Nitrogen and phosphorus interception method and system for orchard based on intelligent monitoring

Abstract

The application relates to the technical field of data processing, and discloses an orchard nitrogen and phosphorus interception method and system based on intelligent monitoring. The method comprises the following steps: collecting orchard multi-sensor data and performing standardized processing to generate a monitoring data set; using a variational autoencoder to learn a nitrogen and phosphorus distribution rule and generate an interception parameter relationship model; calculating control instructions of a trench, a PRB and a sedimentation tank through a multi-objective optimization algorithm; controlling the collaborative operation of each interception device to collect nitrogen and phosphorus-rich sediments; and recycling the sediments to prepare slow-release fertilizers for reapplication in the orchard. Through the variational autoencoder data enhancement and the multi-objective optimization algorithm, the application solves the problems of insufficient training samples and single parameter optimization in the nitrogen and phosphorus interception of the orchard, and significantly improves the prediction accuracy of the intelligent monitoring and the collaborative control efficiency of the interception system.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method and system for intercepting nitrogen and phosphorus in orchards based on intelligent monitoring. Background Technology

[0002] In fruit-growing areas, nitrogen and phosphorus loss is a significant cause of eutrophication in some water bodies. Existing nitrogen and phosphorus interception technologies mainly include centralized treatment facilities, constructed wetlands, and ecological buffer zones, using physical, chemical, or biological methods to remove nitrogen and phosphorus pollutants from runoff. Traditional intelligent monitoring systems typically use a single sensor network to collect environmental parameters and apply basic data analysis algorithms to adjust these parameters, thereby monitoring the operational status of the interception devices. Some advanced systems have begun to apply machine learning algorithms to predict nitrogen and phosphorus concentrations and combine them with automated control technologies to adjust the operating parameters of the interception equipment.

[0003] However, existing technologies have significant shortcomings: the high procurement and operating costs of nitrogen and phosphorus sensors result in insufficient monitoring data samples, which severely restricts the training effect and prediction accuracy of intelligent algorithms; traditional methods lack the ability to deeply learn the distribution patterns of nitrogen and phosphorus in orchard runoff, and cannot generate sufficient training samples to optimize interception parameters; existing single-objective optimization methods struggle to find a balance between nitrogen and phosphorus removal efficiency and operating costs, affecting the economic feasibility of the system.

[0004] Based on the above analysis, the core problem of existing technologies lies in the contradiction between data scarcity and the needs of algorithm optimization. The high cost of sensors limits the scale of data collection, resulting in insufficient training samples for intelligent algorithms. This affects the accuracy of parameter optimization and ultimately restricts the intelligence level and operational efficiency of the entire interception system. Therefore, there is an urgent need to develop an intelligent nitrogen and phosphorus interception method that can solve the problem of insufficient samples through data augmentation techniques and achieve multi-objective collaborative optimization. Summary of the Invention

[0005] This application provides a method and system for nitrogen and phosphorus interception in orchards based on intelligent monitoring. By using variational autoencoder data augmentation and multi-objective optimization algorithms, it solves the problems of insufficient training samples and single parameter optimization in nitrogen and phosphorus interception in orchards, and significantly improves the prediction accuracy of intelligent monitoring and the collaborative control efficiency of the interception system.

[0006] In the first aspect, this application provides a method for intercepting nitrogen and phosphorus in orchards based on intelligent monitoring. The method for intercepting nitrogen and phosphorus in orchards based on intelligent monitoring includes: step S101, collecting data from orchard nitrogen and phosphorus concentration sensors, soil moisture sensors, rainfall sensors and flow sensors, and generating an orchard nitrogen and phosphorus monitoring dataset after outlier detection and standardization.

[0007] Step S101: Collect data from orchard nitrogen and phosphorus concentration sensors, soil moisture sensors, rainfall sensors, and flow sensors, and generate orchard nitrogen and phosphorus monitoring datasets after outlier detection and standardization.

[0008] Step S102: Based on the orchard nitrogen and phosphorus monitoring dataset, use the variational autoencoder algorithm to learn the distribution pattern of nitrogen and phosphorus in orchard runoff and generate a model of the relationship between nitrogen and phosphorus content and interception parameters.

[0009] Step S103: Using the nitrogen and phosphorus content and interception parameter relationship model, calculate the control instructions for the stepped trench water flow velocity, PRB adsorption material filling amount, and sedimentation tank flocculant dosage through a multi-objective optimization algorithm;

[0010] Step S104: Control the operation of the trench regulating valve, PRB filling system and sedimentation tank stirring device according to the control instructions to intercept nitrogen and phosphorus in the runoff and collect nitrogen and phosphorus-rich sediments;

[0011] Step S105: Based on the nitrogen- and phosphorus-rich sediment, slow-release nitrogen fertilizer and phosphorus fertilizer are prepared and reapplied to the orchard using a suction device.

[0012] Optionally, step S101 includes:

[0013] Real-time nitrogen and phosphorus content data in orchard soil and runoff were collected using nitrogen and phosphorus concentration sensors to obtain raw nitrogen and phosphorus concentration data sequences.

[0014] Environmental parameter data are collected using soil moisture sensors and rainfall sensors, and runoff flow change data are obtained using flow sensors to obtain orchard environmental monitoring parameter set;

[0015] The original nitrogen and phosphorus concentration data sequence was processed using the 3σ criterion outlier detection method to remove data points that exceeded the normal range, resulting in cleaned nitrogen and phosphorus concentration data.

[0016] The cleaned nitrogen and phosphorus concentration data and the orchard environmental monitoring parameter set were normalized and standardized to unify the data dimensions to the [0,1] interval, thus obtaining the orchard nitrogen and phosphorus monitoring dataset.

[0017] Optionally, step S102 includes:

[0018] A variational autoencoder network is constructed, and the orchard nitrogen and phosphorus monitoring dataset is used as input for feature extraction, outputting a potential feature vector of orchard nitrogen and phosphorus loss.

[0019] A variational autoencoder decoder network was constructed, and the original data was reconstructed using the potential feature vector of nitrogen and phosphorus loss in the orchard as input. The complete variational autoencoder model was trained by combining the KL divergence loss function.

[0020] The variational autoencoder model is run to learn the spatiotemporal distribution characteristics of nitrogen and phosphorus in orchard runoff, generate synthetic samples of nitrogen and phosphorus concentrations that conform to the characteristics of the orchard environment, and expand to form an enhanced nitrogen and phosphorus training dataset.

[0021] Based on the enhanced nitrogen and phosphorus training dataset, a nonlinear mapping relationship is established between the nitrogen and phosphorus content input and the output of the stepped trench water flow velocity, PRB adsorbent filling amount, and sedimentation tank flocculant dosage, forming a model of the relationship between nitrogen and phosphorus content and interception parameters.

[0022] Optionally, step S103 includes:

[0023] A multi-objective optimization algorithm framework is established, with the relationship model between nitrogen and phosphorus content and interception parameters as constraints, and nitrogen and phosphorus removal efficiency and operating cost as dual objective functions.

[0024] Input the real-time nitrogen and phosphorus concentration data of the current orchard into the nitrogen and phosphorus content and interception parameter relationship model to calculate the theoretical treatment load parameters of the three-stage interception system of stepped trench, PRB reaction wall and sedimentation tank;

[0025] A multi-objective optimization algorithm is used to search for Pareto solutions. Under the constraints of the theoretical processing load parameters, a combination scheme of the range of water flow velocity in the stepped trench, the range of PRB adsorption material filling amount, and the range of flocculant dosage in the sedimentation tank is calculated.

[0026] Select the balance point between nitrogen and phosphorus removal efficiency and operating cost from the combined schemes, and determine the specific values ​​of the stepped trench water flow velocity, PRB adsorption material filling amount, and sedimentation tank flocculant dosage as control instructions.

[0027] Optionally, step S104 includes:

[0028] The stepped channel water flow velocity parameters in the control command are analyzed, and the channel regulating valve is driven to perform opening adjustment action to control the flow rate of runoff through the channel and the contact time between nitrogen and phosphorus and iron oxide coated sand.

[0029] The PRB adsorbent material filling amount parameter in the control command is analyzed, and the PRB filling system is started to perform adsorbent material replenishment or replacement operations to maintain the active adsorption capacity of calcium-based materials and iron oxide coated sand in the PRB reaction wall.

