River crab feed feeding mixed monitoring method, system, equipment and medium

Abstract

The invention provides a river crab feed feeding mixed monitoring method, system and device and a medium, and belongs to the field of aquaculture intelligent monitoring. Data is collected and normalized, and a chain type data collection system is constructed; the edge computing and block chain technology is adopted to realize data distributed storage and transmission; a fuzzy neural network hybrid model is constructed, fuzzification processing, feature mining and defuzzification output are integrated, and a feeding amount prediction value is generated; associating the feed proportioning library with the knowledge graph, and iteratively training the model and generating a multi-target feeding decision parameter; feeding equipment is controlled to execute actions according to the decision parameters, and the real-time monitoring model outputs and triggers an early warning signal; and updating the knowledge graph and the feed matching library based on the feedback data to form closed-loop self-adaptive regulation and control. According to the method, various factors can be comprehensively considered, scientific and intelligent feed feeding decision is realized, feed waste and insufficient feeding are effectively avoided, and the growth quality of the river crabs is improved.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent monitoring of aquaculture, and in particular relates to a method, system, equipment and medium for monitoring the feeding and mixing of river crab feed. Background Art

[0002] River crab farming occupies an important position in the development of the aquaculture industry. River crab farming has gone through a historic stage from natural breeding to raising high-quality crabs. Traditional breeding methods are increasingly restricting the scale and standardization development of river crab farming.

[0003] In river crab farming using related technologies, decisions on feeding crab feed are mostly based on experience and cannot adapt to the complex and changing environment of river crab farming and the uncertainty of its growth status. For example, river crabs have different feed requirements at different growth stages, and environmental factors such as temperature and water quality also affect their feeding and growth. However, traditional models are difficult to accurately consider these factors, resulting in unreasonable feeding amounts, which may lead to feed waste or insufficient feeding, affecting the growth and development of river crabs. Traditional river crab farming monitoring systems lack effective feedback mechanisms and self-optimization capabilities. Once the farming strategies and model parameters are set, it is difficult to dynamically adjust them according to actual farming conditions. As river crabs grow and the environment changes, the original strategies and models may no longer be applicable, but the system cannot perceive and adapt to these changes in a timely manner, resulting in inefficient farming management. Summary of the Invention

[0004] The present invention provides a method for monitoring the feeding and mixing of river crab feed. The method can comprehensively consider multiple factors, realize scientific and intelligent feed feeding decision-making, effectively avoid feed waste and insufficient feeding, and improve the growth quality of river crabs.

[0005] Methods include: S101: Collect and normalize air, water quality, crab growth, and geographic information data through sensors to build a chain data collection system; S102: Use edge computing and blockchain technology to achieve distributed data storage and transmission, and complete multi-source data governance and feature fusion; S103: Constructing a fuzzy neural network hybrid model, integrating fuzzification processing, feature mining and defuzzification output, and generating a feeding amount prediction value; S104: Associating the feed ratio library with the knowledge graph, iteratively training the model and generating multi-objective feeding decision parameters; S105: Control the feeding equipment to execute actions according to the decision parameters, monitor the model output in real time and trigger an early warning signal; S106: Update the knowledge graph and feed ratio library based on the feedback data to form a closed-loop adaptive control.

[0006] It should be further explained that step S103 specifically includes: storing the full cycle data of river crab farming in the blockchain farming cloud platform, which integrates the adaptive consensus algorithm and the hierarchical data protection system; For each river crab farming monitoring node j, the contribution value is calculated using the comprehensive indicator formula Ej=α1×Bj+α2×Sj+α3×Pj; Among them, Bj represents the aquaculture data collection accuracy index of the river crab aquaculture monitoring node, Sj is the parameter of the aquaculture water area that can be covered, Pj represents the parameter of the data processing capacity per unit time, and α1, α2, and α3 are weighting coefficients dynamically configured according to the aquaculture stage; The aquaculture contribution index Cj is defined and obtained by the formula Cj=γ1×Vj+γ2×Nj+γ3×Fj, where Vj is the accuracy of crab growth data monitoring in the previous statistical period of the crab aquaculture monitoring node, Nj is the number of times of abnormal aquaculture data warning, Fj is the effective number of times the node assists in optimizing the feeding strategy, and γ1, γ2, and γ3 are adjustment coefficients used to balance different aquaculture contribution dimensions.

[0007] It should be further explained that, after step S103, the following steps are also included: Based on the collected data, the equipment operating parameters and the correlation of historical events are determined, and a data priority evaluation formula is constructed. The data priority evaluation formula is Pscore=∑(wenv·Eparam)+∑(whist·Hfactor). Pscore represents the data priority score, wenn is the weight assigned based on the impact of the environment and equipment on river crab growth, Eparam is the aquaculture environment and equipment operating parameters, Hfactor represents the correlation of historical events, and whist is the weight assigned based on the impact scope and frequency of historical events. According to the priority score of river crab farming data and the timeliness requirements of data, the data transmission priority is set. According to the priority score Pscore and timeliness requirements, the priority Tle of each data transmission task is determined; Using the optimization algorithm, combined with the current data transmission status Ts and the priority Tle of each data transmission task, the transmission resource allocation Rd is dynamically adjusted, Rd = opt.min(∑Ttr(Tta,Ts)); Among them, Rd is the transmission resource allocation after dynamic adjustment, Ts is the current transmission status, Tta represents the i-th data transmission task, and Ttr represents the task data transmission time.

[0008] It should be further explained that step S104 specifically includes: Load the optimal feed formula for different growth stages of river crabs from the historical database; Retrieve the nutritional demand model and extract rules related to the current environmental growth status from the knowledge graph; The rules include: environmental constraints, health linkages, and multi-objective priorities; Convert knowledge graph rules into numerical features; Normalize the sensor data and convert the knowledge graph rules into If-Then statements in the fuzzy rule base; According to the priority rules of the knowledge graph, the weight of the fuzzy rule base is dynamically adjusted, and the membership function of the fuzzy set is defined using trigonometric functions; Mapping fuzzy rules to the rule layer of fuzzy neural network to form a dynamic update mechanism of rule base; Define the objective function based on maximizing the average daily weight gain of river crabs, minimizing the amount of feed, and maintaining water quality parameters within safe thresholds; The normalized data in step S101 and the numerical features of rule trigger probability and priority weight are used as input; Calculate the decision parameters according to the following formula and adjust the dynamic weight of the priority of the knowledge graph; Decision parameters = w1⋅growth optimization value + w2⋅resource conservation value + w3⋅environmental control value; w1 is the weight of growth optimization value, w2 is the weight of resource conservation value, and w3 is the weight of environmental control value; The feeding amount decision is based on the fuzzy neural network output and converted into the specific feeding amount value through the center of gravity method.

