An intelligent large-scale ecological fishery management system based on multi-source data fusion

Abstract

An intelligent large-scale ecological fishery management system based on multi-source data fusion relates to the field of smart fishery technology, including: dividing a large-scale aquaculture area into several sub-areas, setting environmental monitoring points and biological monitoring points in each sub-area, collecting multi-source data, and constructing an integrated visual map of the multi-source data; outputting real-time threshold intervals corresponding to each water quality monitoring indicator, meteorological monitoring indicator, and biological characteristic indicator in the current collection period; performing real-time monitoring on each sub-area in the integrated visual map of the multi-source data, generating sudden abnormality alarms or performing hidden trend analysis based on the real-time monitoring results; performing intelligent decision support operations on the sub-area or marking the sub-area as an ecological risk area based on the hidden trend analysis results; and obtaining the optimal fishing opportunity using a dual-drive fishing algorithm based on multiple sets of time data in the sub-area, thereby realizing comprehensive monitoring, precise perception, and risk prevention and control of ecological fishery management.

Description

Technical Field

[0001] The present invention relates to the field of smart fishery technology, and in particular to an intelligent large-scale water surface ecological fishery management system based on multi-source data fusion. Background Art

[0002] Chinese patent publication number CN110399774B discloses a fishery monitoring and management system, comprising: a water surface detection device for extracting a water surface area in an on-site output image based on water surface imaging features; a fish body identification device for extracting one or more fish body areas in the on-site output image based on fish body imaging features; and a range analysis device for issuing an emergency rescue instruction when a fish body area located outside the water surface area exists in the on-site output image.

[0003] Existing ecological fishery management systems suffer from the following technical issues: They lack intelligent zoning and point optimization methods based on water physical properties, biological characteristics, and historical pollution frequencies. This can easily lead to monitoring blind spots (e.g., missing data in complex terrain or offshore areas) or redundant data (e.g., uniform point distribution that ignores differences in aquaculture density), making it difficult to accurately capture the spatiotemporal heterogeneity of key ecological parameters. Multi-source data fusion is inefficient: large errors in data time alignment and a lack of three-dimensional visualization layer overlay technology make it difficult to intuitively correlate multi-dimensional data such as water quality, meteorology, and biology. Dynamic prediction and risk identification capabilities are weak: Anomaly warnings rely on fixed thresholds or single models, failing to leverage deep learning to build multi-source data prediction models. The dynamic adjustment of real-time threshold intervals lacks accuracy, and the ability to identify hidden trends (e.g., chronic pollution accumulation, early signs of abnormal fish growth) is inadequate, which can lead to delayed or misjudgment responses to emergencies. Inadequate communication network coverage and reliability: A hybrid air-ground communication network has not been constructed to address the environmental differences between shallow, deep, and offshore areas. Signal blind spots may occur (e.g., insufficient 5G coverage in deep waters, data transmission interruptions in offshore areas), compromising the integrity and transmission efficiency of real-time monitoring data. The lack of in-depth analysis of historical data and real-time trends makes it difficult to determine the best time to fish, which makes fishery management rely on experience and judgment, making it difficult to achieve the goals of intelligent, scientific and ecologically sustainable development. Summary of the Invention

[0004] In order to solve the above technical problems, the purpose of the present invention is to provide an intelligent large-scale water surface ecological fishery management system based on multi-source data fusion, including a cloud and a dynamic response acquisition module. The cloud includes an integrated visual module, a multi-source data prediction module, a real-time monitoring module, a hidden trend analysis module and an intelligent decision support module;

[0005] Building an air-ground integrated network, wherein the cloud and the dynamic response collection module are connected via the air-ground integrated network communication;

[0006] The dynamic response acquisition module is used to divide the large water surface aquaculture area into several sub-areas, set up environmental monitoring points and biological monitoring points in each sub-area, and collect multi-source data, including water quality monitoring indicators, meteorological monitoring indicators and biological characteristic indicators;

[0007] The integrated visualization module is used to time-calibrate the multi-source data of the dynamic response acquisition module, fuse the multi-source data with the same receiving timestamp and sending timestamp into a set of time data, and construct an integrated visualization graph of the multi-source data based on the multiple sets of time data;

[0008] The multi-source data prediction module is used to build a multi-source data prediction model and output the real-time threshold intervals corresponding to various water quality monitoring indicators, meteorological monitoring indicators, and biological characteristic indicators in the current collection period;

[0009] The real-time monitoring module is used to monitor each sub-area in the integrated visual graph of multi-source data in real time, and generate sudden abnormality alarms or perform hidden trend analysis based on the real-time monitoring results;

[0010] The hidden trend analysis module obtains abnormal trend feature sets of several historical abnormal collection cycles, performs hidden trend analysis on the time data of several collection cycles of the sub-region based on the abnormal trend feature sets of several historical abnormal collection cycles, and performs intelligent decision support operations on the sub-region or marks the sub-region as an ecological risk area based on the hidden trend analysis results;

[0011] The intelligent decision support module constructs a dual-drive fishing algorithm of Gompertz growth model + real-time feeding recognition, and uses the dual-drive fishing algorithm to obtain the best fishing time based on multiple sets of time data in the sub-area.

[0012] Furthermore, the large water surface aquaculture area is divided into several sub-areas, and the process of setting up environmental monitoring points in each sub-area includes:

[0013] Obtain the water body physical characteristics (water depth and topography), environmental characteristics (water flow and wind direction, meteorological conditions), and biological characteristics (fish species and ecological niche, stocking density) of large water surface aquaculture areas, and divide the large water surface aquaculture areas into several sub-areas based on the water body physical characteristics;

[0014] The historical pollution frequencies of several sub-regions are obtained based on the historical data of several sub-regions. The water physical characteristics, environmental characteristics and historical pollution frequencies of each sub-region are used as evaluation indicators. The indicator weights of the evaluation indicators are set. The membership matrix of each sub-region to the preset importance level is obtained through fuzzy comprehensive evaluation.

[0015] The importance level of each sub-area is obtained based on the membership matrix and the indicator weight. The number and distribution of environmental monitoring points are obtained based on the importance level of the sub-area and the location characteristics (the location coordinate set that constitutes the sub-area). The environmental monitoring points are used to collect water quality monitoring indicators (water temperature, dissolved oxygen, pH value, ammonia nitrogen, etc.) and meteorological monitoring indicators (air temperature, wind speed, wind direction, air pressure, precipitation, etc.), mark the collection time, and set the collection cycle.

[0016] Furthermore, the process of obtaining the number of environmental monitoring points and their distribution according to the importance level and location characteristics of the sub-areas includes:

[0017] The number of environmental monitoring points corresponding to different importance levels is preset, and the number n of environmental monitoring points in the sub-area is determined according to the importance level of the sub-area;

[0018] Initialize the locations of environmental monitoring points in the sub-area and set the coordinate set P of the initial environmental monitoring points: ;

[0019] Construct a polygonal monitoring area for each initial environmental monitoring point. , define its polygon monitoring area as all The set of points whose distance is less than the distance to other initial environmental monitoring points :

[0020] ;

[0021] in, Initial environmental monitoring point The maximum monitoring area, From point Q to the initial environmental monitoring point The Euclidean distance of : ;

[0022] Optimize the initial environmental monitoring point locations, with the goal of minimizing the area variance of each polygon monitoring area, and iteratively adjust the node positions to make the coverage more uniform. The objective function is:

[0023] ;

[0024] in, is the area of ​​the i-th polygon monitoring area, is the average area.

