All-year production management method of rice, loach and shrimp ecological planting and breeding system

Abstract

The invention discloses a year-round production management method of a rice, loach and shrimp ecological planting and breeding system, and belongs to the technical field of intelligent agriculture. The method comprises the following steps: acquiring multi-dimensional data through a sensing network to obtain a comprehensive state vector; based on the vector and a target induction knowledge base, judging the physiological state of the organism and calculating a coupling matching degree; when the coupling matching degree is lower than a preset threshold value, active induction regulation and dynamic priority arbitration are conducted, an optimal regulation instruction is generated, and the arbitration process combines a long-term production target and ecological conflict punishment, comprises a closed-loop self-adaptive learning mechanism and is used for correcting model parameters and a knowledge base on line; and finally, the instruction is executed by the environment and biocenosis regulation and control execution system. According to the method, the conversion from passive stress regulation to active predictive optimization can be realized, decision conflicts in a multi-species symbiotic system are effectively processed, the method has adaptive learning ability, and the globality, the stability and the intelligent level of system management are improved.

Description

Technical Field

[0001] This invention relates to the field of smart agriculture technology, specifically to a year-round production management method for a rice-loach-shrimp ecological farming system. Background Technology

[0002] Rice-loach-shrimp integrated ecological farming, as an intensive and ecological agricultural production model, improves output and resource utilization per unit area by creating a multi-species symbiotic environment. In current production practices, Internet of Things (IoT) technology has been initially applied to this type of system, mainly through the deployment of various sensors to achieve real-time monitoring of environmental parameters such as dissolved oxygen, pH value, and temperature in the water.

[0003] However, existing technologies have inherent limitations in their management logic. Their control strategies largely rely on preset static thresholds; the system only triggers corresponding intervention devices, such as starting an aerator or water pump, when a monitored parameter exceeds the threshold range. This reactive control mechanism essentially remedies problems after they occur, resulting not only in response lag but also in potential for frequent system oscillations around the threshold, leading to a waste of energy and water resources.

[0004] More importantly, the rice-loach-shrimp system is a complex community of life, with different species having varying environmental needs at different growth stages, sometimes even mutually restricting each other. Current technologies typically treat each regulatory target in isolation, lacking a mechanism for comprehensively assessing the overall system status and making collaborative decisions. For example, specific agricultural practices or materials applied to promote rice growth may pose potential risks to the survival of shrimp or loach, and vice versa. This fragmented decision-making makes regulatory actions often localized and one-sided, failing to achieve global optimization of the entire ecosystem.

[0005] Furthermore, the models and rules of current management systems are typically universal and rarely change once deployed. They cannot self-adjust and optimize based on specific field soil conditions, microclimates, and historical production data, meaning they lack the ability to adapt to site specificity and the mechanism to learn from experience. At the same time, these systems generally lack a long-term planning perspective spanning the entire year; their regulatory actions are mostly focused on short-term maintenance, neglecting strategic arrangements that significantly impact overall annual benefits, such as crop rotation and resource pre-scheduling. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a year-round production management method for a rice-loach-shrimp ecological farming system. This method solves the problems of existing management methods lacking the ability to plan and dynamically optimize the entire year-round production cycle, and having fixed system model parameters that cannot adaptively learn and iterate based on historical environmental changes and regulatory feedback, resulting in low overall system robustness and comprehensive benefits.

[0007] To achieve the above objectives, the present invention provides the following technical solution: The first aspect of the present invention provides a year-round production management method for a rice-loach-shrimp ecological farming system, the method comprising: A sensing network deployed within the rice-crayfish ecological farming system periodically or through event-triggered data collection is used to obtain a comprehensive state vector characterizing the current state of the system. It can be derived from the environment state subvector. Biological state subvectors and production data state sub-vectors The system is structured based on the comprehensive state vector and a pre-defined target-induced knowledge base. This determines the physiological state of organisms within the system and calculates the coupling matching degree, reflecting the difference between the current system state and the ideal state. The target-induced knowledge base... It stores multiple reference state modes corresponding to ideal or excellent system states at different growth stages. The coupling matching degree... By combining the current integrated state vector The calculation process, derived by quantitative comparison with the ideal state patterns in the knowledge base, can be described by the following functional model: ; Among them, the It can be a metric function based on Euclidean distance or Mahalanobis distance, or a classification confidence score based on a machine learning model (such as support vector machine or neural network).

[0008] When the coupling matching degree is lower than the preset matching degree threshold At that time, the next step is triggered.

[0009] Active induction and dynamic priority arbitration are employed to generate optimal control instructions for regulating the rice-fish-shrimp ecological farming system. The specific steps are as follows: First, based on the aforementioned physiological state The system's long-term production goals and real-time environmental risks Calculation Dynamic priority weight vector representing the priorities of different regulatory targets .

[0010] ; Subsequently, based on the dynamic priority weight vector and the comprehensive state vector, a constrained optimization problem is constructed. Its objective is to find an optimal control command by minimizing a composite objective function. The objective function of this constrained optimization problem This can be expressed as: ; In the formula, the first term is the target deviation. For the model's predicted state, The first term represents the desired target state; the second term represents the control cost, expressed as a control command vector. The norm representation; the third term is the ecological conflict penalty. It is a penalty function used to quantify the negative impact of regulatory commands on other organisms. Dynamic priority weight vector The amount.

[0011] Finally, under the conditions of satisfying physical constraints (such as maximum equipment power) and biosafety constraints (such as maximum drug concentration), the constrained optimization problem is solved using numerical optimization algorithms such as Model Predictive Control (MPC) and Sequence Quadratic Programming (SQP) to obtain the optimal control command. .