[0030] The flocculant dosage parameters in the control instructions are analyzed to control the speed of the sedimentation tank stirring device and the flow rate of the polyaluminum chloride dosing pump, thereby promoting the formation of flocculation and sedimentation of residual nitrogen and phosphorus in the runoff and allowing them to settle to the bottom of the tank.

[0031] Through a three-stage synergistic interception and treatment process involving stepped trenches, PRB reactive walls, and sedimentation tanks, nitrogen and phosphorus in the runoff are progressively enriched into insoluble precipitates, which then accumulate at the bottom of the sedimentation tank to form nitrogen- and phosphorus-rich sediments.

[0032] Optionally, the step of analyzing the flocculant dosage parameters in the control command, controlling the stirring speed of the sedimentation tank and the flow rate of the polyaluminum chloride dosing pump, and promoting the formation of flocculation and sedimentation of residual nitrogen and phosphorus in the runoff to the bottom of the tank includes:

[0033] Read the numerical parameters of the flocculant dosage in the sedimentation tank from the control command, calculate the target flow rate and dosage duration of the polyaluminum chloride dosing pump, and generate a flocculant dosing control signal.

[0034] Based on the flocculant addition control signal, the polyaluminum chloride addition pump is started, and polyaluminum chloride solution is quantitatively added into the sedimentation tank according to the calculated target flow rate to form a flocculation reaction environment in the tank.

[0035] According to the stirring device speed parameters in the control command, the rotation speed of the stirring blades in the sedimentation tank is adjusted to control the shear force and mixing intensity of the water flow in the tank, so as to promote the flocculation reaction of residual nitrogen and phosphorus with polyaluminum chloride to generate large flocs.

[0036] The stirring device is stopped and the sedimentation process begins. Gravity causes the large flocs to sink to the bottom of the sedimentation tank, separating and clarifying the upper water and the nitrogen- and phosphorus-rich sediments at the bottom.

[0037] Optionally, step S105 includes:

[0038] Start the bottom suction device of the sedimentation tank to perform negative pressure extraction of the nitrogen- and phosphorus-rich sediment, and transport the sediment from the bottom of the sedimentation tank to the dewatering treatment device to separate the solid and liquid phases to obtain concentrated nitrogen- and phosphorus sludge.

[0039] The concentrated nitrogen and phosphorus sludge is transported to an intelligent granulation system, where biochar carrier and slow-release coating material are added for mixing and granulation. The granulation temperature and pressure parameters are controlled to form uniform slow-release fertilizer granules.

[0040] The nitrogen and phosphorus content of the slow-release fertilizer granules was tested and analyzed. Based on the test results, the nitrogen and phosphorus ratio in the granules was adjusted to generate slow-release nitrogen and phosphorus fertilizer products that meet the nutrient requirements of orchard soil.

[0041] Intelligent fertilization machinery is used to apply slow-release nitrogen and phosphorus fertilizers at specific locations according to the growth stage of fruit trees and the soil nutrient status, completing a closed-loop cycle of nitrogen and phosphorus nutrients from interception and recovery to reuse.

[0042] Secondly, this application provides an orchard nitrogen and phosphorus interception system based on intelligent monitoring, the orchard nitrogen and phosphorus interception system based on intelligent monitoring includes:

[0043] The data acquisition module is used to collect data from orchard nitrogen and phosphorus concentration sensors, soil moisture sensors, rainfall sensors, and flow sensors. After outlier detection and standardization, it generates an orchard nitrogen and phosphorus monitoring dataset.

[0044] The generation module is used to learn the distribution law of nitrogen and phosphorus in orchard runoff using the variational autoencoder algorithm based on the orchard nitrogen and phosphorus monitoring dataset, and generate a model of the relationship between nitrogen and phosphorus content and interception parameters.

[0045] The calculation module is used to calculate the control commands for the flow velocity of the stepped trench, the filling amount of PRB adsorbent material, and the dosage of flocculant in the sedimentation tank by using the nitrogen and phosphorus content and interception parameter relationship model and a multi-objective optimization algorithm.

[0046] The control module is used to control the operation of the trench regulating valve, PRB filling system and sedimentation tank stirring device according to the control instructions, to intercept nitrogen and phosphorus in the runoff and collect nitrogen and phosphorus-rich sediments;

[0047] The processing module is used to recover and process the nitrogen- and phosphorus-rich sediments using a suction device to prepare slow-release nitrogen and phosphorus fertilizers for reuse in the orchard.

[0048] Thirdly, an orchard nitrogen and phosphorus interception device based on intelligent monitoring is provided, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor invokes the instructions in the memory to cause the orchard nitrogen and phosphorus interception device based on intelligent monitoring to execute the above-described orchard nitrogen and phosphorus interception method based on intelligent monitoring.

[0049] Fourthly, a computer-readable storage medium is provided, wherein instructions are stored therein, which, when executed on a computer, cause the computer to perform the above-described method for intercepting nitrogen and phosphorus in orchards based on intelligent monitoring.

[0050] The technical solution provided in this application collects data from orchard nitrogen and phosphorus concentration sensors, soil moisture sensors, rainfall sensors, and flow sensors. The resulting orchard nitrogen and phosphorus monitoring dataset, generated through outlier detection and standardization, provides high-quality foundational data for subsequent intelligent algorithms, avoiding training bias issues caused by inconsistent data quality and dimensional differences in traditional methods. A variational autoencoder algorithm is used to learn the distribution patterns of nitrogen and phosphorus in orchard runoff. The generated model of the relationship between nitrogen and phosphorus content and interception parameters effectively solves the problem of insufficient training samples due to the high cost of nitrogen and phosphorus sensors. Data augmentation technology significantly expands the sample size available for model training, improving the model's adaptability to complex changes in the orchard environment and its prediction accuracy. Using the model of the relationship between nitrogen and phosphorus content and interception parameters, a multi-objective optimization algorithm calculates the control commands for the stepped ditch water flow velocity, PRB adsorption material filling amount, and sedimentation tank flocculant dosage, achieving an intelligent balance between nitrogen and phosphorus removal efficiency and operating costs. This overcomes the technical shortcomings of traditional single-objective optimization methods that cannot simultaneously consider economic and environmental benefits.

[0051] Based on regulatory commands, the system controls the operation of the channel regulating valves, PRB filling system, and sedimentation tank stirring device, achieving a collaborative control mechanism that intercepts nitrogen and phosphorus in runoff and collects nitrogen- and phosphorus-rich sediments. This overcomes the limitations of traditional manual management methods, such as slow response and low regulation accuracy, and realizes real-time linkage and precise control of multi-stage interception devices. A closed-loop utilization method is implemented, where nitrogen- and phosphorus-rich sediments are recovered and processed using suction equipment to prepare slow-release nitrogen and phosphate fertilizers for reuse in orchards. This not only solves the problem of difficult sediment disposal in traditional interception methods but also achieves on-site nutrient recycling, organically combining environmental governance with resource recovery. The variational autoencoder algorithm, specifically applied in orchard nitrogen and phosphorus interception, fully considers the seasonal variations and spatial distribution characteristics of orchard runoff. Through targeted optimization of the algorithm structure, the data augmentation effect is more closely aligned with practical application scenarios. Attached Figure Description

[0052] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 This is a schematic diagram of one embodiment of the orchard nitrogen and phosphorus interception method based on intelligent monitoring in this application.

[0054] Figure 2 This is a schematic diagram of an embodiment of the orchard nitrogen and phosphorus interception and treatment process based on intelligent monitoring in this application.

[0055] Figure 3 This is a schematic diagram of one embodiment of the orchard nitrogen and phosphorus interception system based on intelligent monitoring in this application.

[0056] Figure 4 This is a schematic block diagram of the structure of an orchard nitrogen and phosphorus interception device based on intelligent monitoring in an embodiment of the present invention. Detailed Implementation

[0057] This application provides a method and system for intercepting nitrogen and phosphorus in orchards based on intelligent monitoring. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0058] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the orchard nitrogen and phosphorus interception method based on intelligent monitoring in this application includes:

[0059] Step S101: Collect data from orchard nitrogen and phosphorus concentration sensors, soil moisture sensors, rainfall sensors, and flow sensors, and generate orchard nitrogen and phosphorus monitoring datasets after outlier detection and standardization.