[0025] Furthermore, the process of setting up biological monitoring points in each sub-area includes:

[0026] Preset the point density threshold per unit area (e.g. 1-2 points per 100 hectares), based on the total coverage area of ​​the sub-region Point density threshold per unit area , obtain the basic number of biological monitoring points in the sub-area , , Indicates rounding up, obtaining the terrain characteristics of the sub-region (such as shallows, deep pools, gullies, and water confluences) based on the physical characteristics of the water body in the sub-region, presetting the correction coefficients corresponding to different terrain characteristics, and Basic number of biological monitoring points Make corrections and obtain the corrected number of biological monitoring points ;

[0027] According to the biological characteristics of the sub-area (fish species and ecological niche, breeding density), the fish species and ecological niche (fish habitat water layer (surface layer, middle layer, lower layer, bottom layer) and activity area (such as migration channel, feeding ground, spawning ground)) and breeding density parameters are obtained. According to the fish species and ecological niche and breeding density parameters, the revised number of biological monitoring points is weighted to obtain the final number of biological monitoring points. According to the final number of biological monitoring points, the point distribution of the biological monitoring points in the sub-area is obtained. The biological monitoring points are used to collect biological characteristic indicators, mark the collection time, and set the collection cycle.

[0028] Furthermore, the revised number of biological monitoring points is weighted according to the fish species, ecological niche and stocking density parameters. The process of obtaining the final number of biological monitoring points includes:

[0029] ;

[0030] in, is the final number of biological monitoring points, is the maximum stocking density, is the average stocking density, is the niche width of the hth fish species (obtained in advance through habitat utilization assessment), is the reference niche width (e.g., 1 for a single species);

[0031] Furthermore, the process of building an integrated air-ground network includes:

[0032] The water depth and offshore distance of each sub-area are obtained based on the physical characteristics of the water body and location features of each sub-area. Each sub-area is divided into shallow water area, deep water area and offshore area according to the offshore distance and water depth of each sub-area. Among them, shallow water area: the water depth is usually less than 5 meters and the distance from the shore is ≤10 kilometers, which is suitable for deploying LoRaWAN (coverage radius 3-5 kilometers, low power consumption, strong penetration); deep water area: the water depth is ≥5 meters and the distance from the shore is ≤10 kilometers, and there is 5G signal coverage; offshore area: deep water area with a distance from the shore greater than 10 kilometers or without 5G coverage, relying on satellite communication (such as BGAN satellite terminal, global coverage but high latency);

[0033] According to the coverage area of ​​shallow water area, LoRaWAN nodes are deployed in shallow water area using regular hexagonal grid coverage. The number of nodes required for LoRaWAN nodes is , ,in To round up, is the coverage area of ​​the shallow water area, is the maximum monitoring range of the LoRaWAN node. 5G nodes are deployed in a grid-like distribution in the deep water area according to the coverage area of ​​the deep water area. The number of 5G nodes required is , ,in To round up, is the coverage area of ​​the deep water area, is the coverage radius of the 5G base station. Based on the coverage area of ​​the offshore area, BGAN satellite nodes are deployed in the offshore area using a regular hexagonal grid coverage method. The number of BGAN satellite nodes required is , ,in To round up, is the coverage area of ​​the offshore area, is the maximum monitoring range of the BGAN satellite node;

[0034] The data transmission path in shallow water areas is: LoRaWAN node → shore gateway → 5G / fiber → cloud;

[0035] The data transmission path in the deepwater area (5G coverage) is: 5G node → 5G base station → core network → cloud;

[0036] The data transmission path in the offshore area is: BGAN satellite node → satellite → ground station → Internet → cloud;

[0037] Furthermore, the multi-source data of the dynamic response acquisition module are time-calibrated, and the multi-source data with the same receiving timestamp and sending timestamp are fused into a set of time data. The process of constructing an integrated visual graph of the multi-source data based on the multiple sets of time data includes:

[0038] Obtain multi-source data from the dynamic response acquisition module, and record the sending timestamp of the multi-source data based on the timing function of the Beidou system, and determine whether the cloud has received the multi-source data, wherein the cloud receives Beidou standard timing data according to the acquisition cycle and obtains the regional time on the cloud, compares the Beidou standard timing data with the regional time for consistency, and if inconsistent, updates the regional time on the cloud according to the Beidou standard timing data to obtain the standard regional time, and if so, records the receiving timestamp of the multi-source data based on the timing function of the Beidou system (records the receiving timestamp of the multi-source data according to the standard regional time on the cloud to ensure time consistency between the cloud and the dynamic response acquisition module),

[0039] A 3D model layer of a large aquaculture area is constructed based on Unity3D digital twin technology. Multi-source data is divided into multiple groups of time data based on their receiving and sending timestamps (multi-source data with the same receiving and sending timestamps are merged into one group). The multiple groups of time data are then divided into a water quality monitoring indicator time layer, a meteorological monitoring indicator time layer, and a biological characteristic indicator time layer.

[0040] Taking the three-dimensional model layer as the base layer, the water quality monitoring indicator time layer, the meteorological monitoring indicator time layer and the biological characteristic indicator time layer are superimposed on the base layer to obtain an integrated visual graph of multi-source data.

[0041] Furthermore, the process of constructing a multi-source data prediction model and outputting the real-time threshold intervals corresponding to each water quality monitoring indicator, meteorological monitoring indicator, and biological characteristic indicator in the current collection period includes:

[0042] Construct a multi-source data prediction model based on deep learning, obtain multi-source data from several historical collection cycles as training sets and test sets, input the training sets into the multi-source data prediction model for training until the loss function training is stable, save the model parameters, test the multi-source data prediction model with the test set until it meets the preset requirements, and output the multi-source data prediction model;

[0043] According to the multi-source data model, the predicted numerical time series sequence corresponding to each water quality monitoring indicator, meteorological monitoring indicator and biometric indicator in the current collection period is output. According to the predicted numerical time series sequence corresponding to each water quality monitoring indicator, meteorological monitoring indicator and biometric indicator in the current collection period, the real-time threshold range corresponding to each water quality monitoring indicator, meteorological monitoring indicator and biometric indicator in the current collection period is obtained.

[0044] Furthermore, the process of performing real-time monitoring on each sub-area in the integrated visual graph of multi-source data and generating sudden abnormality alerts or performing hidden trend analysis based on the real-time monitoring results includes:

[0045] Mark each water quality monitoring indicator, meteorological monitoring indicator, and biometric indicator in the water quality monitoring indicator time layer, meteorological monitoring indicator time layer, and biometric indicator time layer as a first indicator, obtain the real-time threshold interval corresponding to each first indicator in the current acquisition period, obtain the numerical time series sequence corresponding to the first indicator in the current acquisition period, compare the numerical time series sequence corresponding to the first indicator with the corresponding real-time threshold interval, and obtain the cumulative time that the numerical time series sequence corresponding to the first indicator is not within the corresponding real-time threshold interval;

[0046] A cumulative time threshold is preset. If the cumulative time corresponding to a first indicator is greater than the cumulative time threshold, the first indicator is marked as a key monitoring indicator, an emergency abnormality alarm is generated, the layer to which the key monitoring indicator belongs and the location of the multi-source data integrated visual map are highlighted in red, and the current collection period is marked as an abnormal collection period;

[0047] If the cumulative time corresponding to each first indicator is less than or equal to the cumulative time threshold, the current collection period is marked as a normal collection period, and a hidden trend analysis is performed.

[0048] Furthermore, the process of the hidden trend analysis module obtaining the abnormal trend feature set of several historical abnormality collection cycles includes:

[0049] Extracting the numerical time series sequences corresponding to the key monitoring indicators and other first indicators within the historical anomaly collection period, and obtaining the trend correlation coefficient between the key monitoring indicators and other first indicators and the stability coefficient of the other first indicators based on the numerical time series sequences;

[0050] Extract the numerical time series corresponding to the key monitoring indicators and other first indicators of the previous k historical normal collection cycles of the historical abnormal collection cycle; obtain the trend correlation coefficients between the key monitoring indicators and other first indicators of the previous k historical normal collection cycles of the historical abnormal collection cycle and the stability coefficients of the other first indicators based on the numerical time series corresponding to the key monitoring indicators and other first indicators of the previous k historical normal collection cycles of the historical abnormal collection cycle; obtain the average trend correlation coefficients between the key monitoring indicators and other first indicators of the previous k historical normal collection cycles of the historical abnormal collection cycle and the average stability coefficients of the other first indicators based on the trend correlation coefficients between the key monitoring indicators and other first indicators of the previous k historical normal collection cycles;

[0051] Obtain the trend correlation change coefficient between the key monitoring indicator and the other first indicators based on the trend correlation coefficient and the average trend correlation coefficient between the key monitoring indicator and the other first indicators; obtain the stability change coefficient of the other first indicators based on the stability coefficient and the average stability coefficient of the other first indicators;

[0052] An abnormal trend feature set of a historical abnormal collection period is constructed based on the trend correlation change coefficient between the key monitoring indicator and other first indicators and the stability change coefficient of other first indicators.