[0012] The environmental and biological community regulation execution system deployed within the system executes the optimal regulation instructions to physically intervene in the system environment or biological community.

[0013] In a preferred embodiment, the method further includes performing ecological configuration and benefit optimization for multiple cropping throughout the year to generate a production operation sequence. and time scheduling scheme The long-term production target is determined and provided as a decision-making basis for the steps of proactive guidance and dynamic priority arbitration. This step also includes: employing a rolling correction mechanism to dynamically adjust the production operation sequence and the time scheduling scheme based on actual production progress and changes in the external environment.

[0014] In another preferred embodiment, the method further includes performing closed-loop adaptive learning. This step receives the optimal control command and the actual state data collected by the sensing network after executing the command. This step calculates the prediction error by comparing the predicted state with the actual state. Simultaneously assess the gap between the actual situation and the expected goals, and calculate the performance score. Based on the performance score and the prediction error, at least one of the following is updated using reinforcement learning or an online learning algorithm: Update the target-inducing knowledge base; Modify the model parameters Θ or arbitration rules used to construct the constrained optimization problem. .

[0015] A second aspect of the present invention provides a year-round production management system for a rice-loach-shrimp ecological farming system, the system comprising: The data preprocessing module is used to collect multidimensional data through a sensing network and preprocess the multidimensional data to obtain a comprehensive state vector that represents the current state of the system. The state discrimination module is used to discriminate the physiological state of organisms within the system based on the comprehensive state vector and the preset target-induced knowledge base, and to calculate the coupling matching degree. The dynamic priority arbitration module is used to perform proactive guidance and dynamic priority arbitration to generate the optimal control instruction; An environmental and biological community regulation execution system is used to execute the optimal regulation instructions.

[0016] This invention provides a year-round production management method for a rice-loach-shrimp ecological farming system. It has the following beneficial effects: 1. This invention uses coupling matching degree to quantitatively assess the overall system state, replacing the passive response mode based solely on a single threshold, thus enabling proactive intervention before significant deviations in the system state occur. Simultaneously, the dynamic priority arbitration step in the method incorporates ecological conflict penalty terms into the optimization objective, balancing the needs of different species when generating optimal control instructions. This ensures that control decisions are geared towards the global optimum of the entire symbiotic system, rather than focusing on local components.

[0017] 2. This invention provides a long-term operational framework encompassing production operation sequences and time scheduling schemes for the entire production cycle through a multi-cropping ecological configuration and benefit optimization process throughout the year. Combined with a rolling correction mechanism, this framework can be dynamically adjusted based on actual production data and environmental changes, integrating long-term planning with short-term regulation, thereby improving the efficiency of resource allocation and the adaptability of production management at the annual scale.

[0018] 3. The closed-loop adaptive learning mechanism included in this invention continuously updates the target-induced knowledge base and model parameters by analyzing the effectiveness scores and prediction errors fed back after each adjustment. This process enables the system to gradually improve the accuracy of its internal model based on operational data accumulated in specific fields, thereby enhancing the robustness and long-term effectiveness of the system's decision-making. Attached Figure Description Figure 1 This is a system functional module architecture diagram of the present invention; Figure 2 This is a schematic diagram of the deployment of the paddy field sensing and execution system of the present invention; Figure 3 This is a schematic diagram illustrating the deployment of the paddy field sensing and execution system in an embodiment of the present invention; Figure 4 This is a logical diagram of the dynamic priority arbitration mechanism of the present invention; Figure 5 This is a block diagram illustrating the principle of the closed-loop adaptive learning mechanism in an embodiment of the present invention. Figure 6 This is an example diagram of the human-computer interaction interface in an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] See attached document Figures 1 to 6 In the present invention, in step S100, the sensing network construction and data preprocessing module is responsible for the real-time acquisition, cleaning, fusion, and standardization of multi-source heterogeneous data of the paddy field ecosystem, providing high-quality, structured data input for subsequent decision-making and control modules. This step is specifically implemented through sub-steps S110 to S130.

[0021] In step S110, the system performs the acquisition of multidimensional heterogeneous data. This step deploys a three-dimensional sensing network covering key elements of the paddy field ecosystem. This network includes: Underwater sensor array: Multi-parameter water quality sensors are deployed at different depths and locations in paddy fields and surrounding ditches to monitor in real time dissolved oxygen (DO), pH value, water temperature, conductivity (EC), turbidity, oxidation-reduction potential (ORP), and the concentration of key ions such as ammonia nitrogen and nitrite.

[0022] Sediment sensor: A specially designed soil sensor probe is inserted into the sediment at different depths to monitor the temperature, moisture content, redox potential, and key vibration signals of the sediment, so as to indirectly assess the intensity of loach burrowing activities.

[0023] Field weather stations: Small weather stations are deployed on field ridges to collect environmental data such as light intensity, air temperature and humidity, wind speed and direction.

[0024] Image acquisition unit: High-definition cameras are installed at fixed high points and can be combined with unmanned aerial vehicles (UAVs) for regular patrols to collect images of the rice canopy and water surface. The above sensors transmit raw data to a local edge computing gateway or cloud server via low-power wide-area network (such as LoRa, NB-IoT) or fieldbus technology at a preset sampling frequency (e.g., water quality data once every 5 minutes, image data once per hour).

[0025] In step S120, the system preprocesses the acquired raw data to eliminate noise, fill in omissions, and correct errors. This step specifically includes: Outlier removal: The system uses statistical methods (such as the 3-σ criterion) to examine the collected sensor data sequences. If a data point deviates from the mean of the data within its time window by more than three times the standard deviation, the data point is identified as an outlier and is marked or removed.

