Marine ranching early warning method and system based on knowledge graph

Abstract

The invention discloses a marine ranching early warning method and system based on a knowledge graph, and relates to the technical field of marine ranching management, and the method comprises the knowledge graph of large-class keywords of water quality management, equipment management and biological resource management and progressive-level subclass keywords of the large-class keywords, terminal subclass association preset acquisition parameter groups and state description; generating a marine ranching plane graph; equipment states and detection parameters are collected to form a normal state parameter group, and a normal planar graph representation model is constructed; generating an exception model based on a preset exception adjustment normal parameter group, and dynamically associating the exception model with small-class keywords in the knowledge graph; and collecting data in real time to construct a real-time model, comparing the real-time model with an abnormal model to trigger an alarm, and marking an abnormal root in the knowledge graph. According to the method, the limitation of traditional single-dimensional monitoring is broken through, multi-source data are integrated through knowledge graph semantization, dynamic threshold and spatial correlation analysis are combined, abnormal accurate positioning and cross-dimensional risk early warning are achieved, and the marine ranch management efficiency and safety are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of marine ranch management, and in particular, to a marine ranch early warning method and system based on a knowledge graph. Background Art

[0002] As an important model for the intensive development of modern fisheries, the management of marine ranches needs to coordinate multi-dimensional elements such as water quality environment, equipment operation, and biological resources. However, the current technical system has significant limitations: traditional monitoring systems usually monitor water quality, equipment, or biological parameters in isolation, lacking the ability to analyze multi-dimensional data fusion, and it is difficult to identify compound risks (such as local water quality deterioration caused by a malfunction of a feeding machine); the static early warning mechanism relying on fixed thresholds (such as dissolved oxygen < 4mg / L) cannot adapt to the dynamic fluctuations of the marine environment (such as day-night temperature difference, seasonal changes), resulting in an increased false alarm rate; abnormal alarm information only indicates that a parameter exceeds the standard (such as "pH abnormal"), but is not associated with specific management links (equipment maintenance or water quality regulation), and manual troubleshooting of the fault source is required, with low response efficiency; in addition, existing technologies ignore the spatial correlation between equipment and detection points (such as the impact of a damaged net cage on the surrounding water flow), making it difficult to predict the abnormal diffusion path and exacerbating systemic risks. Summary of the Invention

[0003] The purpose of the present invention is to provide a method and system that can more accurately early warn a marine ranch based on knowledge graph analysis.

[0004] The present invention discloses a marine ranch early warning method based on a knowledge graph, including:

[0005] Step S100, determining the major management keywords for marine ranch management, and setting several levels of minor management keywords under each major management keyword. Based on the hierarchical relationship between the major management keywords and the minor management keywords, constructing a hierarchical connection line between the major management keywords and the minor management keywords to form a marine ranch knowledge graph. For the minor management keywords at the end of the marine ranch knowledge graph, several groups of preset collection parameter groups are set. The preset collection parameter groups include several types of collection parameter intervals, and a status description is configured for each collection parameter group;

[0006] Step S200, using a GIS map information system to determine the marine ranch floor plan, determining the locations of the equipment in the marine ranch and the locations of the detection points of the parameter detection points, and respectively configuring them on the marine ranch floor plan in the form of equipment mapping points and detection mapping points;

[0007] Step S300: When the marine ranch is operating normally, collect the status performances of different devices and the acquisition parameters corresponding to different detection points. Record the combination of the status performances of the devices and the acquisition parameters that appear simultaneously as the normal status parameter group, and map and associate the normal status parameter group with the corresponding marine ranch floor plan to obtain the normal marine ranch floor plan performance model;

[0008] Step S400: Based on the preset abnormal situations, adjust the normal status parameter group in the normal marine ranch floor plan performance model to obtain the abnormal marine ranch floor plan performance model. Associate the status description corresponding to the abnormal situation with the abnormal marine ranch floor plan performance model, and based on the abnormal status parameter group of the abnormal marine ranch floor plan performance model, determine the subclass keyword to which the abnormal marine ranch floor plan performance model belongs relative to the preset acquisition parameter group mapped in the marine ranch knowledge graph;

[0009] Step S500: When conducting real-time supervision of the marine ranch, construct a real-time marine ranch floor plan performance model based on the real-time status performance of the devices collected in real time and the real-time acquisition parameters. Compare the real-time marine ranch floor plan performance model with different abnormal marine ranch floor plan performance models. If there is a matching abnormal marine ranch floor plan performance model, an abnormal alarm will be issued, and based on the subclass management keyword corresponding to the abnormal marine ranch floor plan performance model, it will be marked and displayed on the marine ranch knowledge graph.