[0053] Furthermore, the process of obtaining the trend correlation coefficient between the key monitoring indicator and the other first indicators and the stability coefficient of the other first indicators includes:

[0054] ;

[0055] ;

[0056] in, Indicates the trend correlation coefficient between the key monitoring indicator a and other first indicators b, represents the stability coefficient of the other first indicator b, Indicates the value of the key monitoring indicator a at the tth moment, represents the value of the other first indicator b at the tth moment, Indicates the average value of the key monitoring indicator a during the collection period. Indicates the average value of other first indicators b during the collection period, Indicates the total number of moments included in the collection period.

[0057] The process of obtaining the trend correlation coefficient of the key monitoring indicator and other first indicators and the stability coefficient of the other first indicators includes:

[0058] ;

[0059] ;

[0060] in, Indicates the trend correlation coefficient between the key monitoring indicator a and the other first indicator b, It represents the average trend correlation coefficient between the key monitoring indicator a and other first indicators b, represents the stability variation coefficient of the other first indicator b, Represents the average stability coefficient of other first indicators b.

[0061] Furthermore, based on the abnormal trend feature sets of several historical abnormal collection cycles, a hidden trend analysis is performed on multiple sets of time data of the sub-region. The process of performing intelligent decision support operations on the sub-region or marking the sub-region as an ecological risk area according to the hidden trend analysis results includes:

[0062] Select the first indicator with the longest cumulative time in the current collection period of the sub-region, and mark the first indicator as the estimated key monitoring indicator;

[0063] Obtain the numerical time series sequence corresponding to the first indicator and the estimated key monitoring indicator in the k collection periods before the current collection period. Based on the numerical time series sequence corresponding to the first indicator and the estimated key monitoring indicator in the current collection period and the k collection periods before the current collection period, obtain the trend correlation change coefficient between the estimated key monitoring indicator and other first indicators in the current collection period and the stability change coefficient of other first indicators;

[0064] Compare the trend correlation change coefficients between the estimated key monitoring indicators and other first indicators in the current collection period and the stability change coefficients of other first indicators with the abnormal trend feature sets of several historical abnormal collection periods to obtain the similarity of each abnormal trend feature set (obtained by using cosine similarity calculation method);

[0065] A similarity threshold is preset. If the similarity of each abnormal trend feature set is less than or equal to the similarity threshold, an intelligent decision support operation is performed. If the similarity of an abnormal trend feature set is greater than the similarity threshold, the sub-region is marked as an ecological risk area, the location of the sub-region in the multi-source data integrated visual map is highlighted in orange, and emergency treatment measures for the sub-region are generated.

[0066] Furthermore, the process of marking the sub-region as an ecological anomaly region and generating emergency treatment measures for the sub-region includes: constructing a sudden ecological anomaly database, obtaining emergency treatment measures corresponding to a number of historical anomaly collection cycles, matching abnormal trend feature sets within the number of historical anomaly collection cycles with the emergency treatment measures corresponding to the number of historical anomaly collection cycles, and storing the matched sets in the sudden ecological anomaly database;

[0067] When the similarity of an abnormal trend feature set is greater than a similarity threshold, a search is performed in the sudden ecological anomaly database according to the abnormal trend feature set to obtain emergency treatment measures corresponding to the abnormal trend feature set.

[0068] Furthermore, the intelligent decision support module constructs a dual-drive fishing algorithm consisting of a Gompertz growth model and real-time feeding recognition. The process of using the dual-drive fishing algorithm to obtain the optimal fishing opportunity based on multiple sets of time data in the sub-area includes:

[0069] Based on the numerical time series corresponding to the first indicator of the current collection period, the dissolved oxygen correction factor, water temperature correction factor and real-time food intake of the current collection period are obtained. Based on the numerical time series corresponding to the first indicator of several historical collection periods, nonlinear fitting is performed to construct a Gompertz growth model. Based on the Gompertz growth model, the average weight and theoretical growth rate of the fish are obtained.

[0070] Among them, the Gompertz growth model formula is:

[0071] ;

[0072] in, Indicates time The average weight of fish, Indicates the maximum weight of the fish, exp() is the natural exponential function, Indicates the breeding time, and The nonlinear fitting is performed by obtaining the numerical time series corresponding to the first indicator of several historical collection periods. It should be noted that is a constant, reflecting the initial growth conditions and early growth rate, The larger it is, the faster the initial growth rate (the curve shifts to the left). The theoretical growth rate determines how quickly the growth curve reaches its maximum weight. The bigger you are, the faster you grow and approach your maximum weight. ;

[0073] The actual growth rate is obtained based on the theoretical growth rate, dissolved oxygen correction factor and water temperature correction factor. The feeding efficiency coefficient is obtained based on the actual growth rate, average fish weight and real-time food intake. The optimal fishing weight is preset, and the optimal fishing time is obtained based on the average fish weight, optimal fishing weight and feeding efficiency coefficient.

[0074] The formula for obtaining the actual growth rate based on the theoretical growth rate, dissolved oxygen correction factor, and water temperature correction factor is:

[0075] ;

[0076] ;

[0077] ;

[0078] in, represents the actual growth rate, is the dissolved oxygen correction factor, is the water temperature correction factor, Indicates the dissolved oxygen value, Indicates the minimum threshold of dissolved oxygen, is the optimum dissolved oxygen value, For water temperature, is the minimum water temperature threshold, For the optimum water temperature, is the maximum water temperature threshold;

[0079] The formula for obtaining the feeding efficiency coefficient based on the actual growth rate, average fish weight, and real-time food intake is:

[0080] ;

[0081] in, Indicates time Real-time food intake, represents the feeding efficiency coefficient;

[0082] The formula for obtaining the best fishing time based on the average fish weight, optimal fishing weight and feeding efficiency coefficient is:

[0083] ;

[0084] in, For the best fishing weight, is the index, is the preset parameter, It is the best time to fish.

[0085] Compared with the prior art, the present invention has the following beneficial effects:

[0086] 1. Comprehensive monitoring and precise perception: By dividing the aquaculture area and scientifically setting up environmental and biological monitoring points, the system comprehensively collects multi-source data such as water quality, meteorological, and biological data. Compared with a single monitoring method, it can grasp the ecological status of the aquaculture area in a more detailed and accurate manner, capture subtle changes in a timely manner, avoid missing key information, and provide richer and more accurate data support for fishery management.

[0087] 2. Efficient data fusion and visualization: The integrated visualization module performs time calibration and fusion of multi-source data to construct an intuitive integrated visualization of multi-source data. This solves the previous problem of scattered data and difficulty in intuitive presentation. Managers can quickly obtain key information, improve decision-making efficiency and accuracy, and realize visual and intelligent management of breeding areas.

[0088] 3. Accurate prediction and risk prevention and control: The multi-source data prediction module uses deep learning to build a model and output real-time threshold ranges. It can not only predict the changing trends of various indicators, but also, by combining with real-time monitoring, promptly detect sudden anomalies and hidden trends. Compared with traditional early warning methods, it can identify risks more early and accurately, reduce fishery losses, and ensure ecological security.

[0089] 4. Advanced communications and stable operation: The construction of an integrated air-ground network adopts appropriate communication technologies based on the characteristics of different regions to ensure stable data transmission. Compared with a single communication method, this improves the reliability and coverage of system communications, ensures stable operation of the system in complex environments, and provides solid communication support for fishery management.