[0060] Step S102: Based on the orchard nitrogen and phosphorus monitoring dataset, use the variational autoencoder algorithm to learn the distribution pattern of nitrogen and phosphorus in orchard runoff and generate a model of the relationship between nitrogen and phosphorus content and interception parameters.

[0061] Step S103: Using the nitrogen and phosphorus content and interception parameter relationship model, calculate the control instructions for the stepped trench water flow velocity, PRB adsorption material filling amount, and sedimentation tank flocculant dosage through a multi-objective optimization algorithm.

[0062] Step S104: Control the operation of the trench regulating valve, PRB filling system and sedimentation tank stirring device according to the control instructions to intercept nitrogen and phosphorus in the runoff and collect nitrogen and phosphorus-rich sediments;

[0063] Step S105: Based on the nitrogen- and phosphorus-rich sediments, slow-release nitrogen and phosphorus fertilizers are prepared and reapplied to the orchard using a suction device.

[0064] It is understood that the implementing entity of this application can be an orchard nitrogen and phosphorus interception system based on intelligent monitoring, or it can be a terminal or a server; the specific implementation is not limited here. This application's embodiments use a server as an example for illustration.

[0065] Specifically, the implementation of nitrogen and phosphorus interception in orchards begins with the collaborative data acquisition of multiple sensors. Nitrogen and phosphorus concentration sensors are deployed in soil profiles and runoff channels for real-time monitoring. , and To monitor changes in nitrogen and phosphorus content, soil moisture sensors are embedded in the active layer of the fruit tree root system, rainfall sensors are installed in open areas, and flow sensors are placed at runoff collection points. All sensors synchronously upload data at 5-minute intervals via IoT nodes. The raw monitoring data enters an anomaly detection phase, where a dynamic threshold method is used to identify and remove outliers exceeding a reasonable range. Missing data is imputed using spatiotemporal correlation analysis. The processed data stream undergoes min-max standardization, uniformly converting parameters of different dimensions to the [0,1] interval, forming a structured orchard nitrogen and phosphorus monitoring dataset.

[0066] A standardized dataset is input into a variational autoencoder for feature learning. The encoder network compresses the 6-dimensional input into an 8-dimensional latent space to capture the spatiotemporal distribution characteristics of nitrogen and phosphorus loss. The decoder network reconstructs the original data distribution from the latent space. During training, the KL divergence term constrains the latent space to follow a standard normal distribution. After training, the model generates synthetic samples by randomly sampling from the latent space to expand the original dataset. This augmented data is used to train the feedforward neural network, establishing a mapping relationship between nitrogen and phosphorus content and environmental parameters to interception parameters. The model outputs three key parameters: stepped channel flow velocity, PRB adsorbent filling amount, and sedimentation tank flocculant dosage.

[0067] The multi-objective optimization algorithm aims to optimize nitrogen and phosphorus removal efficiency and operating cost, using the theoretical treatment load output by the relational model as constraints. An improved NSGA-II algorithm is employed to search for the Pareto optimal solution set. The algorithm initializes the population by randomly generating parameter combinations within the constraints, selecting superior individuals through non-dominated sorting and crowding calculation, and outputting the optimal parameter combination after 200 generations of evolution. The optimization results are then used to select an equilibrium point through decision preference analysis, generating control instructions containing specific numerical values.

[0068] Control commands are sent to the field execution equipment via the bus. The trench regulating valve adjusts its opening according to the set value of water flow velocity, the PRB filling system replenishes adsorbent material according to the calculated amount, and the sedimentation tank dosing pump injects flocculant according to the optimized results. The three-stage interception system operates in concert: the trench removes particulate nitrogen and phosphorus through physical interception, the PRB reactive wall removes dissolved pollutants through chemical adsorption, and the sedimentation tank removes colloidal residues through flocculation and sedimentation, ultimately forming nitrogen- and phosphorus-rich sediments at the bottom of the tank.

[0069] Once the sediment accumulates to a set thickness, the recycling process is initiated. A screw pump transports the sludge to a dewatering unit, where centrifugation separates it into concentrated sludge. The sludge is then mixed with biochar and coating materials in a specific ratio and granulated into slow-release fertilizer granules. A near-infrared spectroscopy monitors the nutrient content of the granules in real time and provides feedback to adjust the formulation. The finished fertilizer is applied using a smart fertilizer applicator based on the fruit trees' nutrient requirements and soil moisture conditions, completing the material cycle from nitrogen and phosphorus interception to nutrient recovery. The entire system achieves efficient interception and resource utilization of nitrogen and phosphorus loss in orchards through closed-loop control via a sensor network, optimization algorithms, and actuators.

[0070] In one specific embodiment, the process of performing step S101 may specifically include the following steps:

[0071] Real-time nitrogen and phosphorus content data in orchard soil and runoff were collected using nitrogen and phosphorus concentration sensors to obtain raw nitrogen and phosphorus concentration data sequences.

[0072] Environmental parameter data are collected using soil moisture sensors and rainfall sensors, and runoff flow change data are obtained using flow sensors to obtain orchard environmental monitoring parameter set;

[0073] The original nitrogen and phosphorus concentration data sequence was processed by 3σ criterion outlier detection to remove data points that exceeded the normal range, resulting in cleaned nitrogen and phosphorus concentration data.

[0074] The cleaned nitrogen and phosphorus concentration data and the orchard environmental monitoring parameter set were normalized and standardized to unify the data dimensions to the [0,1] interval, thus obtaining the orchard nitrogen and phosphorus monitoring dataset.

[0075] Specifically, the construction of the orchard nitrogen and phosphorus monitoring dataset began with the synchronous data acquisition from nitrogen and phosphorus concentration sensors, soil moisture sensors, rainfall sensors, and flow sensors. The nitrogen and phosphorus concentration sensors measure nitrogen and phosphorus concentrations in the soil and runoff in real time using ion-selective electrodes. , and The system outputs raw nitrogen and phosphorus concentration data sequences, including timestamps, monitoring point numbers, and the concentration values ​​of the three ions. Soil moisture sensors are installed at a depth of 0-30 cm in the root zone of fruit trees to measure volumetric water content and generate soil moisture data sequences. Rainfall sensors use tipping bucket rain gauges to record cumulative rainfall and output rainfall intensity data sequences. Flow sensors are installed at runoff collection ditches using ultrasonic flow meters to record runoff velocity and cross-sectional flow rate, forming flow data sequences. Analog signals from all sensors are converted from analog to digital and then synchronously acquired at 5-minute sampling intervals to ensure time-series data alignment.

[0076] In the original nitrogen and phosphorus concentration data sequence , and Concentration data may contain outliers due to electrode drift or electrochemical noise. The 3σ criterion was used to detect outliers for each ion concentration, calculating the arithmetic mean μ and standard deviation σ of historical data, and removing data points falling outside the interval (μ-3σ, μ+3σ). Soil moisture data exceeding 110% of field capacity or falling below 50% of the wilting coefficient were marked as outliers. Extreme records of over 100 mm per hour in rainfall data, and negative values ​​or values ​​exceeding 120% of the ditch design flow rate in flow rate data were discarded. After outlier removal, cleaned nitrogen and phosphorus concentration data were generated. Missing values ​​were supplemented using linear interpolation at adjacent time points on the same sensor to ensure the continuity of the time series.

[0077] After cleaning, nitrogen and phosphorus concentration data were aligned with soil moisture, rainfall intensity, and flow data by timestamp and stitched together to form an orchard environmental monitoring parameter set. Each row of data in this parameter set contains data from the same time point. , and Concentration, soil volumetric water content, rainfall intensity, and runoff flow constitute a 6-dimensional vector set. Due to significant differences in the dimensions of each parameter, min-max normalization is used to linearly map each dimension of data to the [0,1] interval. The normalization formula is ( x - min) / (max - min), where min and max are the historical extreme values ​​of this parameter under typical orchard conditions: The concentration is 0-15 mg / L. The concentration is 0-10 mg / L. The concentrations were 0-5 mg / L, the soil volumetric moisture content was 15%-45%, the rainfall intensity was 0-50 mm / h, and the runoff was 0-0.5 m³ / s. The normalized 6-dimensional vector set formed the orchard nitrogen and phosphorus monitoring dataset, with each row representing a standardized monitoring parameter at a given time point, and the timestamp serving as an implicit index.