[0010] In some embodiments disclosed in the present invention, the major class management keywords include: water quality management, equipment management, and biological resource management.

[0011] In some embodiments disclosed in the present invention, the subclass management keywords include:

[0012] The subclass management keywords corresponding to water quality management include dissolved oxygen content, pH value, and water temperature;

[0013] The subclass management keywords corresponding to equipment management include aquaculture net cage monitoring, feeding system monitoring, and water quality detection equipment calibration;

[0014] The subclass management keywords corresponding to biological resource management include fish farming technology, shellfish harvesting plan, and seaweed farming area planning;

[0015] Among them, the sub-category management keywords corresponding to the monitoring of the aquaculture net also include structural integrity and position stability. The sub-category management keywords corresponding to the monitoring of the feeding system also include daily maintenance and fault handling. The sub-category management keywords corresponding to the calibration of water quality monitoring equipment also include accuracy and reliability. The sub-category management keywords corresponding to fish farming techniques also include seed selection, hatching, feeding, and disease prevention. The sub-category management keywords corresponding to the shellfish harvesting plan also include harvesting time, harvesting method, and harvesting quantity. The sub-category management keywords corresponding to the seaweed aquaculture area planning also include growth requirements, water flow conditions, and light intensity.

[0016] In some embodiments disclosed by the present invention, the method for determining each acquisition parameter range in the preset acquisition parameter group includes:

[0017] Step S101: Analyze the historical operation records of the marine ranch, determine several groups of historical state parameter groups, and classify the historical state parameter groups into normal and abnormal categories to form a normal historical state parameter set and an abnormal historical state parameter set. Then, based on the state descriptions in the historical operation records of the marine ranch, perform secondary classification on the normal historical state parameter set and the abnormal historical state parameter set respectively;

[0018] Step S102: Construct a vertical acquisition parameter axis for each detection point corresponding to the position on the marine ranch floor plan. Map the historical acquisition parameters corresponding to each detection point in the historical state parameter group to the acquisition parameter number axis in the form of historical acquisition parameter mapping points, and connect the historical acquisition parameter mapping points in pairs to obtain a mapping point connection diagram;

[0019] Step S103: Compare the mapping point connection diagrams corresponding to the historical state parameter sets after secondary classification for mutual equivalence. Based on the comparison results, perform tertiary classification on the normal historical state parameter set and the abnormal historical state parameter set to obtain a normal historical state parameter subset and an abnormal historical state parameter subset;

[0020] Step S104: Analyze the acquisition parameter ranges of the acquisition parameter groups in each historical state parameter subset to obtain the preset acquisition parameter group corresponding to the historical state parameter subset.

[0021] In some embodiments disclosed by the present invention, the method for comparing the mapping point connection diagrams corresponding to the historical state parameter sets after secondary classification for mutual equivalence includes:

[0022] Step S1031: Set different key weights for different parameter mapping points, compare the mapping point height differences of each parameter mapping point between the mapping point connection diagrams, and combine the key weights to construct a first equivalence operator;

[0023] Step S1032: Compare the coincidence scale of the mapped point connection lines in the mapped point connection diagram, and based on the coincidence scale, correct the first equivalent operator to obtain the equivalence degree between the mapped point connection diagrams.

[0024] Among them, the expression for calculating the equivalence degree is:

[0025] ;

[0026] Among them, T is the equivalence degree, H is the coincidence scale, K i is the key weight corresponding to the i-th parameter mapped point, h max is the preset maximum height difference, Δh i is the height difference of the i-th parameter mapped point, and n is the number of all parameter mapped points.