[0026] Noise Filtering: For dynamic time-series data such as dissolved oxygen and pH values, which are susceptible to short-term disturbances, the system employs a Kalman filter algorithm. This algorithm, by establishing a state transition model, can estimate the true state of parameters from noisy observations, resulting in a smoother data sequence that more closely approximates the actual trend of change.

[0027] Missing data imputation: When a sensor malfunctions or the network is interrupted, resulting in missing data, the system initiates an appropriate imputation strategy based on the duration of the missing data and the characteristics of the data. For short-term missing data, linear interpolation or cubic spline interpolation is used to fill the gaps; for long-term missing data, a correlation model is used, such as an established water temperature-dissolved oxygen saturation physical model, to estimate the missing dissolved oxygen data based on the measured water temperature.

[0028] In step S130, the system performs multi-source data fusion, standardization, and vectorization. This step aligns the preprocessed data from different sensors in time and space, and integrates them into a unified comprehensive state vector. The process includes: Feature extraction: The system extracts quantitative features from the acquired image data using image processing algorithms. For example, it calculates the Normalized Difference Vegetation Index (NDVI) or Excess Green Index (ExG) of rice canopy images to characterize the growth and health of rice; and analyzes the texture and color of water surface images to assess the risk of algal blooms or oil films.

[0029] Data alignment and fusion: All preprocessed numerical data and extracted image features are tagged with a unified timestamp and integrated into a high-dimensional comprehensive state vector based on their spatial coordinates (e.g., the field number or ditch area to which they belong). The specific structure of this vector can be represented as follows: ; in, Representing the A state parameter, for example The dissolved oxygen concentration in region A. The pH value of area A. For example, the NDVI value of the rice canopy in zone B. This represents the total dimension of the state parameters.

[0030] Data standardization: To eliminate the influence of different parameter units and facilitate subsequent processing by machine learning models, the system standardizes vectors. Each element in the vector is linearly mapped to the interval [0,1] or [-1,1] using the min-max scaling method. Finally, the standardized composite state vector... The output is sent to the ecological-physiological state discrimination and coupling matching module as the basis for subsequent ecological-physiological state discrimination.

[0031] See attached document Figure 1 In step S200 of the method described in this invention, the ecological-physiological state discrimination and coupling matching module receives the comprehensive state vector from the sensing network construction and data preprocessing module. This is then transformed into a specific biological and physiological state, and the degree of matching between environmental conditions and target induction conditions is assessed. The specific implementation of this step includes sub-steps S210 to S240. In step S210, the system synthesizes the state vector... Feature engineering was performed to construct a real-time ecological-physiological state vector. This feature vector is a composite feature obtained by nonlinearly combining, performing time-series statistics, and reducing the dimensionality of the original sensor data. For example, it can be used to construct species behavior pattern features by calculating the vibration frequency of loach burrowing. The temporal variance of the ammonia nitrogen concentration is used to characterize its activity stability; environmental stress characteristics are constructed by measuring the ammonia nitrogen concentration. With water temperature Perform a product operation to obtain the potential toxicity pressure index. The underlying technology is based on the fact that the toxicity of ammonia nitrogen increases significantly with rising water temperature. It involves constructing interspecies interaction characteristics and calculating the plankton density index. With shrimp feeding intensity index The ratio of [value] is used to assess the supply and demand balance of natural bait resources. Composed of dozens of such composite features, it can more comprehensively and essentially reflect the current biological and ecological state of the system.

[0032] In step S220, the system utilizes a biological physiological state discrimination model. Will Mapped to specific biological physiological state categories Discriminant model These could be multilayer perceptrons (MLPs), support vector machines (SVMs), or long short-term memory networks (LSTMs) optimized for time-series data. Taking the LSTM model as an example, this model receives data over a continuous period of time... The sequence is used as input, and its internal forget gate, input gate, and output gate structure captures dynamic patterns of state changes, thereby predicting the specific physiological state of major farmed species such as loach and shrimp under the current time-series characteristics. The output of the discriminant model is shown below. It covers key growth and development stages (e.g., the pre-molting stage of shrimp and the active feeding period of loach), abnormal stress states (e.g., biological stress and sediment hypoxia stress), and ecological balance states (e.g., aquatic microecological stability). The discriminant function is expressed as follows: ; in, This indicates the category of the identified biological physiological state. Represents a real-time ecological-physiological state vector. This represents a biological physiological state discrimination model. The training of the discrimination model is conducted using supervised learning based on historical, labeled paddy field ecological data and biological monitoring results. This training process is a well-known technique and will not be elaborated upon here.

[0033] In step S230, the system queries the target guidance condition knowledge base. This knowledge base stores key target physiological states that the system can actively induce or needs to maintain. and the corresponding set of required environmental parameters . Organized in a structured data format (such as key-value pairs or database tables), each (For example, synchronous egg-carrying induction in freshwater shrimp and spawning and habitat induction in loach) all correspond to a precise range of environmental control parameters or target values, i.e. For example, regarding the target "synchronous egg-carrying shrimp", its The required water hardness (calcium-magnesium ion ratio), dissolved oxygen range, water redox potential, and stress hormone concentration will be clearly defined. The maximum permissible value, etc. The content of the knowledge base is based on a large amount of biological experimental data and the experience of experts in rice-fish integrated farming, and can be updated online through subsequent system closed-loop adaptive learning mechanisms. In step S240, the system calculates the matching degree between the current environmental state and the target induction conditions. When step S220 determines... When it falls into a category requiring active control, the system... Extract the corresponding and calculate and The degree of closeness. Matching degree. The calculation function is as follows: ; in, It is the current integrated state vector. The Middle Standardized values ​​of key environmental parameters; The first required for the target physiological state The optimal induced target value or median value for each environmental parameter; It is the first The weighting coefficients of each parameter reflect the biological criticality and regulatory sensitivity of the parameter to the induction process; This is the range of standardized deviations allowed by this parameter; This represents the total number of key environmental parameters. This exponentially decaying function ensures that an increase in deviation from any key parameter will lead to a decrease in the fit. The rapid decline demonstrates the sensitivity of biological systems to key environmental parameters. When the calculated... Below the preset matching threshold At that time, the system will send an intervention request to the proactive guidance and control and dynamic priority arbitration module, triggering the proactive control process.