[0078] In one specific embodiment, the process of performing step S102 may specifically include the following steps:

[0079] A variational autoencoder network is constructed, and the orchard nitrogen and phosphorus monitoring dataset is used as input for feature extraction. The output is a potential feature vector of orchard nitrogen and phosphorus loss.

[0080] A variational autoencoder decoder network was constructed, and the original data was reconstructed using the potential feature vector of nitrogen and phosphorus loss in the orchard as input. The complete variational autoencoder model was trained by combining the KL divergence loss function.

[0081] The variational autoencoder model is run to learn the spatiotemporal distribution characteristics of nitrogen and phosphorus in orchard runoff, generate synthetic samples of nitrogen and phosphorus concentrations that conform to the characteristics of the orchard environment, and expand to form an enhanced nitrogen and phosphorus training dataset.

[0082] Based on the enhanced nitrogen and phosphorus training dataset, a nonlinear mapping relationship is established between the input nitrogen and phosphorus content and the output of the stepped trench water flow velocity, the PRB adsorbent filling amount, and the sedimentation tank flocculant dosage, forming a model of the relationship between nitrogen and phosphorus content and interception parameters.

[0083] Specifically, the encoder network of the variational autoencoder receives an orchard nitrogen and phosphorus monitoring dataset as input, which contains standardized nitrogen and phosphorus data. , and The latent feature vector z is a 6-dimensional vector containing the concentration of soil moisture, soil volumetric water content, rainfall intensity, and runoff. The encoder network consists of three fully connected layers. The first layer maps the 6-dimensional input to a 32-dimensional latent space, the second layer maintains the 32-dimensionality, and the output layer is divided into two parallel branches, generating an 8-dimensional mean vector μ and an 8-dimensional log-variance vector logσ², respectively. The latent feature vector z is obtained through reparameterization techniques. In the sampling process, the random noise ε distributed according to the standard normal distribution is multiplied by exp(logσ² / 2) and then added to μ to ensure that the gradient can propagate backward.

[0084] The decoder network takes an 8-dimensional latent feature vector z as input, undergoes a nonlinear transformation through two 32-dimensional fully connected layers, and finally maps it back to a 6-dimensional output space. The reconstructed output... , and Concentration, soil volumetric water content, rainfall intensity, and runoff flow rate maintained the same dimensions and physical meaning as the original input. During training, the reconstruction loss used mean squared error to calculate the difference between the original 6-dimensional vector and the reconstructed 6-dimensional vector, while the KL divergence loss constrained the latent space distribution to approximate a standard normal distribution. The weighted sum of the two losses constituted the total loss function. Optimization was completed within 200 training epochs using data with a batch size of 64, and the Adam optimizer with a learning rate set to 0.001 automatically adjusted and updated parameters.

[0085] The trained variational autoencoder generates new latent feature vectors by randomly sampling in the 8-dimensional latent space, and the decoder converts these vectors into synthetic samples. , and Concentration data exhibit the same statistical characteristics as real monitoring data, and soil volumetric water content, rainfall intensity, and runoff data maintain reasonable physical correlations. Synthetic samples were merged with the original monitoring data to form an enhanced nitrogen and phosphorus training dataset, expanding the sample size to 5-8 times that of the original data.

[0086] An enhanced nitrogen and phosphorus training dataset was used to train a three-layer feedforward neural network, establishing a mapping relationship between nitrogen and phosphorus content and environmental parameters to interception parameters. The input layer receives a 6-dimensional feature vector, containing... , and The study considered nitrogen and phosphorus concentrations, soil volumetric moisture content, rainfall intensity, and runoff. The hidden layer employed 32 neurons and the Tanh activation function, while the output layer generated three interception parameters: stepped ditch flow velocity, PRB adsorbent material filling amount, and sedimentation tank flocculant dosage. An early-stop strategy was employed during training to prevent overfitting. The resulting model dynamically predicted the optimal combination of interception parameters based on real-time monitoring data. The model output showed a stepped ditch flow velocity range of 0.1-0.5 m / s, a PRB adsorbent material filling amount range of 10-50 kg / m³, and a sedimentation tank flocculant dosage controlled at 5-20 mg / L.

[0087] In one specific embodiment, the process of executing step S103 may specifically include the following steps:

[0088] A multi-objective optimization algorithm framework was established, with the relationship model between nitrogen and phosphorus content and interception parameters as constraints, and nitrogen and phosphorus removal efficiency and operating cost as dual objective functions.

[0089] Input the real-time nitrogen and phosphorus concentration data of the orchard into the nitrogen and phosphorus content and interception parameter relationship model to calculate the theoretical treatment load parameters of the three-stage interception system of stepped trench, PRB reaction wall and sedimentation tank;

[0090] A multi-objective optimization algorithm is used to search for Pareto solutions. Under the constraints of theoretical load parameters, a combination scheme is calculated for the range of flow velocity in the stepped trench, the range of PRB adsorption material filling amount, and the range of flocculant dosage in the sedimentation tank.

[0091] Choose the balance point between nitrogen and phosphorus removal efficiency and operating cost from the combined scheme, and determine the specific values ​​of the stepped trench water flow velocity, PRB adsorption material filling amount, and sedimentation tank flocculant dosage as control instructions.

[0092] Specifically, the multi-objective optimization algorithm framework uses a model relating nitrogen and phosphorus content to interception parameters as its core constraint. This model is based on real-time monitoring data. , and The model uses six dimensions of data—concentration, soil volumetric moisture content, rainfall intensity, and runoff—to output three dimensions of parameters: stepped ditch flow velocity, PRB adsorbent material filling amount, and sedimentation tank flocculant dosage. The multilayer perceptron network within the model has been trained, and its weight matrix and bias vector are fixed to form a mathematical expression of the parameter mapping. The optimization framework simultaneously maintains two objectives: a nitrogen and phosphorus removal efficiency function and an operating cost function. The removal efficiency function is calculated using a weighted average of the measured removal rates from the three-stage interception system, with weight coefficients allocated by real-time flow. The operating cost function comprehensively calculates the energy consumption of the ditch regulating valve, the cost of replacing the PRB adsorbent material, and the cost of flocculant consumption in the sedimentation tank.

[0093] After real-time nitrogen and phosphorus concentration data are input into the nitrogen and phosphorus content-interception parameter relationship model, the model outputs three theoretical treatment load parameters: the maximum hydraulic load of the stepped trench, the PRB adsorption capacity limit, and the sedimentation tank treatment capacity threshold. These parameters serve as hard constraints to limit the search space of the NSGA-II algorithm, ensuring that all candidate solutions are within the engineering feasible range. During algorithm initialization, a population of 100 individuals is generated, each encoded as a real triple representing the trench water flow velocity, the PRB adsorption material filling amount, and the sedimentation tank flocculant dosage, respectively, with their values ​​strictly limited by the theoretical treatment load parameters.

[0094] During the NSGA-II algorithm's execution, the crossover operation uses a simulated binary crossover operator with a crossover probability of 0.9; the mutation operation uses a polynomial mutation operator with a mutation probability of 0.1. In each generation of evolution, the algorithm evaluates the performance of each individual on two objective functions: nitrogen and phosphorus removal efficiency is calculated using the physicochemical equations of the three-stage interception system, and operating costs are calculated based on historical orchard operation data. After 200 iterations, the algorithm outputs a set of non-dominated solutions on the Pareto front. These solutions form a continuous distribution within the range of ditch water flow velocity (0.05-0.25 m / s), PRB adsorbent material filling amount (5-30 kg / m³), and sedimentation tank flocculant dosage (2-12 mg / L).

[0095] When selecting the final control command from the Pareto solution set, a weighted Euclidean distance method is used for decision-making. After normalizing the nitrogen and phosphorus removal efficiency and operating cost to the [0,1] interval, the distance from each solution to the ideal point is calculated, and the solution with the smallest distance is selected as the optimal compromise. The specific values ​​of the determined stepped trench water flow velocity, PRB adsorbent material filling amount, and sedimentation tank flocculant dosage are encapsulated into a control command frame. The frame structure includes four fields: device identifier, parameter value, execution duration, and check code, and is sent to the field PLC controller via a wireless communication module. After the command is executed, the trench regulating valve adjusts its opening according to the water flow velocity parameter, the PRB filling system replenishes adsorbent material according to the calculated amount, and the sedimentation tank dosing pump injects flocculant according to the set value, realizing the coordinated operation of the three-stage interception system.