[0027] In some embodiments disclosed in the present invention, the method for determining the coincidence scale of the mapped point connection lines in the mapped point connection diagram includes:

[0028] Step S1033: Determine the corresponding mapped point connection lines in the mapped point connection diagram, set a number of coincidence detection points for the mapped point connection lines at a preset interval, determine the coincidence detection distance between the corresponding coincidence detection points, and calculate the average value of the coincidence detection distances to obtain the average coincidence detection distance.

[0029] Step S1034: Set a coincidence detection distance judgment interval for the average coincidence detection distance, and a coincidence parameter is set for each coincidence detection distance judgment interval. Based on the coincidence detection distance judgment interval to which each mapped point connection line belongs, determine the coincidence parameter corresponding to the mapped point connection line.

[0030] Step S1035: Based on the coincidence parameter corresponding to each mapped point connection line, determine the coincidence scale between the mapped point connection diagrams.

[0031] Among them, the expression for calculating the coincidence scale is:

[0032] ;

[0033] Among them, H is the coincidence scale, R is the coincidence parameter influence adjustment coefficient, c is the coincidence parameter influence adjustment constant, s max is the preset maximum coincidence parameter, s x is the coincidence parameter corresponding to the x-th mapped point connection line, and N is the number of mapped point connection lines.

[0034] In some embodiments disclosed in the present invention, the method for adjusting the normal marine ranch plan view performance model based on a preset abnormal situation includes:

[0035] Step S401: Analyze the historical operation records of the marine ranch, determine several groups of abnormal historical state parameters, and determine the corresponding state descriptions. Traverse and compare the preset situations with the determined state descriptions to determine several matching state descriptions, and screen out the preset acquisition parameter groups corresponding to the state descriptions.

[0036] Step S402: Based on several acquisition parameter ranges corresponding to the screened preset acquisition parameter groups, determine the adjustment range for each group of abnormal historical state parameters. After adjustment, several adjusted abnormal state parameter groups are obtained, and the normal state parameter groups are adjusted with the adjusted abnormal state parameter groups as the reference objects.

[0037] In some embodiments disclosed by the present invention, the method for judging the coincidence between the real-time marine ranch floor plan performance model and the abnormal marine ranch floor plan performance model includes:

[0038] Step S501: Compare the real-time marine ranch floor plan performance model with the abnormal marine ranch floor plan performance model, including the comparison of the state performances of the corresponding devices and the comparison of the acquisition parameters of the corresponding parameter detection points. If the performance differences between the state performances are all within the preset range and the parameter difference amounts between the acquisition parameters are all within the preset difference amount, it is determined that the real-time marine ranch floor plan performance model and the abnormal marine ranch floor plan performance model coincide.

[0039] In some embodiments disclosed by the present invention, a marine ranch early warning system based on a knowledge graph is also disclosed, including:

[0040] The first module is used to determine the major management keywords for marine ranch management, and several levels of minor management keywords are set under each major management keyword. Based on the hierarchical relationship between the major management keywords and the minor management keywords, hierarchical connections between the major management keywords and the minor management keywords are constructed to form a marine ranch knowledge graph. Several groups of preset acquisition parameter groups are set for the minor management keywords at the end of the marine ranch knowledge graph. The preset acquisition parameter groups include several types of acquisition parameter ranges, and a state description is configured for each acquisition parameter group.

[0041] The second module is used to use the GIS map information system to determine the marine ranch floor plan, determine the locations of the devices in the marine ranch and the locations of the detection points of the parameter detection points, and configure them on the marine ranch floor plan in the form of device mapping points and detection mapping points respectively.

[0042] The third module is used to collect the state performances of different devices and the acquisition parameters corresponding to different detection points when the marine ranch is operating normally, record the combination of the state performances of the devices that appear simultaneously and the acquisition parameters as the normal state parameter group, and map and associate the normal state parameter group with the corresponding marine ranch floor plan to obtain the normal marine ranch floor plan performance model;

[0043] The fourth module is used to adjust the normal state parameter group in the normal marine ranch floor plan performance model based on a preset abnormal situation to obtain the abnormal marine ranch floor plan performance model, associate the state description corresponding to the abnormal situation with the abnormal marine ranch floor plan performance model, and determine the subclass keyword to which the abnormal marine ranch floor plan performance model belongs based on the abnormal state parameter group of the abnormal marine ranch floor plan performance model relative to the preset acquisition parameter group mapped in the marine ranch knowledge graph;