[0090] 5. The dual-driven fishing algorithm combines the Gompertz growth model with real-time feeding recognition to accurately match fish growth needs with environmental conditions, determine the optimal fishing time based on fish growth patterns, and significantly improve aquaculture efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0091] Figure 1 This is a schematic diagram of an intelligent large-scale ecological fishery management system based on multi-source data fusion in an embodiment of the present application. DETAILED DESCRIPTION

[0092] The following is a clear and complete description of the technical solutions in the embodiments of the present application in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0093] like Figure 1 As shown, an intelligent large-scale ecological fishery management system based on multi-source data fusion includes a cloud and a dynamic response acquisition module. The cloud includes an integrated visual module, a multi-source data prediction module, a real-time monitoring module, a hidden trend analysis module, and an intelligent decision support module.

[0094] Building an air-ground integrated network, wherein the cloud and the dynamic response collection module are connected via the air-ground integrated network communication;

[0095] The dynamic response acquisition module is used to divide the large water surface aquaculture area into several sub-areas, set up environmental monitoring points and biological monitoring points in each sub-area, and collect multi-source data, including water quality monitoring indicators, meteorological monitoring indicators and biological characteristic indicators;

[0096] The integrated visualization module is used to time-calibrate the multi-source data of the dynamic response acquisition module, fuse the multi-source data with the same receiving timestamp and sending timestamp into a set of time data, and construct an integrated visualization graph of the multi-source data based on the multiple sets of time data;

[0097] The multi-source data prediction module is used to build a multi-source data prediction model and output the real-time threshold intervals corresponding to each water quality monitoring indicator, meteorological monitoring indicator, and biological characteristic indicator in the current collection period;

[0098] The real-time monitoring module is used to monitor each sub-area in the integrated visual graph of multi-source data in real time, and generate sudden abnormality alarms or perform hidden trend analysis based on the real-time monitoring results;

[0099] The hidden trend analysis module obtains abnormal trend feature sets of several historical abnormal collection cycles, performs hidden trend analysis on the time data of several collection cycles of the sub-region based on the abnormal trend feature sets of several historical abnormal collection cycles, and performs intelligent decision support operations on the sub-region or marks the sub-region as an ecological risk area based on the hidden trend analysis results;

[0100] The intelligent decision support module is used to construct a dual-drive fishing algorithm of Gompertz growth model + real-time feeding recognition, and the dual-drive fishing algorithm is used to obtain the optimal fishing time based on multiple sets of time data in the sub-area.

[0101] It should be further explained that, based on the above dual-driven fishing algorithm, ecological assessment results (such as water quality health index and fish growth curve) can be further combined with fishery production goals (such as annual production targets and ecological carrying capacity thresholds) to provide multi-dimensional decision support, for example:

[0102] Stocking density recommendations: Based on the sub-regional water volume, dissolved oxygen threshold (obtained through the multi-source data prediction module), and fish growth capacity simulated by the Gompertz model, a density-growth rate regression model is established to output the optimal stocking density range for each sub-region. For example, when the mean dissolved oxygen value is greater than 6 mg / L, the stocking density is recommended to be increased by 10%-15%;

[0103] Disease warning: Using convolutional neural networks (CNNs), we analyze the correlation between water quality indicators (such as sudden increases in ammonia nitrogen and abnormal pH fluctuations) and biological characteristics (such as decreased fish activity) before historical disease outbreaks. This allows us to build a disease risk prediction model. When real-time data triggers the model's warning threshold, it automatically generates disease type predictions and prevention and control recommendations (such as administering probiotics and increasing the frequency of water changes).

[0104] It should be further explained that a visual management interface can be constructed based on the above-mentioned intelligent decision support module: an interactive management interface is constructed based on Unity3D digital twin technology, integrating the following functions, such as:

[0105] 3D Situational Awareness: Based on the integrated visual map of multi-source data, real-time data dynamic rendering is superimposed (such as red highlighting of areas with excessive water quality indicators and heat maps of fish-dense areas). Managers can zoom in and out and click on points to view detailed data (such as the hourly dissolved oxygen change curve at a monitoring point).

[0106] Intuitive display of decision-making information: Key indicators (such as ecological risk level of each sub-region, recommended stocking density, and optimal fishing window) are presented in the form of dashboards, and management tasks within the aquaculture cycle (such as feeding plans and equipment maintenance reminders) are visualized through Gantt charts;

[0107] Mobile terminal adaptation: Develop a lightweight web interface that supports real-time data viewing and alarm reception on mobile phones / tablets. Managers can directly issue decision-making instructions through mobile terminals (such as remote control of aerator start and stop, and adjustment of feed amount).

[0108] It should be further explained that a system self-learning and optimization module can be constructed based on the above-mentioned intelligent decision support module: an online learning mechanism is used to continuously optimize the core model of the system, specifically implementing, for example:

[0109] Closed-loop data feedback: The actual results of managers' implementation of decision-making recommendations (such as yield data after adjusting stocking density according to the recommendations, and the effectiveness scores of disease control measures) are used as feedback signals and input into the multi-source data prediction model and hidden trend analysis module to update model parameters (such as adjusting the growth rate coefficient β of the Gompertz model and optimizing the similarity threshold of the abnormal trend feature set).

[0110] Reinforcement learning to optimize decision rules: A decision-making effectiveness evaluation function (e.g., ecological benefit index × 0.6 + economic benefit index × 0.4) is constructed, and deep reinforcement learning (DRL) algorithms are used to iteratively optimize decision rules such as stocking density and fishing strategies. For example, by simulating the results of aquaculture cycles under different strategies, a more optimal combination of management rules can be automatically generated.

[0111] Model version management: Historical model parameters and optimization records are stored in a versioned manner, supporting one-click rollback to the model status at a specific period, making it easier to compare system performance at different optimization stages.

[0112] It should be further explained that, in the specific implementation process, the large water surface aquaculture area is divided into several sub-areas, and the process of setting up environmental monitoring points in each sub-area includes:

[0113] Obtain the water body physical characteristics (water depth and topography), environmental characteristics (water flow and wind direction, meteorological conditions), and biological characteristics (fish species and ecological niche, stocking density) of large water surface aquaculture areas, and divide the large water surface aquaculture areas into several sub-areas based on the water body physical characteristics;

[0114] The historical pollution frequencies of several sub-regions are obtained based on the historical data of several sub-regions. The water physical characteristics, environmental characteristics and historical pollution frequencies of each sub-region are used as evaluation indicators. The indicator weights of the evaluation indicators are set. The membership matrix of each sub-region to the preset importance level is obtained through fuzzy comprehensive evaluation.

[0115] The importance level of each sub-area is obtained based on the membership matrix and the indicator weight. The number and distribution of environmental monitoring points are obtained based on the importance level of the sub-area and the location characteristics (the location coordinate set that constitutes the sub-area). The environmental monitoring points are used to collect water quality monitoring indicators (water temperature, dissolved oxygen, pH value, ammonia nitrogen, etc.) and meteorological monitoring indicators (air temperature, wind speed, wind direction, air pressure, precipitation, etc.), mark the collection time, and set the collection cycle.

[0116] It should be further explained that, in the specific implementation process, the process of obtaining the importance level of each sub-region according to the membership matrix and indicator weight includes:

[0117] The fuzzy comprehensive evaluation matrix of the evaluation indicators is obtained by integrating the indicator weights and the membership matrix of the evaluation indicators through a formula. The membership of each sub-region for different importance levels is obtained according to the fuzzy comprehensive evaluation matrix. The importance level with the highest membership corresponding to each sub-region is screened out, and the importance level with the highest membership corresponding to each sub-region is used as the importance level of each sub-region.

[0118] Wherein, the formula is:

[0119] ;

[0120] in, is the fuzzy comprehensive evaluation matrix of evaluation indicators, is the indicator weight of the evaluation index, is the membership matrix, " represents the multiplication of the elements at the corresponding positions of the weight matrix of the evaluation index and the membership matrix, It is the weighting parameter used to control the balance between the weight matrix and the membership matrix in the fuzzy comprehensive evaluation matrix of the evaluation index.

