control layer. There is no independent edge computing layer or optimization decision-making layer. All data processing and optimization calculations are performed in the cloud, using centralized control. The cloud server is equipped with a 6-core CPU (Intel Xeon E5-2603 v4, 1.7GHz), 16GB of DDR4 memory, and 256GB of SSD storage.

[0107] Data transmission relies entirely on the 4G network, and all sensor data must be uploaded to the cloud for processing before control commands are issued. The data transmission protocol uses MQTT (Message Queuing Telemetry Transport), with a QoS level of 1 (at least once delivered), and an average data packet size of 1.5KB. The control algorithm employs a traditional multi-objective optimization method, using a weighted summation approach to solve multi-objective problems. The weights are fixed and lack adaptive weight adjustment. The optimization algorithm uses a standard genetic algorithm with a population size of 50 and 150 iterations.

[0108] Comparative test results show that the system response time is 18.5 minutes, 270% longer than the 5 minutes in Example 1; the system cannot operate normally in the event of a network failure, with availability of only 82.6%, 10.0% lower than the 91.8% in Example 1; the data transmission cost is 165% higher than that in Example 1 (average monthly traffic fee of 485 yuan vs. 183 yuan); and control accuracy is significantly reduced when network latency is large (>800ms). Specifically, the data upload delay averages 12.5 seconds (edge ​​processing delay in Example 1 is <2 seconds), the control instruction issuance delay averages 8.8 seconds (local execution delay in Example 1 is <0.2 seconds), the system stops working in the event of a network interruption (Example 1 can run offline for 90 minutes), and the average daily data transmission volume is 120MB (Example 1 only requires 45MB).

[0109] By comparison, it can be seen that the traditional three-tier architecture is inferior to the four-tier intelligent control architecture of the paper mulberry full chain of the present invention in terms of response time, system reliability and operating cost. Especially in the agricultural environment with unstable network, the advantages of edge computing are more obvious, which proves the technical advantages and practical value of the four-tier distributed architecture design.

[0110] Based on the above embodiments and comparative examples, the present invention realizes the collaborative optimization control of the entire industry chain of paper mulberry circular agriculture through core technological innovations such as the four-layer intelligent control architecture of the entire chain of paper mulberry, cross-link parameter correlation modeling, multi-time scale collaborative control strategy, adaptive weight adjustment mechanism and paper mulberry waste intelligent resource processing system. Compared with the existing technology, the present invention has significantly improved resource utilization efficiency, product quality stability, waste utilization rate, system reliability and intelligence level, providing an effective technical solution for the intelligent development of circular agriculture. The system construction cost is about 120,000 yuan / 50 mu, the annual operating cost is 25,000 yuan, and the investment recovery period is 5.2 years. It has good economic feasibility and promotion and application prospects.

Claims

1. An ecological closed-loop system for the entire industrial chain of paper mulberry circular agriculture, characterized in that: include: Data collection layer, including soil sensors, meteorological sensors, plant physiological sensors, and product quality sensors; The edge computing layer includes multiple edge computing nodes, and adopts filtering algorithm and feature extraction algorithm to pre-process the data collected by the data acquisition layer; the optimization decision layer includes a paper mulberry growth-quality prediction correlation model module and a paper mulberry whole chain collaborative optimization objective function module; the execution control layer includes distributed control nodes and executive agencies, and the distributed control nodes drive the executive agencies through the communication protocol according to the control instructions of the optimization decision layer to realize automatic control of the entire paper mulberry planting-harvesting-processing-waste treatment-resource recycling industry chain.

2. The paper mulberry circular agriculture full industry chain ecological closed-loop system according to claim 1 is characterized in that: The soil sensor includes a soil temperature sensor, a soil moisture sensor and a soil pH sensor, with detection accuracy of ±0.5°C, ±3%RH and ±0.2pH respectively; the meteorological sensor includes a light intensity sensor and a temperature and humidity sensor, with detection accuracy of ±10μmol / (m 2 ·s), ±0.5℃, ±5%RH; the plant physiological sensor detects the protein content and fiber content of paper mulberry leaves with a detection accuracy of ±0.5% and ±1.0%, respectively; the product quality sensor detects the moisture content and crude protein content of processed products with a detection accuracy of ±1.0% and ±0.5%, respectively.