[0044] The fifth module is used to, when conducting real-time supervision of the marine ranch, construct a real-time marine ranch floor plan performance model based on the real-time state performance of the devices collected in real time and the real-time acquisition parameters, compare the real-time marine ranch floor plan performance model with different abnormal marine ranch floor plan performance models. If there is a matching abnormal marine ranch floor plan performance model, an abnormal alarm is issued, and based on the subclass management keyword corresponding to the abnormal marine ranch floor plan performance model, it is marked and displayed on the marine ranch knowledge graph.

[0045] The present invention discloses a marine ranch early warning method and system based on a knowledge graph, which relates to the technical field of marine ranch management, including a knowledge graph of general class keywords and progressive hierarchical subclass keywords for water quality management, equipment management, and biological resource management, and the terminal subclass is associated with a preset acquisition parameter group and a state description; generate a marine ranch floor plan; collect device states and detection parameters to form a normal state parameter group, and construct a normal floor plan performance model; adjust the normal parameter group based on a preset abnormality to generate an abnormal model, and dynamically associate it with the subclass keywords in the knowledge graph; collect real-time data to construct a real-time model, compare it with the abnormal model to trigger an alarm, and mark the root cause of the abnormality in the knowledge graph. This method breaks through the limitations of traditional single-dimensional monitoring, semantically integrates multi-source data through the knowledge graph, combines dynamic thresholds and spatial association analysis, realizes accurate abnormal positioning and cross-dimensional risk early warning, and significantly improves the management efficiency and safety of marine ranches.

[0046] The technical solution of the present invention will be further described in detail below through the accompanying drawings and embodiments. Description of the Drawings

[0047] Figure 1 It is a method step diagram of a marine ranch early warning method based on a knowledge graph disclosed in an embodiment of the present invention. Detailed Implementation Modes

[0048] The technical solutions of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0049] The technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention and cannot be construed as limiting the protection scope of the present invention. Those skilled in the art can make some non-essential improvements and adjustments based on the content of the present invention described below. In the present invention, unless otherwise clearly specified and limited, the technical terms used in the present invention should have the ordinary meaning understood by those skilled in the art of the present invention.

[0050] Embodiment:

[0051] The present invention discloses an early warning method for marine ranching based on a knowledge graph. Refer to Figure 1 , including:

[0052] Step S100, determine the major management keywords for marine ranching management, and set several levels of minor management keywords under each major management keyword. Based on the progressive relationship between the major management keywords and the minor management keywords, construct a progressive connection line between the major management keywords and the minor management keywords to form a marine ranching knowledge graph. For the minor management keywords at the end of the marine ranching knowledge graph, several groups of preset collection parameter groups are set. The preset collection parameter groups include several types of collection parameter intervals, and a status description is configured for each collection parameter group.

[0053] This step realizes the semantic integration of management elements by constructing a hierarchical knowledge graph, and solves the problem of data islands in traditional monitoring systems. First, define major management keywords based on the core dimensions of marine ranching management (water quality, equipment, biological resources) to form the backbone of the management framework. For example, water quality management covers core indicators such as dissolved oxygen and pH value that affect the aquaculture environment. Each major category is further subdivided into minor management keywords to construct a progressive hierarchical relationship (such as "equipment management → aquaculture cage monitoring → structural integrity") to form a tree-shaped knowledge network. The minor nodes at the end of the knowledge graph are associated with preset collection parameter groups, and the parameter groups contain multi-dimensional threshold intervals (such as the normal interval of dissolved oxygen [5-7mg / L]) and corresponding status descriptions (such as "dissolved oxygen normal").

[0054] In some embodiments disclosed by the present invention, the major management keywords include: water quality management, equipment management, and biological resource management.

[0055] In some embodiments disclosed by the present invention, the minor management keywords include:

[0056] The subclass management keywords corresponding to water quality management include dissolved oxygen content, pH value, and water temperature.

[0057] The subclass management keywords corresponding to equipment management include aquaculture cage monitoring, feeding system monitoring, and water quality detection equipment calibration.