[0009] It should be further explained that, after step S103, the following steps are also included: A dual-path deep feature extraction framework is constructed. The first path uses an optimized feature extraction network to process image data. Local features are extracted through the convolution layer, the pooling layer performs dimensionality reduction, and the fully connected layer integrates features. The calculation process is expressed as follows: Img_Feature=FCLayer(PoolLayer(ConvLayer(ImageData))) ImageData is the input aquaculture environment or crab morphology image data, and the image feature vector is obtained through multi-level convolution, pooling and full connection operations; The second path uses a gated recurrent unit network to perform feature mining on time series data. Time series information (TimeSeriesData), such as water quality monitoring data and feeding equipment operation logs, is input into the gated recurrent unit. The gated mechanism selectively memorizes and updates information, and outputs a time series feature vector: Seq_Feature=GRU(TimeSeriesData); Based on the data type importance evaluation results, the two feature vectors are weighted fused. The fusion formula is: Merged_Feature=α×Img_Feature+(1-α)×Seq_Feature Among them, α is an adaptive weight parameter, which is dynamically adjusted by analyzing the criticality of the data in the current breeding stage to achieve the optimal combination of features; A classification decision system is constructed based on the fused feature vector, and the normalized exponential function is used to calculate the probability distribution of each breeding status category: Probability(Class|Feature)=Softmax(WeightMatrix×Merged_Feature+BiasVector) Among them, WeightMatrix is ​​the classifier weight matrix, BiasVector is the bias vector, and through matrix operations and Softmax function, the probability value of each category under the current feature is output.

[0010] The classification system was supervised and trained using the cross-entropy loss criterion. Based on the labeled farming status dataset, the model parameters were optimized by minimizing the difference between the predicted probability and the true label.

[0011] It should be further explained that in step S106, based on the key breeding information extracted by the deep learning model, a river crab breeding status analysis model is built. The real-time sensor data and historical breeding data are input into the river crab breeding status analysis model, and the data analysis algorithm is used to mine the potential regularities and abnormal patterns in the data: Among them, the crab breeding status index is defined as CSI and the calculation formula is: CSI=a1·m(Fc)+a2·n(Wt)+a3·p(Ds); Among them, a1, a2, and a3 are weight coefficients, reflecting the degree of influence of each parameter on the crab rearing status; m(Fc), n(Wt), and p(Ds) are functions of the corresponding parameters, which convert the parameters into the impact score on the crab rearing status; Fc represents the nutritional content of the feed, Wt represents the weight of the crab, and Ds represents the turbidity of the water; Introducing the new environmental impact coefficient EnviCoef, the adjusted crab breeding status index CSde is: CSde=CSI·EnviCoef Wherein, CSde represents the adjusted feeding status index, and EnviCoef represents the environmental impact coefficient; Based on historical river crab breeding data and analysis of problem patterns that occur during the feeding process, a new warning threshold CSId for the river crab breeding status index is set; when the adjusted river crab breeding status index CSde is less than the warning threshold CSId, an early warning signal is issued and a notification is sent to the breeding staff.

[0012] It should be further explained that the method also constructs a multi-level knowledge graph, which includes the basic data level and the knowledge graph level; Divide the hierarchical structure according to the crab breeding data categories; construct a knowledge graph that includes equipment operation logic, growth cycle rules and abnormal causal relationships, and use physical equations and expert experience rules to map data layer information into executable knowledge nodes; Generate complex logical expressions based on the node attributes and relationships of multi-level knowledge graphs and real-time data streams; define retrieval strategies and execute them through pre-compiled instruction sets; Dependency parsing is used to extract keywords and map them to metadata tags at the basic data level. A word embedding model is used to convert natural language queries into high-dimensional vectors, which are then matched with the feature vectors of knowledge graph nodes for similarity. Based on the reasoning mechanism of the rule engine, dynamic logical judgment and operation chain are executed; Based on the parsed data vector, locate the associated nodes in the knowledge graph hierarchy; Call the action instructions in the predefined rule library to generate multimodal analysis results.

[0013] The present application also provides a crab feed mixing monitoring system, the system comprising: The data acquisition module is used to collect and normalize air, water quality, river crab growth and geographic information data through sensors to build a chain data collection system; The data fusion module is used to implement distributed data storage and transmission using edge computing and blockchain technology, and complete multi-source data governance and feature fusion; Feeding amount prediction module is used to build a fuzzy neural network hybrid model, integrating fuzzy processing, feature mining and defuzzification output to generate feeding amount prediction values; The feeding decision module is used to associate the feed ratio library with the knowledge graph, iteratively train the model and generate multi-objective feeding decision parameters; The monitoring module is used to control the feeding equipment to perform actions according to the decision parameters, monitor the model output in real time and trigger early warning signals; The self-regulation module updates the knowledge graph and feed ratio library based on feedback data, forming a closed-loop adaptive regulation.

[0014] According to another embodiment of the present application, an electronic device is provided, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the crab feed mixing monitoring method when executing the program.

[0015] According to another embodiment of the present application, a storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the crab feed feeding and mixing monitoring method are implemented.

[0016] It can be seen from the above technical solutions that the present invention has the following advantages: The hybrid monitoring method for feeding river crab feed provided in this application constructs a fuzzy neural network hybrid model and a dual-path deep feature extraction framework, combined with a classification decision system and a supervised training mechanism. The fuzzy neural network hybrid model can process uncertain data in the breeding process and accurately predict the feeding amount; the dual-path deep feature extraction framework fully exploits the characteristics of image and time series data to improve the classification accuracy of the breeding status. It enables the system to comprehensively consider multiple factors, realize intelligent feed feeding decisions, and effectively avoid feed waste and insufficient feeding. Edge computing and blockchain technology are used. Edge computing reduces data transmission volume and delay, and improves the system response speed; blockchain technology realizes distributed storage and secure transmission of data, ensuring the immutability and credibility of data.

[0017] By linking the feed ratio library with the knowledge graph, a closed-loop adaptive control mechanism is established. The system continuously iterates the training model based on feedback data, updates the knowledge graph and feed ratio library, and realizes dynamic adjustment of breeding strategies and model parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for the description. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1 Flow chart of the feeding and mixing monitoring method for river crab feed; Figure 2 This is a schematic diagram of the crab feed mixing monitoring system; Figure 3 This is a schematic diagram of an embodiment of a crab feed mixing monitoring system; Figure 4 Schematic diagram of an electronic device. DETAILED DESCRIPTION

[0020] This application provides a method for developing a system for monitoring the feeding of river crabs by integrating industrial internet technologies to improve the management of river crab feeding. This method enables prediction, early warning, and precise intervention of river crab feed consumption, growth status, and feeding strategies, thereby improving resource utilization and aquaculture management efficiency.