[0034] See attached document Figure 1 In step S300 of the method described in this invention, the active induction regulation and dynamic priority arbitration module receives the biophysiological state discrimination result from the ecological-physiological state discrimination and coupling matching module. Matching degree Information, as well as long-term production targets from the year-round multi-cropping ecological configuration and benefit optimization engine. Based on this information, this module generates and publishes optimal control strategies that take into account the conflicts of multiple objectives. The specific implementation of this step includes sub-steps such as S310 to S330.

[0035] In step S310, the system first defines a set of operable control input vectors. This vector is used to directly regulate the paddy field ecosystem. It contains multiple fine-grained regulation parameters. This indicates the power or operating frequency of the intelligent bottom aeration / circulation system, used to regulate dissolved oxygen in the water and sediment disturbance. Indicates the dosage rate of functional microbial agents, used to improve the micro-ecological environment of water bodies and bottom sediments; It indicates the opening degree of the inlet gate or the flow rate of the inlet pump, controlling the replenishment and exchange of water. Indicates the opening degree of the sluice gate or the flow rate of the drainage pump, used for water discharge and water level regulation; This indicates the amount and frequency of feed administered to shrimp by the automated feeding system; and This represents the drip or spray rate of precision nutrient solutions or bio-inducers, used to supplement specific trace elements or induce biological physiological responses. This control input vector is represented as: ; in, , , , , , , All times The control input variables.

[0036] In step S320, the system introduces a dynamic priority weight vector. This is used to arbitrate potential conflicts between different species or ecological functions in a paddy field ecosystem. This vector can be defined, for example, as: ; in, , , These represent the current priority weights of rice, loach, and shrimp, respectively. and These represent the priority weights for soil health and water quality health, respectively. Each weight ranges from 0 to 1, indicating the relative importance of each objective at the current moment. Weight calculations are based on the risk level determined by the current eco-physiological state discrimination and coupling matching module, the long-term production plan provided by module 500, and preset biological rules. For example, when loaches exhibit hypoxic stress, This will significantly improve the situation when rice is in its critical growth period. The weights will increase accordingly. Weight calculations can be dynamically adjusted using the Sigmoid function or other non-linear mapping functions. This is based on the risk of soil compaction. For example, its corresponding soil health target weight The calculation formula is: ; in, This is a sensitivity coefficient, which determines the steepness of the weight change with risk; This is a preset threshold for soil compaction risk. When multiple objectives have high-priority requirements that conflict with each other (e.g., loaches require a low-dissolved-oxygen muddy environment, while shrimp require a high-dissolved-oxygen aquatic environment), the arbitration mechanism selects the objective that needs to be prioritized as the control target based on the current risk level of each objective, long-term benefit objectives, and niche relationships between species. This dynamic priority arbitration mechanism ensures the adaptability of the system's decision-making and its responsiveness to key biological / ecological events.

[0037] In step S330, the system generates the current optimal control command by solving a constrained optimization problem. The optimization problem aims to minimize the environment vector required for the current environment state and the high-priority objective. The deviation between the two is considered, along with the energy cost of regulation and the potential negative impact on secondary priority objectives. Objective function Represented as: ; in, Based on current control input And the environment state vector predicted by the system dynamic model at the next moment; It is based on the priority arbitration mechanism and The ideal environment vector required for the selected current high-priority target; Denotes the weighted L2 norm, where and It is a diagonal matrix, and its diagonal elements are the weights of the corresponding parameters; It is an ecological conflict penalty function, which quantifies the current regulatory actions. The potential negative impacts of low-priority objectives, such as significantly increasing water hardness to improve shrimp molting rates, might have a slight adverse effect on rice growth. The penalty value for this adverse effect will be calculated; It serves as a weighting coefficient, used to balance the relationship between target deviation, energy consumption control, and ecological conflict.

[0038] The ecological conflict penalty function The calculation method will be explained in detail. This function quantifies the current control operation. The potential negative impact on low-priority objectives. The specific calculation involves identifying the weights of all priorities. Secondary objectives of value Then calculate the current regulation Resulting environmental state changes This leads to negative deviations from these secondary objectives. For example, if the primary objective is "to increase water hardness to induce prawn molting," while the secondary objective is "to maintain a suitable low-hardness water environment for rice cultivation," then... The degree of inhibition caused by the increase in water hardness on the rice growth function will be calculated. One specific implementation is as follows: ; in, It is the set of all secondary objectives; This indicates the level of attention paid to secondary objectives; It is a predefined damage function used to quantify the effect of changes in environmental state on the target. The negative impact value caused.

[0039] This optimization problem is subject to multiple constraints. Physical constraints refer to... In other words, all controlled variables must be within the limits allowed by the physical constraints of the equipment. Biosafety constraints require that certain key environmental parameters must always be maintained within species-acceptable safe ranges; for example, dissolved oxygen concentration must not be lower than 2 mg / L, and pH must be maintained between 6.5 and 8.5 to avoid irreversible damage to any species. This optimization problem is typically solved online using a Model Predictive Control (MPC) framework. At each time step, MPC predicts the system dynamics over a future period and solves the optimization problem in a rolling fashion to obtain the first element of the optimal control sequence. The specific solution algorithms can be sequential quadratic programming (SQP), interior point method, or evolutionary algorithms, etc. These algorithms are well-known technologies in the field of control engineering and will not be elaborated here.