[0058] The subclass management keywords corresponding to biological resource management include fish farming techniques, shellfish harvesting plans, and seaweed farming area planning.

[0059] Among them, the subclass management keywords corresponding to aquaculture cage monitoring also include structural integrity and position stability; the subclass management keywords corresponding to feeding system monitoring also include daily maintenance and fault handling; the subclass management keywords corresponding to water quality monitoring equipment calibration also include accuracy and reliability; the subclass management keywords corresponding to fish farming techniques also include seed selection, hatching, feeding, and disease prevention; the subclass management keywords corresponding to shellfish harvesting plans also include harvesting timing, harvesting methods, and harvesting quantity; the subclass management keywords corresponding to seaweed farming area planning also include growth requirements, water flow conditions, and light intensity.

[0060] In some embodiments disclosed by the present invention, the method for determining each collection parameter interval in the preset collection parameter group includes:

[0061] Step S101: Analyze the historical operation records of the marine ranch, determine several groups of historical state parameter groups, and classify the historical state parameter groups into normal and abnormal categories to form a normal historical state parameter set and an abnormal historical state parameter set, and perform secondary classification on the normal historical state parameter set and the abnormal historical state parameter set respectively based on the state descriptions in the historical operation records of the marine ranch.

[0062] This step constructs a basic data set of parameter intervals through multi-level classification of historical data, solving the subjectivity problem of the traditional method relying on manual experience to set thresholds. First, clean and extract features from the historical operation records of the marine ranch (such as sensor logs, equipment status reports) to form a historical state parameter group containing multi-dimensional parameters. For example, a record at a certain moment may contain parameters such as dissolved oxygen, water temperature, and feeder current. Subsequently, based on preset labels (such as "normal", "hypoxia warning") or automatic clustering algorithms (such as K-means), the parameter group is initially classified to form a normal historical state parameter set and an abnormal historical state parameter set. For example, records where the dissolved oxygen is continuously below 3 mg / L and the fish activity is abnormal are classified into the abnormal set. Further, according to the state description (such as "equipment failure", "algal bloom"), the two data sets are classified again: the normal set is subdivided according to environmental scenarios (such as summer / winter baselines), and the abnormal set is divided according to failure types (such as mechanical failure / water quality pollution). For example, the normal water temperature intervals of [15-18°C] in winter and [20-25°C] in summer are classified into different subclasses to ensure the seasonal adaptability of the parameter intervals. The second classification is achieved through decision trees or semantic matching, laying a data foundation for subsequent refined analysis.

[0063] Step S102: For each position corresponding to a detection point on the marine ranch floor plan, construct a vertical acquisition parameter axis, map the historical acquisition parameters corresponding to each detection point in the historical state parameter group to the acquisition parameter number axis in the form of historical acquisition parameter mapping points, and connect the historical acquisition parameter mapping points pairwise to obtain a mapping point connection diagram.

[0064] This step realizes parameter distribution pattern recognition through spatial visualization, solving the limitations of single-point threshold analysis. For each detection point (such as the No. 1 water quality sensor), establish a vertical acquisition parameter axis at the corresponding position on the GIS floor plan, with the horizontal axis representing the time or event sequence and the vertical axis representing the parameter value (such as dissolved oxygen concentration). Map the historical parameter values to points on the axis (such as dissolved oxygen 6 mg / L corresponding to the vertical height 6), and connect adjacent points in chronological order to form a mapping point connection diagram. For example, the dissolved oxygen fluctuation of a certain detection point within 24 hours can be shown as a broken line, with the peak corresponding to the enhanced photosynthesis during the day and the valley corresponding to the oxygen consumption at night. Connecting the connection diagrams of multiple detection points pairwise (such as the dissolved oxygen curves of the No. 1 and No. 2 sensors) can reveal spatial correlations (such as abnormal synchronization of downstream parameters caused by upstream equipment failures). This visualization not only shows the dynamic changes of parameters but also helps to identify the types of abnormalities through graphical patterns (such as periodic fluctuations, sudden spikes).