[0121] It should be further explained that, in the specific implementation process, the process of obtaining the number of environmental monitoring points and their distribution according to the importance level and location characteristics of the sub-areas includes:

[0122] The number of environmental monitoring points corresponding to different importance levels is preset, and the number n of environmental monitoring points in the sub-area is determined according to the importance level of the sub-area;

[0123] Initialize the locations of environmental monitoring points in the sub-area and set the coordinate set P of the initial environmental monitoring points: ;

[0124] Construct a polygonal monitoring area for each initial environmental monitoring point. , define its polygon monitoring area as all The set of points whose distance is less than the distance to other initial environmental monitoring points :

[0125] ;

[0126] in, Initial environmental monitoring point The maximum monitoring area, From point Q to the initial environmental monitoring point The Euclidean distance of : ;

[0127] Optimize the initial environmental monitoring point locations, with the goal of minimizing the area variance of each polygon monitoring area, and iteratively adjust the node positions to make the coverage more uniform. The objective function is:

[0128] ;

[0129] in, is the area of ​​the i-th polygon monitoring area, is the average area.

[0130] It should be further explained that, in the specific implementation process, the process of setting up biological monitoring points in each sub-area includes:

[0131] Preset the point density threshold per unit area (e.g. 1-2 points per 100 hectares), based on the total coverage area of ​​the sub-region Point density threshold per unit area , obtain the basic number of biological monitoring points in the sub-area , , Indicates rounding up, obtaining the terrain characteristics of the sub-region (such as shallows, deep pools, gullies, and water confluences) based on the physical characteristics of the water body in the sub-region, presetting the correction coefficients corresponding to different terrain characteristics, and Basic number of biological monitoring points Make corrections and obtain the corrected number of biological monitoring points ;

[0132] According to the biological characteristics of the sub-area (fish species and ecological niche, breeding density), the fish species and ecological niche (fish habitat water layers (surface, middle, lower, bottom) and activity areas (such as migration channels, feeding grounds, spawning grounds)) and breeding density parameters are obtained. According to the fish species and ecological niche and breeding density parameters, the revised number of biological monitoring points is weighted to obtain the final number of biological monitoring points. According to the final number of biological monitoring points, the point distribution of the biological monitoring points in the sub-area is obtained. The biological monitoring points (equipped with underwater camera arrays (equipped with AI visual recognition algorithms, resolution ≥4K), sonar monitoring equipment (detecting fish population density, swimming layer distribution)) are used to collect biological characteristic indicators, mark the collection time, and set the collection cycle.

[0133] It should be further explained that, in the specific implementation process, the revised number of biological monitoring points is weighted according to the fish species, ecological niche and stocking density parameters. The process of obtaining the final number of biological monitoring points includes:

[0134] ;

[0135] in, is the final number of biological monitoring points, is the maximum stocking density, is the average stocking density, is the niche width of the hth fish species (obtained in advance through habitat utilization assessment), is the reference niche width (e.g., 1 for a single species);

[0136] It should be further explained that, in the specific implementation process, the process of obtaining the distribution of biological monitoring points in the sub-area based on the final number of biological monitoring points includes:

[0137] Initialize the positions of biological monitoring points in the sub-area and set the coordinate set G of the initial biological monitoring points: ;

[0138] Construct a polygonal monitoring area for each initial biological monitoring point. , define its polygon monitoring area as all The set of points whose distance is less than the distance to other initial biological monitoring points :

[0139] ;

[0140] in, Initial biological monitoring site The maximum monitoring area, From point Q to the initial biological monitoring point The Euclidean distance of

[0141] Optimize the initial biological monitoring point locations, with the goal of minimizing the area variance of each polygon monitoring area, and iteratively adjust the node positions to make the coverage more uniform. The objective function is:

[0142] ;

[0143] in, is the area of ​​the i-th polygon monitoring area, is the average area.

[0144] It should be further explained that, in the specific implementation process, the process of building an integrated air-ground network includes:

[0145] The water depth and offshore distance of each sub-area are obtained based on the physical characteristics of the water body and location features of each sub-area. Each sub-area is divided into shallow water area, deep water area and offshore area according to the offshore distance and water depth of each sub-area. Among them, shallow water area: the water depth is usually less than 5 meters and the distance from the shore is ≤10 kilometers, which is suitable for deploying LoRaWAN (coverage radius 3-5 kilometers, low power consumption, strong penetration); deep water area: the water depth is ≥5 meters and the distance from the shore is ≤10 kilometers, and there is 5G signal coverage; offshore area: deep water area with a distance from the shore greater than 10 kilometers or without 5G coverage, relying on satellite communication (such as BGAN satellite terminal, global coverage but high latency);

[0146] According to the coverage area of ​​shallow water area, LoRaWAN nodes are deployed in shallow water area using regular hexagonal grid coverage. The number of nodes required for LoRaWAN nodes is , ,in To round up, is the coverage area of ​​the shallow water area, is the maximum monitoring range of the LoRaWAN node. 5G nodes are deployed in a grid-like distribution in the deep water area according to the coverage area of ​​the deep water area. The number of 5G nodes required is , ,in To round up, is the coverage area of ​​the deep water area, is the coverage radius of the 5G base station. Based on the coverage area of ​​the offshore area, BGAN satellite nodes are deployed in the offshore area using a regular hexagonal grid coverage method. The number of BGAN satellite nodes required is , ,in To round up, is the coverage area of ​​the offshore area, is the maximum monitoring range of the BGAN satellite node;

[0147] The data transmission path in shallow water areas is: LoRaWAN node → shore gateway → 5G / fiber → cloud;

[0148] The data transmission path in the deepwater area (5G coverage) is: 5G node → 5G base station → core network → cloud;

[0149] The data transmission path in the offshore area is: BGAN satellite node → satellite → ground station → Internet → cloud;

[0150] For large water surfaces with complex terrain (such as deep waters and offshore areas), a hybrid network architecture of LoRaWAN+5G / satellite communication is designed: low-power LoRa is used in shallow waters (coverage radius of 3-5 kilometers), and data is transmitted through 5G nodes and BGAN satellite nodes in deep waters and offshore areas, building communication coverage without blind spots and solving the signal blind spot problem of traditional single networks in open waters (coverage rate increased from 60% to 98%).

[0151] It should be further explained that, in the specific implementation process, the multi-source data of the dynamic response acquisition module is time-calibrated, and the multi-source data with the same receiving timestamp and sending timestamp are fused into a set of time data. The process of constructing an integrated visual graph of the multi-source data based on the multiple sets of time data includes:

[0152] Obtain multi-source data from the dynamic response acquisition module, and record the sending timestamp of the multi-source data based on the timing function of the Beidou system, and determine whether the cloud has received the multi-source data, wherein the cloud receives Beidou standard timing data according to the acquisition cycle and obtains the regional time on the cloud, compares the Beidou standard timing data with the regional time for consistency, and if inconsistent, updates the regional time on the cloud according to the Beidou standard timing data to obtain the standard regional time, and if so, records the receiving timestamp of the multi-source data based on the timing function of the Beidou system (records the receiving timestamp of the multi-source data according to the standard regional time on the cloud to ensure time consistency between the cloud and the dynamic response acquisition module),

[0153] A 3D model layer of a large aquaculture area is constructed based on Unity3D digital twin technology. Multi-source data is divided into multiple groups of time data based on their receiving and sending timestamps (multi-source data with the same receiving and sending timestamps are merged into one group). The multiple groups of time data are then divided into a water quality monitoring indicator time layer, a meteorological monitoring indicator time layer, and a biological characteristic indicator time layer.

[0154] Taking the three-dimensional model layer as the base layer, the water quality monitoring indicator time layer, the meteorological monitoring indicator time layer and the biological characteristic indicator time layer are superimposed on the base layer to obtain an integrated visual graph of multi-source data.