[0021] The following describes in detail the steps of the method for monitoring the mixing and feeding of river crab feed involved in this application. Specific details such as specific system structures and technologies are provided for illustration rather than limitation to facilitate a thorough understanding of the embodiments of this application. However, it should be clear to those skilled in the art that this application can also be implemented in other embodiments without these specific details.

[0022] It should be understood that when used in this specification, the term "comprising" indicates the presence of the described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or their collections. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

[0023] The phrases "one embodiment" or "some embodiments" described in this application mean that the specific features, structures, or characteristics described in the embodiment are included in one or more embodiments of the application. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in other embodiments," etc. that appear in different places in this application do not necessarily refer to the same embodiment, but rather mean "one or more but not all embodiments," unless otherwise specifically emphasized.

[0024] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0025] See also Figure 1 The figure is a flow chart of a method for monitoring the feeding and mixing of river crab feed in a specific embodiment, the method comprising: Step S101: Collect and normalize air, water quality, crab growth and geographic information data through sensors to build a chain data collection system.

[0026] In some embodiments, various sensors are installed within the crab farming area. These sensors include temperature and humidity sensors and wind speed and direction meters for monitoring the air environment; dissolved oxygen meters, pH meters, and ammonia nitrogen detectors for monitoring water quality; image acquisition equipment for monitoring crab growth; individual physiological parameter sensors; and GPS positioning devices for acquiring geographic information. After the sensors collect data in real time, a data preprocessing module normalizes the data, converting data of varying types and magnitudes into a unified format and range. By connecting these sensors through Internet of Things technology, a chained data collection system is established, enabling data to be collected, transmitted, and stored in a predefined logical sequence.

[0027] Step S102: Use edge computing and blockchain technology to achieve distributed data storage and transmission, and complete multi-source data governance and feature fusion.

[0028] This embodiment utilizes edge computing nodes to perform preliminary processing and analysis of raw data at the data collection site, reducing data transmission volume and latency. Simultaneously, the processed data is uploaded to the blockchain platform via 5G and other communication technologies. On the blockchain platform, distributed storage technology is used to store data across multiple nodes, ensuring data security and reliability. Data governance technologies are used to clean, integrate, and annotate multi-source data, achieving data standardization and regularization. Feature fusion algorithms are used to extract and fuse features from different data types to form more representative comprehensive features.

[0029] Step S103: construct a fuzzy neural network hybrid model, integrate fuzzification processing, feature mining and defuzzification output, and generate a feeding amount prediction value.

[0030] The fuzzification processing of this embodiment is to convert the input data into fuzzy sets, and determine the degree to which each input data belongs to each fuzzy set by defining a membership function. For example, the dissolved oxygen content of water quality is divided into three fuzzy sets of low, medium and high, and the membership corresponding to each dissolved oxygen value is determined. The feature mining of this embodiment is to extract valuable features from the fuzzified data and discover the potential patterns and laws in the data through the learning ability of the neural network. The defuzzification output is to convert the result obtained by fuzzy reasoning into an accurate numerical value as the final feeding amount prediction value. The defuzzification method can adopt the weighted average method and the center of gravity method.

[0031] In this way, the model fuzzifies the input data on air quality, water quality, crab growth, and geographic information to obtain the membership degree of a fuzzy set. A neural network is then used to learn and analyze these fuzzy features, uncovering underlying patterns in the data. Finally, a defuzzification method is used to convert the neural network output into an accurate feeding amount prediction. This fuzzy neural network hybrid model can handle uncertainty and imprecision in the data, making it more suitable for the complex and changing environment of crab farming.

[0032] As an embodiment of the present application, step S103 specifically includes: storing the full cycle data of river crab farming in a blockchain farming cloud platform, and the blockchain farming cloud platform integrates an adaptive consensus algorithm and a hierarchical data protection system; For each river crab farming monitoring node j, the contribution value is calculated using the comprehensive indicator formula Ej=α1×Bj+α2×Sj+α3×Pj.

[0033] Among them, Bj represents the aquaculture data collection accuracy index of the river crab aquaculture monitoring node, Sj is the coverable aquaculture water area parameter, Pj represents the data processing capacity parameter per unit time, and α1, α2, and α3 are weighting coefficients dynamically configured according to the aquaculture stage.

[0034] The aquaculture contribution index Cj is defined and obtained by the formula Cj=γ1×Vj+γ2×Nj+γ3×Fj, where Vj is the accuracy of crab growth data monitoring in the previous statistical period of the crab aquaculture monitoring node, Nj is the number of times of abnormal aquaculture data warning, Fj is the effective number of times the node assists in optimizing the feeding strategy, and γ1, γ2, and γ3 are adjustment coefficients used to balance different aquaculture contribution dimensions.

[0035] Data from the entire crab farming cycle is stored on a blockchain-based farming cloud platform that integrates an adaptive consensus algorithm and a hierarchical data protection system. Traditional data storage methods struggle to ensure data security and immutability, but the introduction of blockchain technology fundamentally changes the data storage model. The adaptive consensus algorithm dynamically adjusts participating nodes based on the farming scenario. Compared to traditional fixed consensus mechanisms, it better adapts to the real-time and dynamic nature of crab farming data.

[0036] Taking a large crab breeding base as an example, the specific implementation process of step S103 is as follows: within the breeding base, there are 50 crab breeding ponds, and multiple crab breeding monitoring nodes are deployed around each pond. These nodes are responsible for collecting data such as water temperature, dissolved oxygen, and crab growth images, and performing preliminary processing and transmission of this data. The breeding base has established a blockchain breeding cloud platform, which records the entire crab life cycle, from crab seedling release, growth and development to mature harvest.

[0037] For each monitoring node j, the base calculates its contribution value using the comprehensive indicator formula Ej = α1 × Bj + α2 × Sj + α3 × Pj. For example, node A's aquaculture data collection accuracy indicator Bj reaches 98%, the covered aquaculture water area parameter Sj is 10 mu, and the data processing capacity parameter per unit time Pj is 1000 data records per hour. During the juvenile stage of river crabs, α1 = 0.4, α2 = 0.3, and α3 = 0.3 are set. The calculated contribution value Ej = 0.4 × 98 + 0.3 × 10 + 0.3 × 1000 = 342.2.