[0065] Step S103: Compare the mapping point connection diagrams corresponding to the historical state parameter sets after the second classification with each other, and based on the comparison results, classify the normal historical state parameter set and the abnormal historical state parameter set for the third time to obtain a normal historical state parameter subset and an abnormal historical state parameter subset.

[0066] In some embodiments disclosed by the present invention, the method for comparing the mapping point connection diagrams corresponding to the historically sorted state parameter sets after secondary classification includes:

[0067] Step S1031: Set different key weights for different parameter mapping points, compare the mapping point height difference amounts of each parameter mapping point between the mapping point connection diagrams, and combine the key weights to construct a first equivalence operator.

[0068] Step S1032: Compare the coincidence scale of the mapping point connections of the mapping point connection diagrams, and based on the coincidence scale, correct the first equivalence operator to obtain the equivalence degree between the mapping point connection diagrams.

[0069] Among them, the expression for calculating the equivalence degree is:

[0070] 。

[0071] Among them, T is the equivalence degree, H is the coincidence scale, K i is the key weight corresponding to the i-th parameter mapping point, h max is the preset maximum height difference amount, Δh i is the height difference amount of the i-th parameter mapping point, and n is the number of all parameter mapping points.

[0072] In some embodiments disclosed by the present invention, the method for determining the coincidence scale of the mapping point connections of the mapping point connection diagrams includes:

[0073] Step S1033: Determine the corresponding mapping point connections in the mapping point connection diagrams, set a number of coincidence detection points for the mapping point connections at a preset interval, determine the coincidence detection distances between the corresponding coincidence detection points, and calculate the average value of the coincidence detection distances to obtain the average coincidence detection distance.

[0074] Step S1034: Set a coincidence detection distance judgment interval for the average coincidence detection distance, and a coincidence parameter is set for each coincidence detection distance judgment interval. Based on the coincidence detection distance judgment interval to which each mapping point connection belongs, determine the coincidence parameter corresponding to the mapping point connection.

[0075] Step S1035: Based on the coincidence parameter corresponding to each mapping point connection, determine the coincidence scale between the mapping point connection diagrams.

[0076] Among them, the expression for calculating the coincidence scale is:

[0077] 。

[0078] Among them, H is the coincidence scale, R is the adjustment coefficient of the influence of the coincidence parameter, c is the adjustment constant of the influence of the coincidence parameter, and s max is the preset maximum coincidence parameter, and s x is the coincidence parameter corresponding to the connection line of the x-th mapping point, and N is the number of connection lines of the mapping points.

[0079] Step S104: Analyze the acquisition parameter range of the acquisition parameter group in each historical state parameter subset to obtain the preset acquisition parameter group corresponding to the historical state parameter subset.

[0080] Step S200: Use the GIS map information system to determine the plan view of the marine ranch, determine the location of the equipment in the marine ranch and the location of the detection points of the parameter detection points, and configure them on the plan view of the marine ranch in the form of equipment mapping points and detection mapping points respectively.

[0081] This step uses GIS technology to establish a spatial association model between equipment and detection points to solve the problem of missing analysis of abnormal propagation paths. First, calibrate the physical positions of equipment (such as feeding machines and aerators) and detection points (such as water quality sensors) through geographical coordinates, and generate equipment mapping points and detection mapping points in the digital plan view of the marine ranch. For example, the position coordinates of the aquaculture cage are mapped to a polygon area in the GIS map, and the water quality sensors deployed around it are mapped to point elements. Further construct a spatial topology network: represent the spatial relationship between equipment and detection points through vector connections (such as the water flow influence path between the feeding machine and the downstream water quality detection point), and assign weights to the connections (such as distance attenuation coefficients). This mapping not only realizes the digital twin of the physical world but also provides spatial dimension support for anomaly analysis. For example, when a certain feeding machine fails, the GIS topology network can quickly locate the detection points within its influence range and preferentially compare relevant parameter anomalies (such as a sudden drop in bait concentration). The introduction of spatial weights enables the system to distinguish local anomalies from global risks (such as the mutation weight of parameters around the equipment being higher than the distal fluctuations), improving the accuracy of early warning.