[0040] See attached document Figure 1 In step S400 of the method described in this invention, the environmental and biological community regulation execution system receives the optimal regulation instruction from the active induced regulation and dynamic priority arbitration module. This process transforms abstract control signals into concrete physical equipment actions, enabling sediment activation, water body regulation, and biological intervention in the paddy field ecosystem. The specific implementation of this step includes sub-steps S410 to S430.

[0041] In step S410, the system adjusts the substrate control component in the control input vector. and Perform substrate biological activation and improvement operations. For The module, via a variable frequency drive (VFD), adjusts the output frequency of a Roots blower or vortex pump. The generated compressed air is released through nanoporous aeration hoses laid at the bottom of the annular trench, forming a microbubble flow. This microbubble flow not only increases the dissolved oxygen content of the bottom water but also generates upward vertical flow, driving the exchange of interstitial water in the sediment and disrupting the reducing layer on the sediment surface. When the system indicates a need for high-intensity bottom disturbance (e.g., during loach habitat reconstruction), it activates a submersible thruster mounted on the bottom to generate a horizontal shear flow at a specific angle, moderately lifting surface sediments and promoting the suspension and oxidative decomposition of organic debris. The module controls the operation of the precision peristaltic pump or Venturi fertilizer applicator. The system injects pre-prepared liquid functional microbial agents (including but not limited to concentrated solutions of Bacillus subtilis, nitrifying bacteria, or photosynthetic bacteria) from a pre-set storage tank into the bottom aeration network or independent drip irrigation network at a calculated flow rate. In this way, the microbial agents can be evenly delivered to the surface of the paddy field sediment with the aid of aeration airflow or water flow, achieving targeted inoculation and enhancement of the sediment microbial community, and accelerating the conversion of ammonia nitrogen and nitrite.

[0042] In step S420, the system determines the hydrological control component based on the control input vector. , and This module performs precise regulation of the aquatic micro-ecology and hydrological environment. It utilizes electric valve positioners to accurately control the opening percentage of inlet and outlet gates, or controls flow rate by adjusting the speed of variable frequency pumps. The system employs a PID control algorithm, based on the real-time water level of the paddy field. To provide feedback, adjust dynamically. and This ensures that water levels are maintained within a target range that matches the rice's growth stages (shallow water during tillering, deep water during heading) and the needs of aquatic animals. For The instructions are primarily used to adjust the chemical composition of the water, especially during the sensitive molting period of shrimp. The system controls the metering pump to inject calcium and magnesium ion supplements, trace element mixtures, or pH buffers (such as limewater supernatant) from the storage tank into the inlet channel. To ensure uniform distribution of nutrients, the dosing rate is adjusted. With inflow rate Maintain the linkage ratio, that is: ; in, This is the concentration conversion factor. Furthermore, during high-temperature seasons or when the risk of water quality deterioration is high, this sub-step also includes activating the surface micro-sprinkler irrigation system to reduce canopy and water temperatures through evaporative cooling and to increase air humidity.

[0043] In step S430, the system determines the biological community intervention component based on the control input vector. And auxiliary commands to execute precise feeding and green disease prevention. For The module controls the speed of the 360-degree intelligent feeder's dispensing motor and the opening of the feeding valve. Based on the set feeding area radius and feed rate per unit time, the feeder evenly distributes the formulated feed to the areas where shrimp and loaches congregate. This feeding process is not a fixed program but can respond to a pulsed feeding strategy in the command, i.e., feeding small amounts multiple times in a short period to improve feed utilization and reduce bottom pollution from uneaten feed. Simultaneously, the module integrates control interfaces for a solar-powered insecticidal lamp and an insect pheromone trap. At night or under specific weather conditions, the system illuminates an insect-attracting light source with a wavelength of 365nm-400nm according to preset light control logic. This light attracts lepidopteran or coleopteran pests from around the rice paddy. As the pests are attracted to the light, they either touch the high-voltage grid or fall into the water, becoming natural, high-protein, active bait for shrimp and loaches, thus achieving an ecological cycle of rice-light-fish / shrimp and green pest control. For specific details regarding the mechanical structure of the feeder and the circuit principle of the insecticidal lamp, those skilled in the art can refer to existing agricultural machinery and plant protection equipment technologies, which will not be elaborated here.

[0044] In step S430, the system determines the biological community intervention component based on the control input vector. And auxiliary commands to execute precise feeding and green disease prevention. For The module controls the speed of the 360-degree intelligent feeder's dispensing motor and the opening of the feeding valve. Based on the set feeding area radius and feed rate per unit time, the feeder evenly distributes the formulated feed to the areas where shrimp and loaches congregate. This feeding process is not a fixed program but can respond to a pulsed feeding strategy in the command, i.e., feeding small amounts multiple times in a short period to improve feed utilization and reduce bottom pollution from uneaten feed. Simultaneously, the module integrates control interfaces for a solar-powered insecticidal lamp and an insect pheromone trap. At night or under specific weather conditions, the system illuminates an insect-attracting light source with a wavelength of 365nm-400nm according to preset light control logic. This light attracts lepidopteran or coleopteran pests from around the rice paddy. As the pests are attracted to the light, they either touch the high-voltage grid or fall into the water, becoming natural, high-protein, active bait for shrimp and loaches, thus achieving an ecological cycle of rice, light, and fish / shrimp, and green pest control. For specific details regarding the mechanical structure of the feeder and the circuit principle of the insecticidal lamp, those skilled in the art can refer to existing agricultural machinery and plant protection equipment technologies, which will not be elaborated here.