[0038] Node contributions are also evaluated using the aquaculture contribution index Cj = γ1 × Vj + γ2 × Nj + γ3 × Fj. During the past statistical period, node A's accuracy rate (Vj) for monitoring crab growth data was 95%, its number of abnormal aquaculture data alerts (Nj) was 5, and its number of effective assistance in optimizing feeding strategies (Fj) was 3. Setting γ1 = 0.3, γ2 = 0.4, and γ3 = 0.3, the calculated aquaculture contribution index (Cj) is 0.3 × 95 + 0.4 × 5 + 0.3 × 3 = 31.4. As crabs enter their rapid growth phase, the real-time data flow in the aquaculture area increases significantly, increasing the requirements for data processing and consensus efficiency. At this point, the system activates a dynamic node screening mechanism based on the real-time data from the aquaculture area and the characteristics of the crab's growth cycle. Using a fuzzy decision-making algorithm, the number of monitoring nodes participating in consensus is evaluated and adjusted based on a preset elasticity threshold K. If the current node processing capacity utilization rate exceeds the K value, the system will automatically increase the number of nodes participating in the consensus and select nodes with higher contribution values ​​and breeding contribution indexes from other backup nodes to join, so as to ensure the efficiency of blockchain consensus.

[0039] Step S104: Associating the feed ratio library with the knowledge graph, iteratively training the model and generating multi-objective feeding decision parameters.

[0040] The feed ratio library in this embodiment stores information on feed formulas required for crabs at different growth stages and under different environmental conditions, including the ratios and contents of various nutrients. The knowledge graph models and represents the relationships between crabs, feed, environmental factors, and other factors involved in the crab farming process. Linking the feed ratio library with the knowledge graph allows the model to select appropriate feed formulas from the library based on the crab's growth status and environmental conditions, and to make inferences and decisions based on the knowledge in the knowledge graph.

[0041] The iterative training model in this embodiment continuously adjusts and optimizes the model using new data to improve its accuracy and adaptability. Through a multi-objective optimization algorithm, optimal feeding decision parameters are generated while meeting the crab's growth needs while considering multiple objectives such as feed cost and environmental impact.

[0042] Specifically, possible feed formulas can be screened from the feed ratio library based on the crab's growth stage and environmental conditions. Then, reasoning is performed using the entity relationships and rules in the knowledge graph to determine the feed formula and feeding strategy that best suits the current situation. Next, an iterative training algorithm is used to adjust the model and continuously optimize its parameters. Finally, a multi-objective optimization algorithm is used to generate multi-objective feeding decision parameters, including feeding amount, feeding frequency, and feed formula. In this way, the linked feed ratio library and knowledge graph provide the model with richer knowledge and experience, improving its accuracy and adaptability. Iterative training enables the model to continuously learn and evolve, adapting to various changes in the crab farming process.

[0043] Step S105: Control the feeding equipment to perform actions according to the decision parameters, monitor the model output in real time and trigger an early warning signal.

[0044] The decision parameters in this embodiment include information such as feeding amount, feeding frequency, and feed formula. These parameters are transmitted to the feeding equipment control system to control the feeding equipment's execution. Real-time monitoring of model output involves continuously collecting and analyzing the model's predictions and actual performance to determine whether the model is functioning properly. Triggering an early warning signal means issuing a timely alert when the model output shows an anomaly or predicts a potential problem, notifying relevant personnel to address the issue.

[0045] Specifically, after receiving the decision parameters, the feeding equipment control system controls the operation of the feeding equipment according to a pre-set program to achieve precise feeding. Simultaneously, the system monitors the model's output in real time, comparing actual feeding amounts, crab growth, and other data with the model's predictions. If deviations exceed a certain threshold or other abnormalities are detected, an early warning signal is triggered, notifying aquaculture managers via text message or app push notifications.

[0046] Step S106: Update the knowledge graph and feed ratio library based on the feedback data to form a closed-loop adaptive control.

[0047] The feedback data in this embodiment includes information such as the performance of feeding equipment, crab growth status, and changes in environmental parameters. This data is collected and analyzed and used to update the knowledge graph and feed ratio library. Updating the knowledge graph involves adjusting and improving the entity relationships and rules within the knowledge graph based on new data and experience. Updating the feed ratio library involves adjusting and optimizing the feed formula based on actual feeding results and crab growth.

[0048] As can be seen, the system regularly collects feedback data and cleans, analyzes, and mines it. Based on the analysis results, the entity relationships and rules in the knowledge graph are updated to more accurately reflect actual conditions. Furthermore, the feed ratio database is adjusted and optimized based on the feedback data to improve the nutritional value and applicability of the feed.

[0049] In an embodiment of the present invention, based on step S104, a possible embodiment will be given below to illustrate its specific implementation scheme in a non-limiting manner.

[0050] Step S104 specifically includes: Load the optimal feed formula for different growth stages of river crabs from the historical database.

[0051] Retrieve the nutritional requirement model and extract rules related to the current environmental growth status from the knowledge graph.

[0052] The rules include: environmental constraints, health associations, and multi-objective priorities.

[0053] Convert knowledge graph rules into numerical features. Normalize sensor data and convert knowledge graph rules into If-Then statements in the fuzzy rule base.

[0054] It's important to note that the rules in the knowledge graph are converted into numerical features for computer processing. Simultaneously, real-time data collected by sensors is normalized to ensure data consistency and comparability. The knowledge graph rules are then further converted into if-then statements within a fuzzy rule base, such as "If the water temperature is low and the crabs are in the juvenile stage, then reduce the feed intake and increase the protein content in the feed." In this way, complex aquaculture knowledge is transformed into computer-executable logical rules.

[0055] According to the priority rules of the knowledge graph, the weight of the fuzzy rule base is dynamically adjusted, and the membership function of the fuzzy set is defined using trigonometric functions.

[0056] This embodiment calculates priority weights in real time through multi-dimensional rule triggering probabilities such as environmental constraints and health associations, and uses trigonometric membership functions to dynamically adjust rule weight coefficients to solve the problem that fixed weight models are insufficiently adaptable to complex environments.

[0057] Mapping fuzzy rules to the rule layer of fuzzy neural network to form a dynamic update mechanism of rule base; The objective function was defined based on maximizing the average daily weight gain of river crabs, minimizing the amount of feed, and maintaining water quality parameters within safe thresholds.

[0058] The normalized data in step S101 and the numerical features of rule triggering probability and priority weight are used as input.

[0059] Calculate the decision parameters according to the following formula and adjust the dynamic weight of the priority of the knowledge graph.

[0060] Decision parameters = w1⋅growth optimization value + w2⋅resource conservation value + w3⋅environmental control value; w1 is the weight of growth optimization value, w2 is the weight of resource conservation value, and w3 is the weight of environmental control value; The feeding amount decision is based on the fuzzy neural network output and converted into the specific feeding amount value through the center of gravity method.

[0061] In some embodiments, a composite objective function is constructed, including maximizing average daily weight gain (w1), minimizing feed savings (w2), and maintaining water quality (w3). Dynamic weight assignment enables multi-objective collaborative optimization, improving decision-making comprehensiveness compared to traditional single-objective models. This embodiment maps the fuzzy rule base to the fuzzy neural network rule layer, forming a three-layer mapping mechanism of knowledge graph, fuzzy rules, and neural network, achieving closed-loop iterative optimization driven by both knowledge and data.