[0096] In one specific embodiment, the process of executing step S104 may specifically include the following steps:

[0097] The stepped channel flow velocity parameters in the control command are analyzed to drive the channel regulating valve to perform opening adjustment action, thereby controlling the flow rate of runoff through the channel and the contact time between nitrogen and phosphorus and iron oxide coated sand.

[0098] The PRB adsorbent material filling amount parameter in the control command is analyzed, and the PRB filling system is started to perform adsorbent material replenishment or replacement operations to maintain the active adsorption capacity of calcium-based materials and iron oxide coated sand in the PRB reaction wall.

[0099] Analyze the flocculant dosage parameters in the control instructions for sedimentation tanks, control the speed of the sedimentation tank stirring device and the flow rate of the polyaluminum chloride dosing pump, and promote the formation of flocculation and sedimentation of residual nitrogen and phosphorus in the runoff to settle to the bottom of the tank.

[0100] Through a three-stage synergistic interception and treatment process involving stepped trenches, PRB reactive walls, and sedimentation tanks, nitrogen and phosphorus in the runoff are progressively enriched into insoluble precipitates, which then accumulate at the bottom of the sedimentation tank to form nitrogen- and phosphorus-rich sediments.

[0101] Specifically, the control command frame is transmitted to the field PLC control system via a wireless communication network. The command parsing module extracts three parameters according to a predefined frame structure format: the water flow velocity in the stepped trench, the amount of PRB adsorbent material filling, and the amount of flocculant added to the sedimentation tank. The water flow velocity parameter in the trench is input into the PID control loop, compared with the measured value of the electromagnetic flowmeter at the trench outlet, and outputs a 4-20mA analog signal to drive the electric regulating valve. The valve opening is adjusted according to the preset water flow velocity-opening characteristic curve to stabilize the flow velocity in the trapezoidal cross-section trench within the design range of 0.05-0.25 m / s. A 0.3m thick iron oxide coated sand filter layer is laid at the bottom of the trench. The water flow velocity adjustment directly changes the contact time between nitrogen and phosphorus pollutants and the filter media. When the flow velocity decreases to 0.1 m / s, The coordination exchange time with the iron oxide surface is extended to 90 seconds, and the adsorption efficiency is improved by 15%.

[0102] The PRB adsorbent filling parameter triggers an automatic replenishment program. The PLC calculates the replenishment amount of calcium-based material and iron oxide coated sand based on the difference between the current reading of the weight sensor inside the reaction wall and the set value. The two materials are added to the reaction zone via a screw conveyor at a 1:2 mass ratio. During the conveying process, a differential pressure sensor built into the wall monitors the permeability coefficient change in real time. Feeding stops when the differential pressure increase exceeds 0.5 kPa / min, ensuring the permeability coefficient is maintained at a certain level. That's all. After replenishment, the spray system will automatically start based on feedback from the wall humidity sensor, controlling the moisture content of the adsorbent material within the optimal activity range of 15-20%.

[0103] The flocculant dosing system in the sedimentation tank receives dosing parameters from the PLC. This parameter is divided by the influent flow meter reading, and the output signal controls the speed of the variable frequency screw pump, injecting a 10% concentration of polyaluminum chloride solution into the mixing zone at a flow rate of 0.5-2.5 L / min. Simultaneously, the stirring device's speed is adjusted in conjunction with the dosing rate. Initially, high-intensity mixing is maintained at 120 rpm for 30 seconds, then reduced to 60 rpm for 90 seconds to maintain floc growth, before finally stopping stirring to enter the sedimentation stage. A turbidity sensor installed in the mixing zone monitors the floc formation status; when the NTU value decreases at a rate below 5% / min, the stirring intensity is automatically increased by 10%. After sedimentation, the supernatant is discharged through an overflow weir, and the nitrogen- and phosphorus-rich sediment formed at the bottom is monitored in real-time by an ultrasonic sludge level gauge. When the sediment layer thickness reaches 30 cm, a sludge pump start signal is triggered.

[0104] The operational status of the three-stage interception system is monitored in real time via the Modbus RTU protocol. Installations are located at the trench outlet, PRB effluent outlet, and sedimentation tank overflow outlet. , and The online analyzer uploads monitoring data to the edge computing node every 5 minutes. The node runs a sliding window algorithm to analyze the changing trends of the most recent 6 sets of data. When the concentration at any monitoring point exceeds a threshold (e.g., in water), the system detects the data. >1.0mg / L >1.0mg / L When the concentration is >0.2 mg / L, the process is automatically re-optimized, generating new control commands to be issued. After the sediment at the bottom of the sedimentation tank accumulates to the set thickness, the sludge pump starts according to the preset program, transporting the nitrogen- and phosphorus-rich sediment with a solids content of 15-20% to the dewatering treatment unit, completing the entire process from nitrogen and phosphorus interception to sediment collection.

[0105] In one specific embodiment, the process of parsing the sedimentation tank flocculant dosage parameter in the control command may specifically include the following steps:

[0106] Read the numerical parameters of the flocculant dosage in the sedimentation tank from the control command, calculate the target flow rate and dosage duration of the polyaluminum chloride dosing pump, and generate the flocculant dosing control signal.

[0107] Based on the flocculant dosing control signal, the polyaluminum chloride dosing pump is started, and polyaluminum chloride solution is quantitatively added into the sedimentation tank according to the calculated target flow rate to form a flocculation reaction environment in the tank.

[0108] According to the stirring speed parameters in the control instructions, adjust the rotation speed of the stirring blades in the sedimentation tank, control the shear force and mixing intensity of the water flow in the tank, and promote the flocculation reaction of residual nitrogen and phosphorus with polyaluminum chloride to generate large flocs.

[0109] The stirring device is stopped and the sedimentation stage begins. Gravity causes large flocs to sink to the bottom of the sedimentation tank, separating and clarifying the upper water and the nitrogen- and phosphorus-rich sediments at the bottom.

[0110] Specifically, during the flocculant dosing control process in the sedimentation tank, the flocculant dosing parameter in the control command frame is extracted by the PLC parsing module. This numerical parameter, along with real-time monitoring data from the sedimentation tank influent flow meter, is input into the dosing calculation module. The dosing calculation module performs a division operation to convert the flocculant concentration in mg / L units into the polyaluminum chloride solution dosing flow rate in L / min units, considering a fixed parameter of 10% concentration of the commercial polyaluminum chloride solution during the calculation. The calculation result generates a flocculant dosing control signal, which includes two key parameters: the target flow rate and the dosing duration. The target flow rate is limited to between 0.5-2.5 L / min, corresponding to a sedimentation tank influent flow rate of 20-100 m³ / h.

[0111] The flocculant dosing control signal drives the variable frequency screw pump control system. The pump speed and target flow rate are correlated through a preset flow-speed curve. An electromagnetic flowmeter installed at the screw pump outlet provides real-time feedback of the actual flow rate, which is compared with the target flow rate to form a closed-loop control. A PID control algorithm keeps the flow error within ±2%. Polyaluminum chloride solution is uniformly injected into the sedimentation tank inlet channel through a perforated pipe, with the injection point located 30cm above the channel bottom to ensure thorough mixing of the reagent and water. During the dosing process, the PLC records the cumulative dosing amount, and the pump automatically stops when the preset dosing duration is reached.

[0112] The stirring speed parameter is extracted from the control command and converted into a frequency converter control signal output to the stirring motor. The stirring blade speed is adjusted in two stages: the first stage is set to 120 rpm for 30 seconds to generate high-intensity turbulence, allowing polyaluminum chloride to diffuse rapidly and fully contact residual nitrogen and phosphorus; the second stage is reduced to 60 rpm and maintained for 90 seconds to reduce shear force and promote the growth of micro-flocs into large flocs of 0.5-1 mm. The stirring intensity is dynamically optimized based on the monitoring data of the online turbidity meter in the tank. When the NTU value decreases at a rate of less than 5% / min, the speed is automatically increased by 10%.