[0045] See attached document Figure 1 In step S500 of the method described in this invention, the annual multi-cropping ecological configuration and benefit optimization engine generates and dynamically optimizes a long-cycle production operation sequence based on meteorological forecast data, market supply and demand models, and ecological carrying capacity constraints. This step addresses the issue of time and resource allocation within the limited space of paddy fields by constructing a comprehensive objective function that incorporates ecosystem service value. The specific implementation of this step includes sub-steps S510 to S530.

[0046] In step S510, the system constructs a quantitative model for the comprehensive benefits throughout the year. This model differs from traditional models that only consider economic output as a single objective, by introducing ecosystem service value as a key evaluation indicator. The system calculates the comprehensive benefits for the entire year using a prediction algorithm. This benefit consists of the sum of the net economic gains from each cropping crop and the value of the ecosystem services provided by the paddy field ecosystem. The calculation formula is expressed as follows: ; in, This indicates the total number of crop rotations planned for the entire year; Number the stubble sections; Indicates the first The predicted yield of crop rotation products is estimated using a crop growth model based on historical accumulated temperature data, photosynthetically active radiation, and basic soil fertility values. Indicates the first The market forecast price of the crop rotation product is obtained by the system connecting to the agricultural product market database and predicting the price at a specific market launch time through the ARIMA (Autoregressive Differential Moving Average) model. Indicates the first The projected total production cost for each crop cycle includes seedlings, feed, labor, and energy inputs. It represents the quantitative value of the ecological service functions provided by the paddy field ecosystem, specifically including the value of farming cost savings converted from the soil permeability increase through loach disturbance, the value of pesticide reduction converted from the predation of pests by shrimp, and the value of water purification converted from the absorption of nitrogen and phosphorus by rice. It represents the weighting coefficient for converting the value of ecosystem services into economic value, used to balance short-term economic gains with long-term ecological sustainability.

[0047] Quantification of the aforementioned ecosystem service functions The calculation method will be explained in detail. It consists of multiple sub-items, and one specific implementation is as follows: ; in, The water purification value is calculated by multiplying the total amount of nitrogen and phosphorus absorbed by the system throughout the entire production cycle by the social governance cost per unit of nitrogen and phosphorus. To determine the value of biological pest control, the difference between the predicted pest occurrence and the actual monitoring value is calculated to offset the corresponding pesticide cost savings. To assess the value of soil improvement, the monitoring of improvements in indicators such as soil organic matter content and aggregate structure is used to calculate the potential benefits of future farming cost savings and improved soil fertility.

[0048] In step S520, the system performs multi-objective constraint optimization to determine the optimal production sequence. and time nodes Optimize variables Defined as a series of species combinations with different crop rotations throughout the year (such as early rice + shrimp rotation, mid-season rice + loach co-cultivation) and their corresponding stocking densities. Defined as the specific start time of each cropping season. and end time This optimization process must meet strict physical and biological constraints. The time continuity constraint requires that sufficient time be allowed between adjacent crop rotations for soil remediation or substrate treatment. , is represented as: ; Environmental carrying capacity constraints require that the total environmental load for the entire year must not exceed the self-purification threshold of the paddy field ecosystem. For example, the total nitrogen and phosphorus emission constraint is expressed as: ; Biological growth physiological constraints require that the duration of each crop rotation must meet the effective accumulated temperature required for crops or aquatic animals to complete a specific growth stage. Its integral constraint form is: ; in, For the predicted real-time temperature, This represents the biological zero degree of the species. For the aforementioned nonlinear programming problem, this embodiment of the invention employs a non-dominated sorting genetic algorithm (NSGA-11) with an elitist strategy to solve it. The algorithm generates a Pareto optimal solution set through population initialization, crossover, mutation, and non-dominated sorting. The system automatically selects a solution from the solution set as the initial annual production plan based on a preset risk preference strategy (such as risk-averse or risk-neutral).

[0049] In step S530, the system implements rolling revisions and dynamic closed-loop management of the plan. The year-round multi-cropping ecological configuration and benefit optimization engine does not generate a one-time plan at the beginning of the year, but establishes a rolling time window mechanism. During system operation, this module periodically (e.g., weekly or bi-weekly) receives actual environmental data and biological growth data from the sensing network construction and data preprocessing module. When the deviation between the actual observed values ​​(such as actual accumulated temperature, actual market price fluctuations) and the initial model predictions exceeds a preset threshold... At this point, the system triggers a replanning mechanism. The replanning process uses the remaining time window after the current moment as a new optimization cycle, uses the latest state data as initial conditions, and re-runs the optimization algorithm in step S520 to fine-tune the type, density, or market launch time of subsequent cropping. The corrected long-term target parameters are transmitted in real time to module 300: the active induced regulation and dynamic priority arbitration module, as a guiding benchmark for short-term regulation strategies.

[0050] See attached document Figure 1 In step S600 of the method described in this invention, the system's closed-loop adaptive learning mechanism performs post-evaluation of the regulatory effect and online evolution of the system's internal logic. This module constructs a feedback loop from the execution result to the decision model, ensuring that the continuously running system can adapt to the time-varying characteristics of the paddy field ecological environment and the succession patterns of the biological community. The specific implementation of this step includes sub-steps S610 to S630.