[0062] This example converts the trigger probability and priority weights of knowledge graph rules into computable numerical features, which serve as neural network input parameters. This addresses the difficulty of quantitative assessment in traditional expert systems. The objective function explicitly constrains water quality parameter safety thresholds, such as dissolved oxygen ≥ 5 mg / L, to achieve a safe aquaculture environment through mathematical programming.

[0063] Step S104 implements the decision of feeding river crabs through the following process: The optimal feed ratio for each growth stage of river crabs, such as the juvenile crab stage and the fattening stage, is extracted from the database. For example, the protein ratio in the juvenile crab stage is 45%.

[0064] Recall rules related to the current environment. For example, if the environmental constraint is that when dissolved oxygen is less than 5 mg / L, the feed amount should be reduced by 20%, or if the health constraint is that calcium demand should be increased by 15% during the molting period. The multi-objective priority is growth stage priority > resource conservation > environmental control. For example, in the juvenile crab stage, w1 = 0.6.

[0065] In this example, rules are converted into a fuzzy rule base by converting the rules in the knowledge graph into fuzzy If-Then statements. For example, if water temperature ∈ high temperature (membership > 0.7) AND dissolved oxygen ∈ low (membership > 0.6), then reduce the feeding amount by 15%. The membership of the fuzzy set is defined using trigonometric functions. For example, the high temperature membership function is a triangular distribution with a peak at 30°C. Real-time data, such as a water temperature of 28°C, is normalized to 0.8 and aligned with the numerical features of the knowledge graph rules, such as the high temperature threshold = 30°C, to a range of 0-1.

[0066] This embodiment dynamically calculates priority weights based on the knowledge graph's priority rules, adjusting w1, w2, and w3 in real time. For example, during the fattening period, w1 = 0.6 prioritizes growth, w2 = 0.3 prioritizes resource conservation, and w3 = 0.1 prioritizes environmental control. When dissolved oxygen is < 5 mg / L, w3 is forced to 0.5, prioritizing the environment.

[0067] The formula decision parameter = w1 × growth optimization value + w2 × resource conservation value + w3 × environmental control value quantifies multiple objectives into a single decision value. For example, the growth optimization value is calculated based on historical data. The resource conservation value is based on the current feed amount and feeding rate model. The environmental control value is predicted by the water quality model.

[0068] This embodiment embeds the fuzzy rule base into the rule layer of the fuzzy neural network to form a correspondence between rules and neurons. For example, the rules of high temperature and low dissolved oxygen, and reduced feeding correspond to the connection weights of specific neurons in the network. Input the normalized sensor data and the rule trigger probability. The centroid method defuzzification is to convert the fuzzy output into a specific value. For example: the output range is [7-9kg], and 8.3kg is obtained by gravity calculation. The way to update the knowledge graph is to incrementally learn new rules based on the execution results and manual feedback. For example, when the water temperature is 28℃ + the residual bait rate is >25%, the next feeding amount will be reduced by 10%. Retrain the fuzzy neural network with historical data and new rules to form a closed loop of data, model, and knowledge. For example, if a rule causes the water quality to deteriorate, the weight of the rule is reduced.

[0069] Step S104 achieves the precision, intelligence and adaptability of river crab feeding through the structured rules of knowledge graph, uncertainty modeling of fuzzy neural network and multi-objective optimization of dynamic weights.

[0070] Furthermore, as a refinement and extension of the specific implementation of the above embodiment, in order to fully illustrate the specific implementation process in this embodiment, step S103 further includes: Evaluate the status of aquaculture equipment and the impact of historical events based on real-time collected data, and dynamically adjust the data collection frequency and transmission priority by building a comprehensive scoring system for data importance; Optimize transmission resource allocation based on data importance score and urgency.

[0071] Based on the collected data, the equipment operating parameters and the correlation of historical events are determined, and a data priority evaluation formula is constructed. The data priority evaluation formula is Pscore=∑(wenv·Eparam)+∑(whist·Hfactor). Pscore represents the data priority score, wenn is the weight assigned based on the impact of the environment and equipment on river crab growth, Eparam is the aquaculture environment and equipment operating parameters, Hfactor represents the correlation of historical events, and whist is the weight assigned based on the impact scope and frequency of historical events. According to the priority score of river crab farming data and the timeliness requirements of data, the data transmission priority is set. According to the priority score Pscore and timeliness requirements, the priority Tle of each data transmission task is determined; Using the optimization algorithm, combined with the current data transmission status Ts and the priority Tle of each data transmission task, the transmission resource allocation Rd is dynamically adjusted, Rd = opt.min(∑Ttr(Tta,Ts)); Among them, Rd is the dynamically adjusted transmission resource allocation, Ts is the current transmission status, Tta represents the i-th data transmission task, and Ttr represents the task data transmission time. This embodiment clarifies that the aquaculture environment and equipment operating parameters Eparam include but are not limited to key indicators such as water temperature, dissolved oxygen concentration, and the working status of feed feeding equipment. Each indicator is assigned a different weight wenn according to its impact on the growth of river crabs and the stability of aquaculture. Based on information such as disease records and equipment failure handling records in the aquaculture history, the degree of correlation between current data and historical events, that is, the historical event correlation Hfactor, is evaluated. Each historical event is assigned a different weight whist according to the degree of loss it causes to aquaculture and the frequency of occurrence.

[0072] This embodiment can analyze the completeness and accuracy of collected data, assess the effectiveness of the current data collection strategy, and promptly adjust data priority scoring criteria and collection frequency. Based on the actual completion time and number of transmission failures during the data transmission process, the data transmission status assessment model and resource allocation strategy can be optimized to achieve more efficient data transmission scheduling.

[0073] In some specific embodiments, after step S103, the following steps are further performed: constructing a dual-path deep feature extraction framework, wherein the first path uses an optimized feature extraction network to process the image data; extracting local features through a convolutional layer, performing dimensionality reduction through a pooling layer, and integrating features through a fully connected layer. The calculation process is expressed as follows: Img_Feature=FCLayer(PoolLayer(ConvLayer(ImageData))) ImageData is the input aquaculture environment or crab morphology image data, and the image feature vector is obtained through multi-level convolution, pooling and full connection operations; The second path uses a gated recurrent unit network to perform feature mining on time series data. Time series information (TimeSeriesData), such as water quality monitoring data and feeding equipment operation logs, is input into the gated recurrent unit. The gated mechanism selectively memorizes and updates information, and outputs a time series feature vector: Seq_Feature=GRU(TimeSeriesData); Based on the data type importance evaluation results, the two feature vectors are weighted fused. The fusion formula is: Merged_Feature=α×Img_Feature+(1-α)×Seq_Feature Among them, α is an adaptive weight parameter, which is dynamically adjusted by analyzing the criticality of the data in the current breeding stage to achieve the optimal combination of features; A classification decision system is constructed based on the fused feature vector, and the normalized exponential function is used to calculate the probability distribution of each breeding status category: Probability(Class|Feature)=Softmax(WeightMatrix×Merged_Feature+BiasVector) Among them, WeightMatrix is ​​the classifier weight matrix, BiasVector is the bias vector, and through matrix operations and Softmax function, the probability value of each category under the current feature is output.