[0155] It should be further explained that, in the specific implementation process, the process of obtaining an integrated visual graph of multi-source data includes:

[0156] Obtain the water body physical characteristics and facility data of the large water surface aquaculture area. With the geometric center of the large water surface aquaculture area as the origin, the X-axis corresponds to the UTM easting direction, the Z-axis corresponds to the northing direction, and the Y-axis is the elevation (vertical direction). Construct a global coordinate system. Perform terrain modeling and facility modeling based on the water body physical characteristics and facility data. Obtain the terrain model of the large water surface aquaculture area, the three-dimensional model of the aquaculture facilities, and the three-dimensional model of the monitoring points (including the three-dimensional model of the environmental monitoring points and the three-dimensional model of the biological monitoring points). Determine the model coordinates of the three-dimensional model of the aquaculture facilities and the three-dimensional model of the monitoring points in the terrain model of the large water surface aquaculture area based on the facility data. Match the three-dimensional model of the aquaculture facilities and the three-dimensional model of the monitoring points with the terrain model of the large water surface aquaculture area based on the model coordinates to obtain the three-dimensional model layer of the large water surface aquaculture area.

[0157] The system obtains matching relationships between multi-source data and the 3D models of monitoring locations. Based on these matching relationships, it overlays time layers for water quality monitoring indicators, meteorological monitoring indicators, and biometric indicators on the 3D model layer to create an integrated visual representation of the multi-source data. This integrated visual representation, built using Unity3D digital twin technology, achieves a precise 1:1 mapping of the virtual farming scene to the physical farming site, improving management efficiency by 50%.

[0158] It should be further explained that, in the specific implementation process, the process of constructing a multi-source data prediction model and outputting the real-time threshold ranges corresponding to each water quality monitoring indicator, meteorological monitoring indicator, and biological characteristic indicator in the current collection period includes:

[0159] Construct a multi-source data prediction model based on deep learning, obtain multi-source data from several historical collection cycles as training sets and test sets, input the training sets into the multi-source data prediction model for training until the loss function training is stable, save the model parameters, test the multi-source data prediction model with the test set until it meets the preset requirements, and output the multi-source data prediction model;

[0160] According to the multi-source data model, the predicted numerical time series sequence corresponding to each water quality monitoring indicator, meteorological monitoring indicator and biometric indicator in the current collection period is output. According to the predicted numerical time series sequence corresponding to each water quality monitoring indicator, meteorological monitoring indicator and biometric indicator in the current collection period, the real-time threshold range corresponding to each water quality monitoring indicator, meteorological monitoring indicator and biometric indicator in the current collection period is obtained.

[0161] In building a multi-source data prediction model based on deep learning, the present invention selects a convolutional neural network (CNN) suitable for time series analysis as the deep learning architecture and uses the MES loss function. Training begins by inputting a prepared training set into the selected deep learning model. During training, weights are continuously updated using a backpropagation algorithm, gradually reducing the loss function until a steady state is reached. During this process, techniques such as early stopping are utilized to prevent overfitting. In addition to the basic training process, various model parameters, including the learning rate, batch size, and regularization coefficient, are tuned using grid search.

[0162] After model training is complete and parameter adjustments are complete, a final evaluation is performed on the test set to obtain the model's evaluation results. These results include classification metrics such as accuracy, recall, and F1 score. Based on the evaluation results on the test set, it is determined whether the model meets the expected standards. If the requirements are met, the model parameters are saved and prepared for deployment. If not, it is necessary to return to a previous stage to re-examine issues such as data quality, model structure, or training strategy.

[0163] It should be further explained that, in the specific implementation process, the process of real-time monitoring of each sub-area in the multi-source data integrated visual graph and generating sudden abnormality alerts or performing hidden trend analysis based on the real-time monitoring results includes:

[0164] Mark each water quality monitoring indicator, meteorological monitoring indicator, and biometric indicator in the water quality monitoring indicator time layer, meteorological monitoring indicator time layer, and biometric indicator time layer as a first indicator, obtain the real-time threshold interval corresponding to each first indicator in the current acquisition period, obtain the numerical time series sequence corresponding to the first indicator in the current acquisition period, compare the numerical time series sequence corresponding to the first indicator with the corresponding real-time threshold interval, and obtain the cumulative time that the numerical time series sequence corresponding to the first indicator is not within the corresponding real-time threshold interval;

[0165] A cumulative time threshold is preset. If the cumulative time corresponding to a first indicator is greater than the cumulative time threshold, the first indicator is marked as a key monitoring indicator, an emergency abnormality alarm is generated, the layer to which the key monitoring indicator belongs and the location of the multi-source data integrated visual map are highlighted in red, and the current collection period is marked as an abnormal collection period;

[0166] If the cumulative time corresponding to each first indicator is less than or equal to the cumulative time threshold, the current collection period is marked as a normal collection period, and a hidden trend analysis is performed.

[0167] It should be further explained that, in a specific implementation process, the process of the hidden trend analysis module obtaining the abnormal trend feature set of several historical abnormal collection cycles includes:

[0168] Extracting the numerical time series sequences corresponding to the key monitoring indicators and other first indicators within the historical anomaly collection period, and obtaining the trend correlation coefficient between the key monitoring indicators and other first indicators and the stability coefficient of the other first indicators based on the numerical time series sequences;

[0169] Extract the numerical time series corresponding to the key monitoring indicators and other first indicators of the previous k historical normal collection cycles of the historical abnormal collection cycle; obtain the trend correlation coefficients between the key monitoring indicators and other first indicators of the previous k historical normal collection cycles of the historical abnormal collection cycle and the stability coefficients of the other first indicators based on the numerical time series corresponding to the key monitoring indicators and other first indicators of the previous k historical normal collection cycles of the historical abnormal collection cycle; obtain the average trend correlation coefficients between the key monitoring indicators and other first indicators of the previous k historical normal collection cycles of the historical abnormal collection cycle and the average stability coefficients of the other first indicators based on the trend correlation coefficients between the key monitoring indicators and other first indicators of the previous k historical normal collection cycles;

[0170] Obtain the trend correlation change coefficient between the key monitoring indicator and the other first indicators based on the trend correlation coefficient and the average trend correlation coefficient between the key monitoring indicator and the other first indicators; obtain the stability change coefficient of the other first indicators based on the stability coefficient and the average stability coefficient of the other first indicators;

[0171] An abnormal trend feature set of a historical abnormal collection period is constructed based on the trend correlation change coefficient between the key monitoring indicator and other first indicators and the stability change coefficient of other first indicators.

[0172] It should be further explained that, in the specific implementation process, the process of obtaining the trend correlation coefficient between the key monitoring indicator and the other first indicators and the stability coefficient of the other first indicators includes:

[0173] ;

[0174] ;

[0175] in, Indicates the trend correlation coefficient between the key monitoring indicator a and other first indicators b, represents the stability coefficient of the other first indicator b, Indicates the value of the key monitoring indicator a at the tth moment, represents the value of the other first indicator b at the tth moment, Indicates the average value of the key monitoring indicator a during the collection period. Indicates the average value of other first indicators b during the collection period, Indicates the total number of moments included in the collection period.

[0176] The process of obtaining the trend correlation coefficient of the key monitoring indicator and other first indicators and the stability coefficient of the other first indicators includes:

[0177] ;

[0178] ;

[0179] in, Indicates the trend correlation coefficient between the key monitoring indicator a and the other first indicator b, It represents the average trend correlation coefficient between the key monitoring indicator a and other first indicators b, represents the stability variation coefficient of the other first indicator b, Represents the average stability coefficient of other first indicators b.