[0051] In step S610, the system performs quantitative monitoring and deviation assessment of the regulatory effect. The environmental and biological community regulation execution system executes regulatory commands. The subsequent preset time window Internally, the system's closed-loop adaptive learning mechanism continuously tracks the comprehensive state vector. The actual change trajectory. The system will record the actual observed environmental response state. The theoretical state predicted by the system dynamic model Compare the results and calculate the prediction error vector. Simultaneously, the system assesses whether the control objectives have been achieved, that is, it calculates the actual state versus the target-induced conditions. The Euclidean distance between them. To comprehensively evaluate the effectiveness of regulation, the system calculates the effectiveness score of a single regulation. The calculation formula is as follows: ; in, Characterizes the control accuracy, reflecting whether the environment has successfully converged to the target state; Characterizes the energy consumption or material costs incurred in regulation; This refers to the system response time, which is the time span required for the environmental state to reach the target value. These are the corresponding weighting coefficients. This score... It is recorded in the historical control database as a reward signal for subsequent strategy optimization.

[0052] In step S620, the system uses an online learning algorithm to correct internal model parameters to eliminate model bias. This applies to the biological physiological state discrimination model in the eco-physiological state discrimination and coupling matching module. In the environmental dynamic prediction model within the active-induced regulation and dynamic priority arbitration module, the system employs recursive least squares (RLS) or stochastic gradient descent (SGD) for parameter updates. When the prediction error vector... When the norm exceeds the allowable threshold, it indicates that the current system dynamics model fails to accurately describe the physical / biochemical evolution of the paddy field environment (e.g., changes in wind resistance coefficient due to rice growth or changes in oxygen consumption due to sediment aging). The system utilizes the current input data. and observation data As new training samples, the model parameter matrix Perform calibration. Parameter updates follow these rules: ; in, For the first The model parameter set at the next iteration; The learning rate is set to a small positive number to ensure the stability of the update process; For loss function The gradient of the parameters (typically the mean squared error of the prediction error). Through this mechanism, the system can capture the slow drift (Driftinq) of environmental parameters with seasonality or cropping years, ensuring prediction accuracy over long-term operation.

[0053] In step S630, the system uses the target induction condition knowledge base. The system iteratively optimizes the priority arbitration rules. This step enables the self-accumulation of knowledge. When the system recognizes that a certain regulation has caused an organism to produce an ideal physiological response (e.g., detecting synchronous molting signals in a group of shrimp or rapid weight gain in loaches), the system records the corresponding actual environmental parameter vector at that moment. These are marked as positive samples. The system uses a moving average update strategy to fine-tune the target-inducing conditions stored in the knowledge base. This gradually brings it closer to the actual optimal operating condition. The update formula is: ; in, and These are the target environment parameters before and after the update, respectively. The update factor, typically ranging from 0.01 to 0.05, is used to smooth out random fluctuations. In addition, the system also... Historical cumulative values ​​are used to adjust the risk threshold for dynamic priority arbitration in the proactive guidance and control and dynamic priority arbitration modules. If historical data shows that no intervention at a certain risk level would lead to severe subsequent ecological losses (i.e., If the value remains low for an extended period, the system will reduce... This makes the system more sensitive and aggressive in similar future situations; conversely, it raises the threshold to reduce unnecessary intervention.

[0054] See attached document Figure 1 In step S700 of the method described in this invention, the human-computer interaction and management interface provides system administrators or technicians with a visual operation platform that integrates monitoring, intervention, and management. This module is a key node that combines autonomous system decision-making with human expert experience, ensuring the transparency, controllability, and reliability of the entire production management process. The specific implementation of this step includes sub-steps S710 to S730.

[0055] In step S710, the system provides a multi-dimensional visualization of the paddy field ecosystem. This module integrates real-time and historical data collected by the sensing network construction and data preprocessing module, as well as the biological physiological states determined by the active induction regulation and dynamic priority arbitration module, into a dynamic, GIS-driven "digital twin" paddy field interface. On this interface, managers can intuitively see: Spatial distribution map: Using a 2D or 3D map of the paddy field as a base, the spatial distribution of key environmental parameters (such as dissolved oxygen, water temperature, and soil redox potential) in different areas is rendered in real time through heat maps, helping managers to quickly locate problem areas.

[0056] Equipment Status Diagram: The location of various sensors, aeration discs, water pumps, feeders and other operating equipment is marked on the map in the form of icons, and their current working status (such as running, standby, fault) is indicated by different colors or animation effects.

[0057] Status trend curve: Managers can select any one or more monitoring parameters to view their trend curves over a period of time (such as 24 hours, a week, or a month), and can overlay and compare curves of different parameters (such as ammonia nitrogen concentration and aerator operating power) to analyze their correlation.

[0058] Biological state label: The system combines the biological physiological state determined by the ecological-physiological state discrimination and coupling matching module. (Situations such as the molting period of freshwater shrimp and anaerobic conditions in the bottom sediment) are displayed on the interface in the form of eye-catching text labels, replacing the complex data interpretation process. In step S720, the system provides an interpretable display of the intelligent alarm and decision-making process. When the system detects that key parameters exceed the safety threshold, or the ecological-physiological state discrimination and coupling matching module identifies a high-risk biological stress state, the system sends alarm information to the manager through preset methods such as mobile application push, SMS, or on-site audible and visual alarms. This alarm information not only includes a risk description (e.g., dissolved oxygen in area A of the ring ditch is below 2.0 mg / L, indicating a risk of oxygen deficiency for freshwater shrimp), but also includes a recommended control strategy generated by the active induction regulation and dynamic priority arbitration module (e.g., starting micropore aeration in area A, setting the power to 75%, and continuing for 30 minutes). More importantly, this module provides a decision tracing function, allowing managers to query why the system made this decision. The interface will display the dynamic priority weight vector in the active induction regulation and dynamic priority arbitration module in the form of flowcharts or text. The current value, and the ecological conflict penalty function. The calculation results enable managers to understand the system's trade-off and decision-making logic in multi-objective conflicts.