[0074] The classification system was supervised and trained using the cross-entropy loss criterion. Based on the labeled farming status dataset, the model parameters were optimized by minimizing the difference between the predicted probability and the true label.

[0075] As can be seen, a deep learning analysis architecture was created to acquire processed sensor data from the blockchain crab farming cloud platform. Data was preprocessed, and then a basic model was constructed by combining convolutional neural network and recurrent neural network structures. Feature extraction was performed on image data using a convolutional neural network, and time series data was processed using the long-short-term memory module of the recurrent neural network. The two types of features were then weighted and fused. Based on the fused feature vectors, a multi-classifier system was constructed, and a softmax function was used to achieve multi-category output prediction. Using a well-labeled farming dataset, the multi-classifiers were trained using supervised learning with a cross-entropy loss function as the optimization objective, achieving intelligent recognition and key information extraction from crab farming data.

[0076] This embodiment, based on a parallel image-time series data processing framework, enables simultaneous analysis of environmental visual information and equipment operating status information. An optimized convolutional network employs depthwise separable convolution and an attention mechanism to improve the efficiency of feature extraction in aquaculture scenarios. A gated recurrent unit introduces a time decay factor to enhance the ability to capture short-term dependencies such as water quality fluctuations and equipment anomalies. An α-adaptive adjustment algorithm is proposed to dynamically determine the fusion weights of image and time series features based on Bayesian optimization, addressing the inadequacy of fixed-weight models in complex environments.

[0077] In some specific embodiments, in step S106, based on the key breeding information extracted by the deep learning model, a river crab breeding status analysis model is built, the real-time sensor data and historical breeding data are input into the river crab breeding status analysis model, and the data analysis algorithm is used to mine the potential regularities and abnormal patterns in the data: The definition of the crab feeding status index is based on parameters that are more closely related to feed feeding. Let the crab feeding status index be CSI (CrabFeedingStatusIndex), and the calculation formula is: CSI=a1·m(Fc)+a2·n(Wt)+a3·p(Ds) Among them, a1, a2, and a3 are weight coefficients, reflecting the degree of influence of each parameter on the feeding status of river crabs; m(Fc), n(Wt), and p(Ds) are functions of the corresponding parameters, which convert the parameters into impact scores on the feeding status of river crabs; Fc represents the nutritional components of the feed (such as protein content, fat content, etc.), Wt represents the weight of the river crab (reflecting the growth status), and Ds represents the turbidity of the water quality (affecting the feeding environment of the river crab).

[0078] Taking into account the impact of environmental factors related to feed feeding on the crab rearing status, a new environmental impact coefficient EnviCoef is introduced. The adjusted crab rearing status index CSde is: CSde=CSI·EnviCoef CSde represents the adjusted feeding status index, and EnviCoef represents the environmental impact coefficient. This coefficient can be adjusted based on factors such as the impact of water temperature changes on feed digestion and the influence of light intensity on crab feeding behavior. A new warning threshold, CSId, for the crab feeding status index is set based on historical crab feeding data and analysis of problem patterns encountered during feeding. When the adjusted crab feeding status index, CSde, falls below the warning threshold, an early warning signal is issued and a notification is sent to the keeper, alerting them to a possible feeding issue that requires attention.

[0079] The method of this embodiment also constructs a multi-level knowledge graph, which includes a basic data level and a knowledge graph level; Divide the hierarchical structure according to the crab breeding data categories; construct a knowledge graph that includes equipment operation logic, growth cycle rules and abnormal causal relationships, and use physical equations and expert experience rules to map data layer information into executable knowledge nodes; Generate complex logical expressions based on the node attributes and relationships of multi-level knowledge graphs and real-time data streams; define retrieval strategies and execute them through pre-compiled instruction sets; Dependency parsing is used to extract keywords and map them to metadata tags at the basic data level. A word embedding model is used to convert natural language queries into high-dimensional vectors, which are then matched with the feature vectors of knowledge graph nodes for similarity. Based on the reasoning mechanism of the rule engine, dynamic logical judgment and operation chain are executed; Based on the parsed data vector, locate the associated nodes in the knowledge graph hierarchy; Call the action instructions in the predefined rule library to generate multimodal analysis results.

[0080] For the above method, let's take a crab farm's summer feeding monitoring (July 15th, water temperature 29°C, dissolved oxygen 5.5 mg / L) as an example. Real-time data includes: water temperature 29°C, dissolved oxygen 5.5 mg / L, average crab weight 80g, yesterday's feed amount 7kg, and residual feed rate 18%. Historical data shows a downward trend in dissolved oxygen over the past week (from 6.2 to 5.5 mg / L), and crab feeding time has increased from 2 hours to 2.5 hours. The equipment logical nodes defined in the knowledge graph hierarchy are: maximum capacity of feeder 8kg / time, association rule between aerator power and dissolved oxygen. Growth cycle rule: According to the knowledge base, July is the "mid-rearing period," and "plant-based feed should be increased to 55%." According to the knowledge base, the feed amount during high temperature periods is 8%-10% of crab body weight, i.e., 7.2-8kg (80g x 1000 crabs x 8%-10%). The abnormal relationships are: decreased dissolved oxygen, reduced feeding amount, lower residual feed rate, and improved water quality.

[0081] Based on natural language queries and semantic matching, the user inputs: "The current water temperature is 29°C, the dissolved oxygen is 5.5 mg / L, and the crabs are eating slowly. How should I adjust the feeding?" The processing flow is as follows: dependency parsing extracts the keywords "water temperature 29°C," "dissolved oxygen 5.5 mg / L," and "slow eating." The word embedding matches the word "slow eating" and maps it to the knowledge graph node "high leftover bait rate." The association rule is: When the leftover bait rate is >15%, reduce the feeding amount by 10%. "5.5 mg / L" is associated with the "5 mg / L dissolved oxygen safety threshold," triggering the "resource conservation rule."