[0180] It should be further explained that, in the specific implementation process, hidden trend analysis is performed on multiple sets of time data of the sub-region based on the abnormal trend feature sets of several historical abnormal collection cycles. The process of performing intelligent decision support operations on the sub-region or marking the sub-region as an ecological risk area based on the hidden trend analysis results includes:

[0181] Select the first indicator with the longest cumulative time in the current collection period of the sub-region, and mark the first indicator as the estimated key monitoring indicator;

[0182] Obtain the numerical time series sequence corresponding to the first indicator and the estimated key monitoring indicator in the k collection periods before the current collection period. Based on the numerical time series sequence corresponding to the first indicator and the estimated key monitoring indicator in the current collection period and the k collection periods before the current collection period, obtain the trend correlation change coefficient between the estimated key monitoring indicator and other first indicators in the current collection period and the stability change coefficient of other first indicators;

[0183] Compare the trend correlation change coefficients between the estimated key monitoring indicators and other first indicators in the current collection period and the stability change coefficients of other first indicators with the abnormal trend feature sets of several historical abnormal collection periods to obtain the similarity of each abnormal trend feature set (obtained by using cosine similarity calculation method);

[0184] A similarity threshold is preset. If the similarity of each abnormal trend feature set is less than or equal to the similarity threshold, an intelligent decision support operation is performed. If the similarity of an abnormal trend feature set is greater than the similarity threshold, the sub-region is marked as an ecological risk area, the location of the sub-region in the multi-source data integrated visual map is highlighted in orange, and emergency treatment measures for the sub-region are generated.

[0185] It should be further explained that, in the specific implementation process, the process of marking a sub-region as an ecological anomaly region and generating emergency treatment measures for the sub-region includes: building a sudden ecological anomaly database, obtaining emergency treatment measures corresponding to several historical anomaly collection cycles, matching abnormal trend feature sets within several historical anomaly collection cycles with the emergency treatment measures corresponding to several historical anomaly collection cycles, and storing them in the sudden ecological anomaly database;

[0186] When the similarity of an abnormal trend feature set is greater than a similarity threshold, a search is performed in the sudden ecological anomaly database according to the abnormal trend feature set to obtain emergency treatment measures corresponding to the abnormal trend feature set.

[0187] It should be further explained that, during the specific implementation process, the intelligent decision support module constructs a dual-drive fishing algorithm consisting of a Gompertz growth model and real-time feeding recognition. The process of using the dual-drive fishing algorithm to obtain the optimal fishing opportunity based on multiple sets of time data in the sub-area includes the following:

[0188] Based on the numerical time series corresponding to the first indicator of the current collection period, the dissolved oxygen correction factor, water temperature correction factor and real-time food intake of the current collection period are obtained. Based on the numerical time series corresponding to the first indicator of several historical collection periods, nonlinear fitting is performed to construct a Gompertz growth model. Based on the Gompertz growth model, the average weight and theoretical growth rate of the fish are obtained.

[0189] Among them, the Gompertz growth model formula is:

[0190] ;

[0191] in, Indicates time The average weight of fish, Indicates the maximum weight of the fish, exp() is the natural exponential function, Indicates the breeding time, and The nonlinear fitting is performed by obtaining the numerical time series corresponding to the first indicator of several historical collection periods. It should be noted that is a constant, reflecting the initial growth conditions and early growth rate, The larger it is, the faster the initial growth rate (the curve shifts to the left). The theoretical growth rate determines how quickly the growth curve reaches its maximum weight. The bigger you are, the faster you grow and approach your maximum weight. ;

[0192] The actual growth rate is obtained based on the theoretical growth rate, dissolved oxygen correction factor and water temperature correction factor. The feeding efficiency coefficient is obtained based on the actual growth rate, average fish weight and real-time food intake. The optimal fishing weight is preset, and the optimal fishing time is obtained based on the average fish weight, optimal fishing weight and feeding efficiency coefficient.

[0193] The formula for obtaining the actual growth rate based on the theoretical growth rate, dissolved oxygen correction factor, and water temperature correction factor is:

[0194] ;

[0195] ;

[0196] ;

[0197] in, represents the actual growth rate, is the dissolved oxygen correction factor, is the water temperature correction factor, Indicates the dissolved oxygen value, Indicates the minimum threshold of dissolved oxygen, is the optimum dissolved oxygen value, For water temperature, is the minimum water temperature threshold, For the optimum water temperature, is the maximum water temperature threshold;

[0198] The formula for obtaining the feeding efficiency coefficient based on the actual growth rate, average fish weight, and real-time food intake is:

[0199] ;

[0200] in, Indicates time Real-time food intake, represents the feeding efficiency coefficient;

[0201] The formula for obtaining the best fishing time based on the average fish weight, optimal fishing weight and feeding efficiency coefficient is:

[0202] ;

[0203] in, For the best fishing weight, is the index, is the preset parameter, To find the best time to fish, the above formulas are all calculated by removing dimensions and taking numerical values. The formula is a formula that is closest to the actual situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and preset thresholds in the formula are set by technicians in this field according to actual conditions or obtained by simulating a large amount of data.

[0204] The above embodiments are only used to illustrate the technical method of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. An intelligent large-scale ecological fishery management system based on multi-source data fusion, characterized by: It includes cloud and dynamic response acquisition modules. The cloud includes an integrated visualization module, a multi-source data prediction module, a real-time monitoring module, a hidden trend analysis module, and an intelligent decision support module. Building an air-ground integrated network, wherein the cloud and the dynamic response collection module are connected via the air-ground integrated network communication; The dynamic response acquisition module is used to divide the large water surface aquaculture area into several sub-areas, set up environmental monitoring points and biological monitoring points in each sub-area, and collect multi-source data, including water quality monitoring indicators, meteorological monitoring indicators and biological characteristic indicators; The integrated visualization module is used to time-calibrate the multi-source data of the dynamic response acquisition module, fuse the multi-source data with the same receiving timestamp and sending timestamp into a set of time data, and construct an integrated visualization graph of the multi-source data based on the multiple sets of time data; The multi-source data prediction module is used to build a multi-source data prediction model and output the real-time threshold intervals corresponding to various water quality monitoring indicators, meteorological monitoring indicators, and biological characteristic indicators in the current collection period; The real-time monitoring module is used to monitor each sub-area in the integrated visual graph of multi-source data in real time, and generate sudden abnormality alarms or perform hidden trend analysis based on the real-time monitoring results; The hidden trend analysis module obtains the abnormal trend feature set of several historical abnormality collection cycles, performs hidden trend analysis on the sub-region, and performs intelligent decision support operations on the sub-region or marks the sub-region as an ecological risk area based on the hidden trend analysis results; The intelligent decision support module is used to build a dual-drive fishing algorithm that combines the Gompertz growth model with real-time feeding recognition. This algorithm is used to determine the optimal fishing opportunity based on multiple sets of time data from sub-areas, including: Based on the numerical time series corresponding to the first indicator of the current collection period, the dissolved oxygen correction factor, water temperature correction factor and real-time food intake of the current collection period are obtained. Based on the numerical time series corresponding to the first indicator of several historical collection periods, nonlinear fitting is performed to construct a Gompertz growth model. Based on the Gompertz growth model, the average weight and theoretical growth rate of the fish are obtained. The actual growth rate is obtained based on the theoretical growth rate, dissolved oxygen correction factor and water temperature correction factor. The feeding efficiency coefficient is obtained based on the actual growth rate, average fish weight and real-time food intake. The optimal fishing weight is preset, and the optimal fishing time is obtained based on the average fish weight, optimal fishing weight and feeding efficiency coefficient.

2. The intelligent large-scale ecological fishery management system based on multi-source data fusion according to claim 1 is characterized in that: The process of dividing the large aquaculture area into several sub-areas and setting up environmental monitoring points in each sub-area includes: Obtain the water physical characteristics, environmental characteristics, and biological characteristics of large water surface aquaculture areas, and divide the large water surface aquaculture areas into several sub-areas based on the water physical characteristics; The historical pollution frequencies of several sub-regions are obtained based on the historical data of several sub-regions. The water physical characteristics, environmental characteristics and historical pollution frequencies of each sub-region are used as evaluation indicators. The indicator weights of the evaluation indicators are set. The membership matrix of each sub-region to the preset importance level is obtained through fuzzy comprehensive evaluation. The importance level of each sub-area is obtained according to the membership matrix and the indicator weight. The number and distribution of environmental monitoring points are obtained according to the importance level and location characteristics of the sub-area. The environmental monitoring points are used to collect water quality monitoring indicators and meteorological monitoring indicators, mark the collection time, and set the collection cycle.