[0059] In step S730, the system provides manual intervention and management permission configuration functions. The system supports at least two intervention modes: Manual control mode: In case of emergency or equipment maintenance, the administrator can bypass the system's autonomous decision-making logic and directly send control commands to any of the execution devices in module 400 through virtual switches, sliders or numerical input boxes on the operation interface, such as forcibly turning on or off a water pump.

[0060] Strategy Parameter Adjustment Mode: Managers can fine-tune the parameters of the system's high-level decision-making model based on their own experience or external market information. For example, managers can adjust the market forecast prices for different cropping products in the interface of the year-round multi-cropping ecological configuration and benefit optimization engine. Or adjust the weighting coefficient of ecosystem service value. This guides the long-term optimization engine to generate production plans that better align with its current business objectives.

[0061] In addition, managers can also utilize the target-inducing knowledge base of the eco-physiological state discrimination and coupling matching module. In the management interface, you can review and modify the environmental parameters required for specific physiological states. Alternatively, in the feedback interface of the system's closed-loop adaptive learning mechanism, the actual effect of a certain adjustment can be manually labeled (e.g., excellent, good, poor), providing a channel for the input of subjective experience and the accelerated learning of the model. For the software development of the user interface, the construction of the database, and the implementation of the communication protocol, those skilled in the art can use general Web technologies (such as HTML5, Vue.js, ECharts) and backend technologies (such as Python, Java) combined with industrial communication standards (such as MQTT, OPCUA) to complete this task. The specific implementation is well-known in this field and will not be elaborated upon here.

Claims

1. A method for managing a rice, loach and shrimp ecological breeding system throughout the year, comprising the following steps: collecting multi-dimensional data through a perception network deployed in the rice, loach and shrimp ecological breeding system, and pre-processing the multi-dimensional data to obtain a comprehensive state vector representing the current state of the system; based on the comprehensive state vector and a pre-set target induction knowledge base, determining the physiological state of the organisms in the system, and calculating a coupling matching degree reflecting the gap between the current state of the system and the ideal state; performing active induction regulation and dynamic priority arbitration to generate optimal regulation instructions for regulating the rice, loach and shrimp ecological breeding system; executing the optimal regulation instructions by an environment and biological community regulation execution system deployed in the system to physically intervene in the system environment or biological community.

2. The method according to claim 1, wherein the rice-fish-shrimp ecological species breeding system is characterized in that, The step of calculating the coupling matching degree specifically comprises: comparing and quantitatively evaluating the comprehensive state vector with a plurality of ideal state patterns stored in the target induction knowledge base.

3. The method according to claim 1, wherein the rice-fish-shrimp ecological species breeding system is managed throughout the year. The step of performing active induction regulation and dynamic priority arbitration is triggered when the coupling matching degree is lower than a pre-set matching degree threshold.

4. The method according to claim 1, wherein the rice-fish-shrimp ecological species breeding system is managed throughout the year. The step of performing active induction regulation and dynamic priority arbitration specifically comprises: based on the physiological state, the long-term production target of the system and the real-time environmental risk, calculating a dynamic priority weight vector representing the priority of different regulation targets; based on the dynamic priority weight vector and the comprehensive state vector, constructing a constrained optimization problem; by solving the constrained optimization problem, the optimal regulation instructions are obtained.

5. The method according to claim 4, wherein the rice-fish-shrimp ecological species breeding system is managed throughout the year. The objective function of the constrained optimization problem includes the weighted sum of at least two of the following: target deviation between the expected regulation result and the regulation target; regulation cost required for executing the optimal regulation instructions; ecological conflict penalty caused by interest conflicts between different biological populations due to regulation actions.

6. The method according to claim 1, wherein the rice-fish-shrimp ecological species breeding system is managed throughout the year. The method further comprises the steps of: performing annual multiple-crop ecological configuration and benefit optimization to generate a long-term production target including a production operation sequence and a time scheduling scheme, and providing the long-term production target to the step of performing active induction regulation and dynamic priority arbitration as a decision basis.

7. The method according to claim 6, wherein the rice-fish-shrimp ecological species breeding system is managed throughout the year. The step of performing annual multiple-crop ecological configuration and benefit optimization further comprises: in the production cycle, dynamically adjusting the production operation sequence and the time scheduling scheme using a rolling correction mechanism according to the actual production progress and external environmental changes.

8. The method according to claim 1, wherein the rice-fish-shrimp ecological species breeding system is managed throughout the year. The method further comprises the steps of: performing closed-loop adaptive learning, which receives the optimal regulation instructions and actual state data collected by the perception network after executing the instructions, calculates the performance score and prediction error of this regulation, and updates the system based on the performance score and the prediction error.

9. The method according to claim 8, wherein the rice-fish-shrimp ecological species breeding system is managed throughout the year. The update process of the closed-loop adaptive learning specifically comprises at least one of the following: updating the target induction knowledge base; correcting the model parameters or arbitration rules used to construct the constrained optimization problem.

10. The all-year-round production management system of the rice- eel-shrimp eco-species breeding system according to any one of claims 1 to 9, characterized by, comprises: a data pre-processing module for collecting multi-dimensional data through a perception network and pre-processing the multi-dimensional data to obtain a comprehensive state vector representing the current state of the system; The state discrimination module is configured to discriminate the physiological state of the organism in the system and calculate a coupling matching degree based on the comprehensive state vector and a preset target induction knowledge base. The dynamic priority arbitration module is configured to perform active induction regulation and dynamic priority arbitration to generate optimal regulation instructions. The environment and organism community regulation execution system is configured to execute the optimal regulation instructions.

Images