[0113] After the flocculation reaction is completed, the stirring device stops operating and the sedimentation stage begins. The sedimentation tank is designed as a horizontal flow type with an effective water depth of 2.5m and a hydraulic retention time of 2 hours. During sedimentation, the flocs sink at a speed of 0.3-0.5mm / s under the influence of gravity, and the upper water gradually clarifies. An ultrasonic sludge level gauge installed at the bottom of the tank monitors the sediment interface height in real time. When the interface rises to 30cm from the bottom of the tank, a sludge discharge signal is triggered. The clarified supernatant is discharged evenly through a sawtooth overflow weir, and the effluent NTU value is controlled below 5. The nitrogen- and phosphorus-rich sediment at the bottom has a solids content of 15-20%, which is periodically pumped by a sludge pump to the dewatering treatment unit, completing the entire process control from flocculation addition to solid-liquid separation.

[0114] In one specific embodiment, the process of executing step S105 may specifically include the following steps:

[0115] Start the bottom suction equipment of the sedimentation tank to perform negative pressure extraction of nitrogen- and phosphorus-rich sediments, and transport the sediments from the bottom of the sedimentation tank to the dewatering treatment device to separate the solid and liquid phases and obtain concentrated nitrogen- and phosphorus sludge.

[0116] Concentrated nitrogen and phosphorus sludge is transported to an intelligent granulation system, where biochar carriers and slow-release coating materials are added for mixing and granulation. Granulation temperature and pressure parameters are controlled to form uniform slow-release fertilizer granules.

[0117] The nitrogen and phosphorus content of slow-release fertilizer granules was tested and analyzed. Based on the test results, the nitrogen and phosphorus ratio in the granules was adjusted to generate slow-release nitrogen and phosphorus fertilizer products that meet the nutrient requirements of orchard soil.

[0118] Intelligent fertilization machinery is used to apply slow-release nitrogen and phosphorus fertilizers at specific locations according to the growth stage of fruit trees and the soil nutrient status, completing a closed-loop cycle of nitrogen and phosphorus nutrients from interception and recovery to reuse.

[0119] Specifically, when the ultrasonic sludge level gauge at the bottom of the sedimentation tank detects that the accumulated thickness of nitrogen- and phosphorus-rich sediment reaches 30 cm, it triggers the start signal of the suction equipment. Under a negative pressure of -0.06 MPa, the screw pump transports the sediment through a DN80 pipeline to the dewatering machine, with the flow velocity controlled below 1.2 m / s to prevent floc breakage. The dewatering machine drum speed is dynamically adjusted according to the torque signal within the range of 2800-3200 rpm, and the back pressure plate gap is adjusted synchronously, reducing the sediment moisture content from 96% to 75-78%, forming concentrated nitrogen- and phosphorus sludge containing 42 g / kg total nitrogen and 8 g / kg total phosphorus. During the dewatering process, the PLC records the dry sludge production data in real time and uploads it to the cloud server via a 4G module.

[0120] Concentrated nitrogen and phosphorus sludge is metered by a belt scale and then fed into a twin-shaft paddle mixer with a metering accuracy controlled within ±0.5 kg. The mixer simultaneously receives biochar carrier and slow-release coated urea-formaldehyde resin from a loss-in-weight scale. The biochar has a moisture content of 8% and is added at 30% of the dry sludge mass; the urea-formaldehyde resin has a particle size of 0.6 mm and is added at 4% of the finished granule mass. The paddles mix at 30 rpm for 180 seconds to ensure uniform material distribution. The mixture then falls into a flat die granulator, where the die temperature is maintained at 105±2℃ and the pressure roller pressure is 8 MPa, extruding into cylindrical granules with a diameter of 4 mm and a length of 6 mm. Torque and temperature sensors in the granulator provide real-time feedback to the control system. When the torque exceeds a set threshold, the feed rate is automatically reduced; when the temperature is below a set value, the heating power is increased to ensure a stable granule density of 1.28 g / cm³.

[0121] After granulation, the granules are graded by a vibrating screen, and qualified granules enter the near-infrared detection station. A spectrometer scans the particle surface in the 1200-2400nm wavelength range, and the collected absorbance data is input into a pre-trained partial least squares regression model. The model outputs the nitrogen and phosphorus content detection results. The detection system uses a nitrogen-to-phosphorus ratio of 3:1 as the control target. When the detection value deviates from the target, it automatically adjusts the resin coating thickness and dry mud blending ratio of subsequent batches. The coating thickness is fine-tuned by adjusting the granulator mold temperature; for every 1°C decrease, the coating amount is reduced by 0.2%. The dry mud blending ratio is adjusted by regulating the speed of the belt scale, ultimately ensuring that the nitrogen and phosphorus content error of the finished granules is controlled within ±3%.

[0122] Qualified slow-release fertilizer granules are loaded into the storage bin of the intelligent fertilizer applicator, which receives prescription map data from the orchard management system. The prescription map is generated by integrating soil conductivity, pH test results, and the nutrient requirements of fruit trees at different growth stages, controlling the applicator to precisely deliver fertilizer at a depth of 20cm below the root zone projection of the fruit trees. The rotation speed of the spiral fertilizer delivery pipe is proportional to the fertilizer application rate required by the prescription map. The fertilizer application rate is dynamically adjusted based on the measured values ​​of available nitrogen and phosphorus in the soil, increasing the nitrogen fertilizer ratio during the budding stage and increasing the phosphorus fertilizer application during the fruit enlargement stage. The fertilization operation record includes location coordinates, fertilizer application rate, and timestamp, and is uploaded to the cloud database via an IoT module, forming a complete closed loop with the previous nitrogen and phosphorus interception monitoring data.

[0123] The above describes the orchard nitrogen and phosphorus interception method based on intelligent monitoring in the embodiments of this application. The following describes the orchard nitrogen and phosphorus interception processing procedure based on intelligent monitoring in the embodiments of this application. Please refer to [link / reference]. Figure 2 One embodiment of the orchard nitrogen and phosphorus interception process based on intelligent monitoring in this application includes:

[0124] The orchard nitrogen and phosphorus interception and treatment process based on intelligent monitoring includes five core processes: data preprocessing, variational autoencoder modeling, multi-objective optimization decision-making, three-level interception and collaborative control, and sediment resource utilization. The entire process forms a closed-loop system from data perception to intelligent decision-making and then to physical interception, achieving efficient removal and resource recovery of nitrogen and phosphorus pollutants in orchard runoff.

[0125] In the data preprocessing process, nitrogen and phosphorus concentration sensors, soil moisture sensors, rainfall sensors, and flow sensors were deployed in the orchard soil profile, runoff channels, and surrounding environment to collect data synchronously at 5-minute intervals. , , Concentration, soil volumetric water content, rainfall intensity, and runoff flow were used to form the original data sequence; then, the 3σ criterion was applied to detect outliers and remove outliers, missing values ​​were filled using linear interpolation, and finally, min-max normalization was performed. Mapped to 0–15 mg / L Mapped to 0–10 mg / L By mapping to 0–5 mg / L, soil volumetric water content to 15%–45%, rainfall intensity to 0–50 mm / h, and runoff flow rate to 0–0.5 m³ / s, a 6-dimensional orchard nitrogen and phosphorus monitoring dataset was obtained.

[0126] The variational autoencoder modeling process involves inputting the aforementioned 6-dimensional orchard nitrogen and phosphorus monitoring dataset into the encoder network, which is then compressed into an 8-dimensional latent feature vector through three fully connected layers. The decoder network reconstructs the original 6-dimensional data using the latent feature vector as input. During training, mean squared error and KL divergence loss are combined, and latent space sampling generates synthetic samples which are then merged with real data, expanding the sample size to 5–8 times the original to form an enhanced nitrogen and phosphorus training dataset. A feedforward neural network is then trained based on this enhanced nitrogen and phosphorus training dataset, with the network input... , and Using concentration, soil volumetric moisture content, rainfall intensity, runoff flow rate, output stepped ditch water flow velocity 0.1–0.5 m / s, PRB adsorbent filling amount 10–50 kg / m³, and sedimentation tank flocculant dosage 5–20 mg / L, a model was established to establish the relationship between nitrogen and phosphorus content and interception parameters.