[0082] Dynamic logical reasoning and operation chain generation. The compound logical expression is: IF (water temperature > 28°C AND dissolved oxygen < 6 mg / L) OR residual feed rate > 15%), THEN feeding amount = current feeding amount × (0.9 + 0.1 × dissolved oxygen / safety threshold). The calculation result is: original feeding amount 7 kg, new feeding amount = 7 × (0.9 + 0.1 × 5.5 / 5) = 7 × 0.955 ≈ 6.69 kg. The multimodal analysis result is: "Recommend today's feeding amount of 6.7 kg (a 4% decrease from yesterday), and increase plant-based feed such as wheat and corn to 55%." "Monitor dissolved oxygen; if it remains below 5.5 mg / L, activate the aerator." Charts display: feeding amount adjustment trend chart; negative correlation curve between dissolved oxygen and feeding amount. A water quality monitoring alarm triggers a text message notification when dissolved oxygen falls below 5 mg / L.

[0083] Knowledge graph iteration and exception handling can be based on the system recording "feed reduction strategy effective" when the residual feed rate drops to 12% after feeding. If the dissolved oxygen further drops to 4.8mg / L, the "emergency mode rule" in the knowledge graph is triggered: the feed amount is reduced to 5.5kg, the aerator is automatically activated, and a "water quality abnormality" report is sent to the administrator.

[0084] This method achieves precision, automation and explainability of river crab feeding monitoring through structured modeling of multi-level knowledge graphs, natural language-driven intelligent interaction and dynamic logical reasoning.

[0085] The following is an embodiment of the crab feed mixing monitoring system provided by the embodiments of the present disclosure. This system and the crab feed mixing monitoring method of the above-mentioned embodiments belong to the same inventive concept. For details not fully described in the embodiments of the crab feed mixing monitoring system, please refer to the embodiments of the above-mentioned crab feed mixing monitoring method.

[0086] like Figure 2 and Figure 3 As shown, the system includes: a data acquisition module, which is used to collect and normalize air, water quality, river crab growth and geographic information data through sensors to build a chain data acquisition system.

[0087] The data fusion module is used to use edge computing and blockchain technology to achieve distributed data storage and transmission, and complete multi-source data governance and feature fusion.

[0088] The feeding amount prediction module is used to build a fuzzy neural network hybrid model, integrating fuzzy processing, feature mining and defuzzification output to generate feeding amount prediction values.

[0089] The feeding decision module is used to associate the feed ratio library with the knowledge graph, iteratively train the model and generate multi-objective feeding decision parameters.

[0090] The monitoring module is used to control the feeding equipment to perform actions according to the decision parameters, monitor the model output in real time and trigger early warning signals.

[0091] The self-regulation module updates the knowledge graph and feed ratio library based on feedback data, forming a closed-loop adaptive regulation.

[0092] like Figure 4 As shown, the present application also provides an electronic device, including a display module 103, a memory 102, a processor 101 and a computer program stored in the memory and executable on the processor 101. When the processor 101 executes the program, the steps of the crab feed feeding and mixing monitoring method are implemented.

[0093] In the embodiments of the present invention, electronic devices include, but are not limited to, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the embodiments of the present application described and / or claimed herein.

[0094] In the embodiment of the present application, the processor 101 can be implemented by using at least one of a special purpose integrated circuit, a programmable logic device, a field programmable gate array, a processor, a controller, a microcontroller, a microprocessor, and an electronic unit designed to perform the functions described herein. In some cases, such an embodiment can be implemented in a controller. For software implementation, an embodiment such as a process or function can be implemented with a separate software module that allows the execution of at least one function or operation. The software code can be implemented by a software application (or program) written in any appropriate programming language, and the software code can be stored in a memory and executed by a controller.

[0095] The display module 103 is used to display information input by the user or information provided to the user. The display module 103 may include a display panel, which may be configured in the form of a liquid crystal display, an organic light emitting diode, etc.

[0096] The memory 102 can be used to store software programs and various data. The memory 102 can include high-speed random access memory and non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0097] The present application also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the crab feed feeding and mixing monitoring method.

[0098] The storage medium can be any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or component, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0099] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for monitoring the feeding and mixing of river crab feed, characterized in that: include: S101: Collect and normalize air, water quality, crab growth, and geographic information data through sensors to build a chain data collection system; S102: Use edge computing and blockchain technology to achieve distributed data storage and transmission, and complete multi-source data governance and feature fusion; S103: Constructing a fuzzy neural network hybrid model, integrating fuzzification processing, feature mining and defuzzification output, and generating a feeding amount prediction value; S104: Associating the feed ratio library with the knowledge graph, iteratively training the model and generating multi-objective feeding decision parameters; S105: Control the feeding equipment to execute actions according to the decision parameters, monitor the model output in real time and trigger an early warning signal; S106: Update the knowledge graph and feed ratio library based on the feedback data to form a closed-loop adaptive control.

2. The method for monitoring the feeding and mixing of river crab feed according to claim 1, wherein: Step S103 specifically includes: storing the full cycle data of river crab farming in the blockchain farming cloud platform, which integrates the adaptive consensus algorithm and the hierarchical data protection system; For each river crab farming monitoring node j, the contribution value is calculated using the comprehensive indicator formula Ej=α1×Bj+α2×Sj+α3×Pj; Among them, Bj represents the aquaculture data collection accuracy index of the river crab aquaculture monitoring node, Sj is the parameter of the aquaculture water area that can be covered, Pj represents the parameter of the data processing capacity per unit time, and α1, α2, and α3 are weighting coefficients dynamically configured according to the aquaculture stage; The aquaculture contribution index Cj is defined and obtained by the formula Cj=γ1×Vj+γ2×Nj+γ3×Fj, where Vj is the accuracy of crab growth data monitoring in the previous statistical period of the crab aquaculture monitoring node, Nj is the number of times of abnormal aquaculture data warning, Fj is the effective number of times the node assists in optimizing the feeding strategy, and γ1, γ2, and γ3 are adjustment coefficients used to balance different aquaculture contribution dimensions.

3. The method for monitoring the feeding and mixing of river crab feed according to claim 1, wherein: After step S103, the following steps are also included: Based on the collected data, the equipment operating parameters and the correlation of historical events are determined, and a data priority evaluation formula is constructed. The data priority evaluation formula is Pscore=∑(wenv·Eparam)+∑(whist·Hfactor). Pscore represents the data priority score, wenn is the weight assigned based on the impact of the environment and equipment on river crab growth, Eparam is the aquaculture environment and equipment operating parameters, Hfactor represents the correlation of historical events, and whist is the weight assigned based on the impact scope and frequency of historical events. According to the priority score of river crab farming data and the timeliness requirements of the data, the data transmission priority is set; according to the priority score Pscore and timeliness requirements, the priority Tle of each data transmission task is determined; Using the optimization algorithm, combined with the current data transmission status Ts and the priority Tle of each data transmission task, the transmission resource allocation Rd is dynamically adjusted, Rd = opt.min(∑Ttr(Tta,Ts)); Among them, Rd is the transmission resource allocation after dynamic adjustment, Ts is the current transmission status, Tta represents the i-th data transmission task, and Ttr represents the task data transmission time.