[0082] Step S300: Collect the state performances of different equipment and the acquisition parameters corresponding to different detection points when the marine ranch is operating normally, record the combination of the state performances of the equipment and the acquisition parameters that appear simultaneously as the normal state parameter group, and map and associate the normal state parameter group with the corresponding plan view of the marine ranch to obtain the normal marine ranch plan view performance model.

[0083] In this step, a baseline model is constructed through multi-source data fusion to solve the problem of the lack of a dynamic benchmark in traditional methods. During the normal operation of the marine ranch, device status data (such as motor current, cage tension) and detection point parameters (such as dissolved oxygen, water temperature) are synchronously collected to form a normal state parameter group. For example, when the feeder is working properly, the current range is [1.8 - 2.2 A], and the dissolved oxygen at the associated detection point is maintained at [5 - 7 mg / L]. After these parameter groups are spatially and temporally aligned (such as timestamp synchronization, spatial position matching), they are mapped and associated with the GIS floor plan to generate a performance model of the normal marine ranch floor plan. This model is essentially a multi-dimensional tensor, including a device status matrix, an environmental parameter matrix, and a spatial relationship matrix. During the model construction process, a sliding window algorithm is used to dynamically update the parameter interval to adapt to environmental gradual changes (such as the baseline drift of water temperature caused by seasonal changes). For example, the normal interval of water temperature in winter may be adjusted from [20 - 25 °C] to [15 - 20 °C]. The normal model provides a dynamic benchmark for subsequent anomaly detection, avoiding false alarms caused by fixed thresholds.

[0084] Step S400: Based on the preset abnormal situations, adjust the normal state parameter group in the performance model of the normal marine ranch floor plan to obtain a performance model of the abnormal marine ranch floor plan, associate the status description corresponding to the abnormal situation with the performance model of the abnormal marine ranch floor plan, and determine the subclass keywords to which the performance model of the abnormal marine ranch floor plan belongs based on the abnormal state parameter group of the performance model of the abnormal marine ranch floor plan and the preset acquisition parameter group mapped in the marine ranch knowledge graph.

[0085] In some embodiments disclosed in the present invention, the method for adjusting the performance model of the normal marine ranch floor plan based on the preset abnormal situations includes:

[0086] Step S401: Analyze the historical operation records of the marine ranch to determine a number of abnormal historical state parameter groups, determine the corresponding status descriptions, traverse and compare the preset situations with the determined status descriptions to determine a number of matching status descriptions, and screen out the preset acquisition parameter groups corresponding to the status descriptions.

[0087] Step S402: Based on the several acquisition parameter intervals corresponding to the preset acquisition parameter groups screened out, determine the adjustment range for each abnormal historical state parameter group. After adjustment, a number of adjusted abnormal state parameter groups are obtained, and the normal state parameter group is adjusted with the adjusted abnormal state parameter group as the reference object.

[0088] Step S500, when conducting real-time supervision of the marine ranch, based on the real-time status performance of the devices and the real-time acquisition parameters collected in real time, construct a real-time marine ranch floor plan performance model, and compare the real-time marine ranch floor plan performance model with different abnormal marine ranch floor plan performance models. If there is a matching abnormal marine ranch floor plan performance model, an abnormal alarm is issued, and based on the minor category management keywords corresponding to the abnormal marine ranch floor plan performance model, it is marked and displayed on the marine ranch knowledge graph.

[0089] In this step, by matching the real-time data stream with the abnormal model, instant risk perception and traceability are realized. The device status (such as sensor readings) and detection parameters (such as pH value) are collected in real time to construct a real-time marine ranch floor plan performance model. It is compared with the pre-stored abnormal model library in multiple dimensions:

[0090] Parameter value comparison: Check whether the real-time parameters fall within the threshold range of the abnormal model;

[0091] Spatial pattern comparison: Verify the abnormal propagation path through the GIS topological network (such as whether a device failure causes a parameter mutation at an associated detection point);

[0092] Device status linkage analysis: Confirm the spatio-temporal consistency between device anomalies and parameter anomalies (such as whether the current zeroing and the bait concentration decrease occur simultaneously). If the real-time model matches an abnormal model (such as simultaneously meeting the numerical, spatial, and device status conditions), an alarm is triggered, and the abnormal path (such as "device management → feeding system → failure shutdown") is highlighted in the knowledge graph. The alarm information is synchronously pushed to the management terminal to guide targeted handling (such as giving priority to repairing the specified device).