[0127] The multi-objective optimization decision-making process uses nitrogen and phosphorus removal efficiency and operating cost as dual objective functions, and the theoretical treatment load output by the model of the relationship between nitrogen and phosphorus content and interception parameters as constraints. The NSGA-II algorithm initializes 100 individuals, and after 200 generations of evolution with a simulated binary crossover probability of 0.9 and a polynomial mutation probability of 0.1, it outputs Pareto solution sets with channel water flow velocity of 0.05–0.25 m / s, PRB adsorption material filling amount of 5–30 kg / m³, and sedimentation tank flocculant dosage of 2–12 mg / L. Fuzzy membership degree is used to calculate the comprehensive satisfaction index, and a compromise scheme is selected and encapsulated into an optimization control command frame.

[0128] In the three-stage interception and collaborative control process, the PLC analyzes the stepped channel water flow velocity parameters in the control command frame. Through PID control and electromagnetic flowmeter closed-loop control, the opening of the electric regulating valve is maintained at the command value ±0.02 m / s within the trapezoidal cross-section channel, with an iron oxide coating sand thickness of 0.3 m. The PLC also analyzes the PRB adsorbent material filling quantity parameters. The difference between the weight sensor feedback and the set value is used to calculate the replenishment amount of calcium-based material and iron oxide coating sand. These are replenished via a screw conveyor at a 1:2 mass ratio. A differential pressure sensor monitors the permeability coefficient. The flocculant dosage parameters in the sedimentation tank were analyzed. The variable frequency screw pump was adjusted to deliver 0.5–2.5 L / min of 10% polyaluminum chloride solution according to the influent flow rate. The stirring speed was changed from 120 rpm / 30 s to 60 rpm / 90 s. When the turbidity sensor NTU value decreased by <5% / min, the speed was increased by 10%. The trench, PRB, and sedimentation tank operated in a coordinated manner, and online... , , The analyzer transmits data back every 5 minutes, and the edge node sliding window algorithm recalculates the control command when the concentration exceeds the threshold.

[0129] In the sediment resource utilization process, when the thickness of nitrogen- and phosphorus-rich sediment at the bottom of the sedimentation tank reaches 30 cm as monitored by an ultrasonic mud level gauge, a screw pump delivers the sediment to a screw press dewatering machine via a DN80 pipeline at a negative pressure of −0.06 MPa. The moisture content is reduced from 96% to 75%, resulting in concentrated nitrogen- and phosphorus sludge containing 42 g / kg total nitrogen and 8 g / kg total phosphorus. The concentrated nitrogen- and phosphorus sludge is then granulated with 30% biochar (8% moisture content) and 4% urea-formaldehyde resin (0.6 mm particle size) at 105 ℃ and 8 MPa using a flat die to form slow-release fertilizer granules with a diameter of 4 mm, a length of 6 mm, and a density of 1.28 g / cm³. Near-infrared spectroscopy is used to scan the 1200–2400 nm band, and the nitrogen and phosphorus content is output using a partial least squares model. The algorithm adjusts the coating thickness and dry mud ratio to maintain a nitrogen-to-phosphorus ratio of 3:1 with an error of ±3%. An intelligent fertilizer applicator receives soil conductivity, pH, and fruit tree growth period prescription maps, and the spiral fertilizer delivery pipe is positioned 20 cm below the root zone projection of the fruit trees. Precise application at a depth of cm, the amount of fertilizer applied is inversely proportional to the measured value of available nitrogen and phosphorus in the soil and directly proportional to the fertilizer requirement curve of fruit trees. The operation record is uploaded to the cloud via a 4G module, completing a closed loop cycle of nitrogen and phosphorus from interception and recycling to reuse.

[0130] The above describes the orchard nitrogen and phosphorus interception method based on intelligent monitoring in the embodiments of this application. The following describes the orchard nitrogen and phosphorus interception system 300 based on intelligent monitoring in the embodiments of this application. Please refer to [link to relevant documentation]. Figure 3 One embodiment of the orchard nitrogen and phosphorus interception system 300 based on intelligent monitoring in this application includes:

[0131] The data acquisition module 301 is used to acquire data from orchard nitrogen and phosphorus concentration sensors, soil moisture sensors, rainfall sensors, and flow sensors, and generate orchard nitrogen and phosphorus monitoring datasets after outlier detection and standardization.

[0132] The generation module 302 is used to learn the distribution law of nitrogen and phosphorus in orchard runoff using the variational autoencoder algorithm based on the orchard nitrogen and phosphorus monitoring dataset, and generate a model of the relationship between nitrogen and phosphorus content and interception parameters.

[0133] The calculation module 303 is used to calculate the control commands for the flow velocity of the stepped trench, the filling amount of PRB adsorbent material, and the dosage of flocculant in the sedimentation tank by using the nitrogen and phosphorus content and interception parameter relationship model and a multi-objective optimization algorithm.

[0134] Control module 304 is used to control the operation of the trench regulating valve, PRB filling system and sedimentation tank stirring device according to the control instructions, to intercept nitrogen and phosphorus in the runoff and collect nitrogen and phosphorus-rich sediments;

[0135] Processing module 305 is used to recover and process the nitrogen- and phosphorus-rich sediments using a suction device to prepare slow-release nitrogen fertilizer and phosphorus fertilizer for reapplication in the orchard.

[0136] above Figure 3 The orchard nitrogen and phosphorus interception system based on intelligent monitoring in this embodiment of the invention will be described in detail from the perspective of modular functional entities. The orchard nitrogen and phosphorus interception device based on intelligent monitoring in this embodiment of the invention will be described in detail from the perspective of hardware processing.

[0137] Reference Figure 4 This invention also provides an orchard nitrogen and phosphorus interception device 400 based on intelligent monitoring. This intelligent monitoring-based orchard nitrogen and phosphorus interception device can be a server, and its internal structure can be as follows: Figure 4 As shown. The intelligent monitoring-based orchard nitrogen and phosphorus interception device includes a processor 402, a memory 403, a display screen 404, an input device 405, a network interface 406, and a database 407 connected via a system bus 401. The processor in this computer design provides computing and control capabilities. The memory of the intelligent monitoring-based orchard nitrogen and phosphorus interception device includes a non-volatile storage medium 4031 and internal memory 4032. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database 407 of the intelligent monitoring-based orchard nitrogen and phosphorus interception device is used to store the corresponding data in this embodiment. The network interface 406 of the intelligent monitoring-based orchard nitrogen and phosphorus interception device is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements the above-described method.

[0138] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present invention and does not constitute a limitation on the orchard nitrogen and phosphorus interception device based on intelligent monitoring applied thereto.

[0139] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the steps of the orchard nitrogen and phosphorus interception method based on intelligent monitoring.