4. The method for monitoring the feeding and mixing of river crab feed according to claim 3, wherein: Step S104 specifically includes: Load the optimal feed formula for different growth stages of river crabs from the historical database; Retrieve the nutritional demand model and extract rules related to the current environmental growth status from the knowledge graph; The rules include: environmental constraints, health linkages, and multi-objective priorities; Convert knowledge graph rules into numerical features; Normalize the sensor data and convert the knowledge graph rules into If-Then statements in the fuzzy rule base; According to the priority rules of the knowledge graph, the weight of the fuzzy rule base is dynamically adjusted, and the membership function of the fuzzy set is defined using trigonometric functions; Mapping fuzzy rules to the rule layer of fuzzy neural network to form a dynamic update mechanism of rule base; Define the objective function based on maximizing the average daily weight gain of river crabs, minimizing the amount of feed, and maintaining water quality parameters within safe thresholds; The normalized data in step S101 and the numerical features of rule trigger probability and priority weight are used as input; Calculate the decision parameters according to the following formula and adjust the dynamic weight of the priority of the knowledge graph; Decision parameters = w1⋅growth optimization value + w2⋅resource conservation value + w3⋅environmental control value; w1 is the weight of growth optimization value, w2 is the weight of resource conservation value, and w3 is the weight of environmental control value; The feeding amount decision is based on the fuzzy neural network output and converted into the specific feeding amount value through the center of gravity method.

5. The method for monitoring the feeding and mixing of river crab feed according to claim 1, wherein: After step S103, the following steps are also included: A dual-path deep feature extraction framework is constructed. The first path uses an optimized feature extraction network to process image data. Local features are extracted through the convolution layer, the pooling layer performs dimensionality reduction, and the fully connected layer integrates features. The calculation process is expressed as follows: Img_Feature=FCLayer(PoolLayer(ConvLayer(ImageData))) ImageData is the input aquaculture environment or crab morphology image data, and the image feature vector is obtained through multi-level convolution, pooling and full connection operations; The second path uses a gated recurrent unit network to perform feature mining on time series data. Time series information (TimeSeriesData), such as water quality monitoring data and feeding equipment operation logs, is input into the gated recurrent unit. The gated mechanism selectively memorizes and updates information, and outputs a time series feature vector: Seq_Feature=GRU(TimeSeriesData); Based on the data type importance evaluation results, the two feature vectors are weighted fused. The fusion formula is: Merged_Feature=α×Img_Feature+(1-α)×Seq_Feature Among them, α is an adaptive weight parameter, which is dynamically adjusted by analyzing the criticality of the data in the current breeding stage to achieve the optimal combination of features; A classification decision system is constructed based on the fused feature vector, and the normalized exponential function is used to calculate the probability distribution of each breeding status category: Probability(Class|Feature)=Softmax(WeightMatrix×Merged_Feature+BiasVector) Among them, WeightMatrix is ​​the classifier weight matrix, BiasVector is the bias vector, and through matrix operations and Softmax function, the probability value of each category under the current feature is output.

6. The method for monitoring the feeding and mixing of river crab feed according to claim 1, wherein: In step S106, based on the key breeding information extracted by the deep learning model, a river crab breeding status analysis model is built. The real-time sensor data and historical breeding data are input into the river crab breeding status analysis model, and the data analysis algorithm is used to mine the potential regularities and abnormal patterns in the data: Among them, the crab breeding status index is defined as CSI and the calculation formula is: CSI=a1·m(Fc)+a2·n(Wt)+a3·p(Ds); Among them, a1, a2, and a3 are weight coefficients, reflecting the degree of influence of each parameter on the crab rearing status; m(Fc), n(Wt), and p(Ds) are functions of the corresponding parameters, which convert the parameters into the impact score on the crab rearing status; Fc represents the nutritional content of the feed, Wt represents the weight of the crab, and Ds represents the turbidity of the water; Introducing the new environmental impact coefficient EnviCoef, the adjusted crab breeding status index CSde is: CSde=CSI·EnviCoef Wherein, CSde represents the adjusted feeding status index, and EnviCoef represents the environmental impact coefficient; Based on historical river crab breeding data and analysis of problem patterns that occur during the feeding process, a new warning threshold CSId for the river crab breeding status index is set; when the adjusted river crab breeding status index CSde is less than the warning threshold CSId, an early warning signal is issued and a notification is sent to the breeding staff.

7. The method for monitoring the feeding and mixing of river crab feed according to claim 1, wherein: The method also constructs a multi-level knowledge graph, which includes a basic data level and a knowledge graph level; Divide the hierarchical structure according to the crab breeding data categories; construct a knowledge graph that includes equipment operation logic, growth cycle rules and abnormal causal relationships, and use physical equations and expert experience rules to map data layer information into executable knowledge nodes; Generate compound logical expressions based on the node attributes and relationships of multi-level knowledge graphs and real-time data streams; Define retrieval strategies and execute them through precompiled instruction sets; Dependency parsing is used to extract keywords and map them to metadata tags at the basic data level. A word embedding model is used to convert natural language queries into high-dimensional vectors, which are then matched with the feature vectors of knowledge graph nodes for similarity. Based on the reasoning mechanism of the rule engine, dynamic logical judgment and operation chain are executed; Based on the parsed data vector, locate the associated nodes in the knowledge graph hierarchy; Call the action instructions in the predefined rule library to generate multimodal analysis results.

8. A crab feed feeding and mixing monitoring system, characterized in that: The system is used to implement the crab feed mixing monitoring method according to any one of claims 1 to 7; The system includes: The data acquisition module is used to collect and normalize air, water quality, river crab growth and geographic information data through sensors to build a chain data collection system; The data fusion module is used to implement distributed data storage and transmission using edge computing and blockchain technology, and complete multi-source data governance and feature fusion; Feeding amount prediction module is used to build a fuzzy neural network hybrid model, integrating fuzzy processing, feature mining and defuzzification output to generate feeding amount prediction values; The feeding decision module is used to associate the feed ratio library with the knowledge graph, iteratively train the model and generate multi-objective feeding decision parameters; The monitoring module is used to control the feeding equipment to perform actions according to the decision parameters, monitor the model output in real time and trigger early warning signals; The self-regulation module updates the knowledge graph and feed ratio library based on feedback data, forming a closed-loop adaptive regulation.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the crab feed feeding and mixing monitoring method as described in any one of claims 1 to 7 are implemented.

10. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the crab feed feeding and mixing monitoring method as described in any one of claims 1 to 7 are implemented.

Images