3. The paper mulberry circular agriculture full industry chain ecological closed-loop system according to claim 1 is characterized in that: The edge computing node adopts a moving average filtering algorithm for noise filtering with a window length of 10 sampling points, adopts a principal component analysis method for feature extraction, and retains the principal components with a cumulative contribution rate greater than 85%.

4. The paper mulberry circular agriculture full industry chain ecological closed-loop system according to claim 1 is characterized in that: The paper mulberry growth-quality prediction association model module establishes a machine learning regression relationship based on the protein content and fiber content of paper mulberry leaves and the nitrogen content, phosphorus content, potassium content, organic matter content, light intensity, temperature, humidity and pH value of the soil during the planting period. 2 Greater than 0.

90.

5. The paper mulberry circular agriculture full industry chain ecological closed-loop system according to claim 1 is characterized in that: The paper mulberry full chain collaborative optimization objective function module adopts a multi-objective optimization algorithm for collaborative optimization, and the objective function is F(x)=w1·G(x)+w2·P(x)+w3·R(x)+w4·E(x), Where G(x) is the growth quality function, P(x) is the processing efficiency function, R(x) is the resource utilization function, E(x) is the environmental impact function, and w1, w2, w3, and w4 are weight coefficients.

6. The paper mulberry circular agriculture full industry chain ecological closed-loop system according to claim 5 is characterized in that: The paper mulberry full-chain collaborative optimization objective function module includes an adaptive weight adjustment mechanism, which dynamically adjusts the weight coefficients w1, w2, w3, and w4 according to the paper mulberry growth stage and environmental conditions. The weight adjustment triggering condition is a change in the growth stage or a product quality variation coefficient greater than 15%.

7. The paper mulberry circular agriculture full industry chain ecological closed-loop system according to claim 4 is characterized in that: The paper mulberry growth-quality prediction correlation model module adopts multi-time scale prediction control, including a short-term control module and a long-term optimization module. The control period of the short-term control module is 30 to 60 minutes, and Kalman filtering is used for processing. The prediction period of the long-term optimization module is 7 to 14 days, preferably 10 days.

8. The paper mulberry circular agriculture full industry chain ecological closed-loop system according to claim 1 is characterized in that: The edge computing layer works in collaboration with the cloud computing platform. The edge computing nodes process real-time control tasks, and the cloud computing platform performs model training and parameter optimization. When the communication distance is less than 3km and the data volume is less than 500KB, the LoRa network is preferentially selected for data transmission. When the communication distance is greater than 3km or the data volume is greater than 500KB, the 4G network is selected for data transmission.

9. The paper mulberry circular agriculture full industry chain ecological closed-loop system according to claim 1, characterized in that: The execution control layer includes redundant control nodes. When a main control node fails, the redundant control node automatically takes over the control task within a preset time. Fault detection adopts a heartbeat detection mechanism.

10. A paper mulberry whole chain collaborative optimization control method based on the system of claim 1, characterized in that: The following steps are involved: Step 1: collecting real-time data of each link of paper mulberry planting, processing and waste treatment through the data collection layer; Step 2: The edge computing layer performs filtering and feature extraction on the collected data; Step 3: The paper mulberry growth-quality prediction association model module establishes a machine learning regression model based on the quantitative relationship between the protein content of paper mulberry leaves and soil nutrients; Step 4: The paper mulberry whole chain collaborative optimization objective function module adopts a multi-objective optimization algorithm for collaborative optimization; Step 5: Generate hierarchical control instructions based on the optimization results, and the execution control layer drives the actuator to realize automatic control.

Images