[0140] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0141] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a smart monitoring-based orchard nitrogen and phosphorus interception device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0142] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for intercepting nitrogen and phosphorus in orchards based on intelligent monitoring, characterized in that, The method includes: Step S101: Collect data from orchard nitrogen and phosphorus concentration sensors, soil moisture sensors, rainfall sensors, and flow sensors, and generate orchard nitrogen and phosphorus monitoring datasets after outlier detection and standardization. Step S102: Based on the orchard nitrogen and phosphorus monitoring dataset, a variational autoencoder algorithm is used to learn the distribution pattern of nitrogen and phosphorus in orchard runoff, generating a model of the relationship between nitrogen and phosphorus content and interception parameters. This includes: constructing a variational autoencoder encoder network, using the orchard nitrogen and phosphorus monitoring dataset as input for feature extraction, and outputting a potential feature vector of nitrogen and phosphorus loss in the orchard; constructing a variational autoencoder decoder network, using the potential feature vector of nitrogen and phosphorus loss in the orchard as input to reconstruct the original data, and training a complete variational autoencoder model using the KL divergence loss function; running the variational autoencoder model to learn the spatiotemporal distribution characteristics of nitrogen and phosphorus in orchard runoff, generating a synthetic sample of nitrogen and phosphorus concentration that conforms to the characteristics of the orchard environment, and expanding it to form an enhanced nitrogen and phosphorus training dataset; and establishing a nonlinear mapping relationship between the nitrogen and phosphorus content input and the output of the stepped ditch water flow velocity, PRB adsorption material filling amount, and sedimentation tank flocculant dosage based on the enhanced nitrogen and phosphorus training dataset, forming a model of the relationship between nitrogen and phosphorus content and interception parameters. Step S103: Using the nitrogen and phosphorus content and interception parameter relationship model, calculate the control commands for the stepped ditch water flow velocity, PRB adsorbent material filling amount, and sedimentation tank flocculant dosage through a multi-objective optimization algorithm: Establish a multi-objective optimization algorithm framework, using the nitrogen and phosphorus content and interception parameter relationship model as constraints, and set nitrogen and phosphorus removal efficiency and operating cost as dual objective functions; input the current real-time nitrogen and phosphorus concentration data of the orchard into the nitrogen and phosphorus content and interception parameter relationship model to calculate the theoretical treatment load parameters of the three-stage interception system of stepped ditch, PRB reactive wall, and sedimentation tank; run the multi-objective optimization algorithm to search for Pareto solutions, and calculate the combination schemes of the stepped ditch water flow velocity range, PRB adsorbent material filling amount range, and sedimentation tank flocculant dosage range under the constraints of the theoretical treatment load parameters; select the balance point between nitrogen and phosphorus removal efficiency and operating cost from the combination schemes, and determine the specific values ​​of the stepped ditch water flow velocity, PRB adsorbent material filling amount, and sedimentation tank flocculant dosage as control commands; Step S104: Control the operation of the trench regulating valve, PRB filling system and sedimentation tank stirring device according to the control instructions to intercept nitrogen and phosphorus in the runoff and collect nitrogen and phosphorus-rich sediments; Step S105: Based on the nitrogen- and phosphorus-rich sediment, slow-release nitrogen fertilizer and phosphorus fertilizer are prepared and reapplied to the orchard using a suction device.

2. The orchard nitrogen and phosphorus interception method based on intelligent monitoring according to claim 1, characterized in that, Step S101 includes: Real-time nitrogen and phosphorus content data in orchard soil and runoff were collected using nitrogen and phosphorus concentration sensors to obtain raw nitrogen and phosphorus concentration data sequences. Environmental parameter data are collected using soil moisture sensors and rainfall sensors, and runoff flow change data are obtained using flow sensors to obtain orchard environmental monitoring parameter set; The original nitrogen and phosphorus concentration data sequence was processed using the 3σ criterion outlier detection method to remove data points that exceeded the normal range, resulting in cleaned nitrogen and phosphorus concentration data. The cleaned nitrogen and phosphorus concentration data and the orchard environmental monitoring parameter set were normalized and standardized to unify the data dimensions to the [0,1] interval, thus obtaining the orchard nitrogen and phosphorus monitoring dataset.

3. The orchard nitrogen and phosphorus interception method based on intelligent monitoring according to claim 1, characterized in that, Step S104 includes: The stepped channel water flow velocity parameters in the control command are analyzed, and the channel regulating valve is driven to perform opening adjustment action to control the flow rate of runoff through the channel and the contact time between nitrogen and phosphorus and iron oxide coated sand. The PRB adsorbent material filling amount parameter in the control command is analyzed, and the PRB filling system is started to perform adsorbent material replenishment or replacement operations to maintain the active adsorption capacity of calcium-based materials and iron oxide coated sand in the PRB reaction wall. The flocculant dosage parameters in the control instructions are analyzed to control the speed of the sedimentation tank stirring device and the flow rate of the polyaluminum chloride dosing pump, thereby promoting the formation of flocculation and sedimentation of residual nitrogen and phosphorus in the runoff and allowing them to settle to the bottom of the tank. Through a three-stage synergistic interception treatment using stepped trenches, PRB reactive walls, and sedimentation tanks, nitrogen and phosphorus in the runoff are gradually removed and converted into insoluble precipitates, which accumulate at the bottom of the sedimentation tank to form nitrogen- and phosphorus-rich sediments.

4. The orchard nitrogen and phosphorus interception method based on intelligent monitoring according to claim 3, characterized in that, The process of analyzing the flocculant dosage parameters in the control instructions for the sedimentation tank, controlling the rotation speed of the sedimentation tank stirring device and the flow rate of the polyaluminum chloride dosing pump, and promoting the formation of flocculated sedimentation of residual nitrogen and phosphorus in the runoff and their settling to the bottom of the tank includes: Read the numerical parameters of the flocculant dosage in the sedimentation tank from the control command, calculate the target flow rate and dosage duration of the polyaluminum chloride dosing pump, and generate a flocculant dosing control signal. Based on the flocculant addition control signal, the polyaluminum chloride addition pump is started, and polyaluminum chloride solution is quantitatively added into the sedimentation tank according to the calculated target flow rate to form a flocculation reaction environment in the tank. According to the stirring device speed parameters in the control command, the rotation speed of the stirring blades in the sedimentation tank is adjusted to control the shear force and mixing intensity of the water flow in the tank, so as to promote the flocculation reaction of residual nitrogen and phosphorus with polyaluminum chloride to generate large flocs. The stirring device is stopped and the sedimentation process begins. Gravity causes the large flocs to sink to the bottom of the sedimentation tank, separating and clarifying the upper water and the nitrogen- and phosphorus-rich sediments at the bottom.

5. The orchard nitrogen and phosphorus interception method based on intelligent monitoring according to claim 1, characterized in that, Step S105 includes: Start the bottom suction device of the sedimentation tank to perform negative pressure extraction of the nitrogen- and phosphorus-rich sediment, and transport the sediment from the bottom of the sedimentation tank to the dewatering treatment device to separate the solid and liquid phases to obtain concentrated nitrogen- and phosphorus sludge. The concentrated nitrogen and phosphorus sludge is transported to an intelligent granulation system, where biochar carrier and slow-release coating material are added for mixing and granulation. The granulation temperature and pressure parameters are controlled to form uniform slow-release fertilizer granules. The nitrogen and phosphorus content of the slow-release fertilizer granules was tested and analyzed. Based on the test results, the nitrogen and phosphorus ratio in the granules was adjusted to generate slow-release nitrogen and phosphorus fertilizer products that meet the nutrient requirements of orchard soil. Intelligent fertilization machinery is used to apply slow-release nitrogen and phosphorus fertilizers at specific locations according to the growth stage of fruit trees and the soil nutrient status, completing a closed-loop cycle of nitrogen and phosphorus nutrients from interception and recovery to reuse.

6. A nitrogen and phosphorus interception system for orchards based on intelligent monitoring, characterized in that, For implementing the orchard nitrogen and phosphorus interception method based on intelligent monitoring as described in any one of claims 1-5, the orchard nitrogen and phosphorus interception system based on intelligent monitoring comprises: The data acquisition module is used to collect data from orchard nitrogen and phosphorus concentration sensors, soil moisture sensors, rainfall sensors, and flow sensors. After outlier detection and standardization, it generates an orchard nitrogen and phosphorus monitoring dataset. The generation module is used to learn the distribution law of nitrogen and phosphorus in orchard runoff using the variational autoencoder algorithm based on the orchard nitrogen and phosphorus monitoring dataset, and generate a model of the relationship between nitrogen and phosphorus content and interception parameters. The calculation module is used to calculate the control commands for the flow velocity of the stepped trench, the filling amount of PRB adsorbent material, and the dosage of flocculant in the sedimentation tank by using the nitrogen and phosphorus content and interception parameter relationship model and a multi-objective optimization algorithm. The control module is used to control the operation of the trench regulating valve, PRB filling system and sedimentation tank stirring device according to the control instructions, to intercept nitrogen and phosphorus in the runoff and collect nitrogen and phosphorus-rich sediments; The processing module is used to recover and process the nitrogen- and phosphorus-rich sediments using a suction device to prepare slow-release nitrogen and phosphorus fertilizers for reuse in the orchard.

7. A nitrogen and phosphorus interception device for orchards based on intelligent monitoring, characterized in that, The method includes a memory and a processor, wherein the memory stores a computer program that can run on the processor, and the processor executes the computer program to implement the orchard nitrogen and phosphorus interception method based on intelligent monitoring as described in any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by the processor, it causes the processor to execute the orchard nitrogen and phosphorus interception method based on intelligent monitoring as described in any one of claims 1 to 5.

Images