3. The intelligent large-scale ecological fishery management system based on multi-source data fusion according to claim 2 is characterized in that: The process of setting up biological monitoring points in each sub-area includes: A point density threshold per unit area is preset, and the basic number of biological monitoring points in the sub-area is obtained based on the total covered area of ​​the sub-area using the point density threshold per unit area. The terrain characteristics of the sub-area are obtained based on the physical characteristics of the water body in the sub-area, and correction coefficients corresponding to different terrain characteristics are preset. The basic number of biological monitoring points is corrected according to the correction coefficients corresponding to the terrain characteristics of the sub-area to obtain the corrected number of biological monitoring points. According to the biological characteristics of the sub-area, the fish species, ecological niche and breeding density parameters are obtained, and the revised number of biological monitoring points is weighted according to the fish species, ecological niche and breeding density parameters to obtain the final number of biological monitoring points. According to the final number of biological monitoring points, the point distribution of the biological monitoring points in the sub-area is obtained. The biological monitoring points are used to collect biological characteristic indicators, mark the collection time, and set the collection cycle.

4. The intelligent large-scale ecological fishery management system based on multi-source data fusion according to claim 3 is characterized in that: The process of building an integrated air-ground network includes: The water depth and offshore distance of each sub-area are obtained according to the water body physical characteristics and location characteristics of each sub-area, and each sub-area is divided into shallow water area, deep water area and offshore area according to the offshore distance and water depth of each sub-area; LoRaWAN nodes are deployed in shallow water areas using a regular hexagonal grid coverage method based on the coverage area of ​​shallow water areas. 5G nodes are deployed in deep water areas using a grid distribution method based on the coverage area of ​​deep water areas. BGAN satellite nodes are deployed in offshore areas using a regular hexagonal grid coverage method based on the coverage area of ​​offshore areas.

5. The intelligent large-scale ecological fishery management system based on multi-source data fusion according to claim 4 is characterized in that: The process of time-calibrating the multi-source data of the dynamic response acquisition module, fusing the multi-source data with the same receiving and sending timestamps into a set of time data, and constructing an integrated visual graph of the multi-source data based on the multiple sets of time data includes: Obtain multi-source data from the dynamic response acquisition module, and record the sending timestamp of the multi-source data based on the timing function of the Beidou system, determine whether the cloud has received the multi-source data, and if so, record the receiving timestamp of the multi-source data based on the timing function of the Beidou system, Construct a 3D model layer for a large aquaculture area, divide the multi-source data into multiple groups of time data according to the receiving and sending timestamps of the multi-source data, and divide the multiple groups of time data into a water quality monitoring indicator time layer, a meteorological monitoring indicator time layer, and a biological characteristic indicator time layer; Taking the three-dimensional model layer as the base layer, the water quality monitoring indicator time layer, the meteorological monitoring indicator time layer and the biological characteristic indicator time layer are superimposed on the base layer to obtain an integrated visual graph of multi-source data.

6. The intelligent large-scale ecological fishery management system based on multi-source data fusion according to claim 5 is characterized in that: The process of building a multi-source data prediction model to output the real-time threshold intervals corresponding to each water quality monitoring indicator, meteorological monitoring indicator, and biological characteristic indicator in the current collection period includes: Construct a multi-source data prediction model based on deep learning, obtain multi-source data from several historical collection cycles as training sets and test sets, input the training sets into the multi-source data prediction model for training until the loss function training is stable, save the model parameters, test the multi-source data prediction model with the test set until it meets the preset requirements, and output the multi-source data prediction model; According to the multi-source data model, the predicted numerical time series sequence corresponding to each water quality monitoring indicator, meteorological monitoring indicator and biometric indicator in the current collection period is output. According to the predicted numerical time series sequence corresponding to each water quality monitoring indicator, meteorological monitoring indicator and biometric indicator in the current collection period, the real-time threshold range corresponding to each water quality monitoring indicator, meteorological monitoring indicator and biometric indicator in the current collection period is obtained.

7. The intelligent large-scale ecological fishery management system based on multi-source data fusion according to claim 6 is characterized in that: The process of real-time monitoring of each sub-area in the integrated visual graph of multi-source data and generating sudden abnormality alerts or performing hidden trend analysis based on the real-time monitoring results includes: Mark each water quality monitoring indicator, meteorological monitoring indicator, and biometric indicator in the water quality monitoring indicator time layer, meteorological monitoring indicator time layer, and biometric indicator time layer as a first indicator, obtain the real-time threshold interval corresponding to each first indicator in the current acquisition period, obtain the numerical time series sequence corresponding to the first indicator in the current acquisition period, compare the numerical time series sequence corresponding to the first indicator with the corresponding real-time threshold interval, and obtain the cumulative time that the numerical time series sequence corresponding to the first indicator is not within the corresponding real-time threshold interval; A cumulative time threshold is preset. If the cumulative time corresponding to a first indicator is greater than the cumulative time threshold, the first indicator is marked as a key monitoring indicator, a sudden abnormality alarm is generated, and the current collection period is marked as an abnormal collection period; If the cumulative time corresponding to each first indicator is less than or equal to the cumulative time threshold, the current collection period is marked as a normal collection period, and a hidden trend analysis is performed.

8. The intelligent large-scale ecological fishery management system based on multi-source data fusion according to claim 7 is characterized in that: The process of obtaining anomaly trend feature sets for several historical anomaly collection periods includes: Extracting the numerical time series sequences corresponding to the key monitoring indicators and other first indicators within the historical anomaly collection period, and obtaining the trend correlation coefficient between the key monitoring indicators and other first indicators and the stability coefficient of the other first indicators based on the numerical time series sequences; Obtain the average trend correlation coefficient between the key monitoring indicators and other first indicators in the previous k historical normal collection periods of the historical abnormal collection period and the average stability coefficient of other first indicators; Obtain the trend correlation change coefficient between the key monitoring indicator and the other first indicators based on the trend correlation coefficient and the average trend correlation coefficient between the key monitoring indicator and the other first indicators; obtain the stability change coefficient of the other first indicators based on the stability coefficient and the average stability coefficient of the other first indicators; An abnormal trend feature set of a historical abnormal collection period is constructed based on the trend correlation change coefficient between the key monitoring indicator and other first indicators and the stability change coefficient of other first indicators.

9. The intelligent large-scale ecological fishery management system based on multi-source data fusion according to claim 8 is characterized in that: The process of performing hidden trend analysis on multiple sets of time data of a sub-region and performing intelligent decision support operations on the sub-region or marking the sub-region as an ecological risk area based on the hidden trend analysis results includes: Select the first indicator with the longest cumulative time in the current collection period of the sub-region, and mark the first indicator as the estimated key monitoring indicator; Obtain the numerical time series sequence corresponding to the first indicator and the estimated key monitoring indicator in the k collection periods before the current collection period. Based on the numerical time series sequence corresponding to the first indicator and the estimated key monitoring indicator in the current collection period and the k collection periods before the current collection period, obtain the trend correlation change coefficient between the estimated key monitoring indicator and other first indicators in the current collection period and the stability change coefficient of other first indicators; Compare the similarity between the trend correlation change coefficient between the estimated key monitoring indicators and other first indicators in the current collection period and the stability change coefficient of other first indicators with the abnormal trend feature sets of several historical abnormal collection periods to obtain the similarity of each abnormal trend feature set; A similarity threshold is preset. If the similarities of all abnormal trend feature sets are less than or equal to the similarity threshold, an intelligent decision support operation is performed. If the similarity of any abnormal trend feature set is greater than the similarity threshold, the sub-region is marked as an ecological risk area, and emergency treatment measures for the sub-region are generated.

Images