[0093] In some embodiments disclosed in the present invention, the method for determining the matching situation between the real-time marine ranch floor plan performance model and the abnormal marine ranch floor plan performance model includes:

[0094] Step S501, compare the real-time marine ranch floor plan performance model with the abnormal marine ranch floor plan performance model, including the comparison of the status performance of the corresponding devices and the comparison of the acquisition parameters of the corresponding parameter detection points. If the performance differences between the status performances are all within the preset range and the parameter difference amounts between the acquisition parameters are all within the preset difference amount, it is determined that the real-time marine ranch floor plan performance model and the abnormal marine ranch floor plan performance model match.

[0095] In some embodiments disclosed in the present invention, a marine ranch early warning system based on a knowledge graph is also disclosed, including:

[0096] The first module is used to determine the major management keywords for ocean ranch management, and several levels of minor management keywords are set under each major management keyword. Based on the progressive relationship between the major management keywords and the minor management keywords, progressive connections between the major management keywords and the minor management keywords are constructed to form an ocean ranch knowledge graph. Several groups of preset acquisition parameter groups are set for the minor management keywords at the end of the ocean ranch knowledge graph. The preset acquisition parameter groups include several types of acquisition parameter intervals, and a status description is configured for each acquisition parameter group.

[0097] The second module is used to use the GIS map information system to determine the ocean ranch floor plan, determine the locations of the equipment in the ocean ranch and the locations of the detection points of the parameter detection points, and configure them on the ocean ranch floor plan in the form of equipment mapping points and detection mapping points respectively.

[0098] The third module is used to collect the status performances of different equipment and the acquisition parameters corresponding to different detection points when the ocean ranch is operating normally. The combination of the status performances of the equipment and the acquisition parameters that appear simultaneously is recorded as a normal status parameter group, and the normal status parameter group is mapped and associated with the corresponding ocean ranch floor plan to obtain a normal ocean ranch floor plan performance model.

[0099] The fourth module is used to adjust the normal status parameter group in the normal ocean ranch floor plan performance model based on preset abnormal situations to obtain an abnormal ocean ranch floor plan performance model, associate the status description corresponding to the abnormal situation with the abnormal ocean ranch floor plan performance model, and determine the minor keywords to which the abnormal ocean ranch floor plan performance model belongs based on the abnormal status parameter group of the abnormal ocean ranch floor plan performance model and the preset acquisition parameter group mapped in the ocean ranch knowledge graph.

[0100] The fifth module is used to construct a real-time ocean ranch floor plan performance model based on the real-time status performance of the equipment and the real-time acquisition parameters collected in real time during the real-time supervision of the ocean ranch, and compare the real-time ocean ranch floor plan performance model with different abnormal ocean ranch floor plan performance models. If there is a matching abnormal ocean ranch floor plan performance model, an abnormal alarm is issued, and a mark display is performed on the ocean ranch knowledge graph based on the minor management keywords corresponding to the abnormal ocean ranch floor plan performance model.

[0101] The present invention discloses an ocean ranch early warning method and system based on a knowledge graph, which relates to the technical field of ocean ranch management. It includes a knowledge graph of major keywords such as water quality management, equipment management, and biological resource management, as well as minor keywords at progressive levels. The terminal minor categories are associated with preset acquisition parameter groups and status descriptions; an ocean ranch floor plan is generated; the equipment status and detection parameters are collected to form a normal status parameter group, and a normal floor plan performance model is constructed; an abnormal model is generated by adjusting the normal parameter group based on a preset abnormality, and is dynamically associated with the minor keywords in the knowledge graph; real-time data is collected to construct a real-time model, which is compared with the abnormal model to trigger an alarm, and the root cause of the abnormality is marked in the knowledge graph. This method breaks through the limitations of traditional single-dimensional monitoring, semantically integrates multi-source data through the knowledge graph, and combines dynamic thresholds and spatial association analysis to achieve precise abnormal positioning and cross-dimensional risk early warning, significantly improving the management efficiency and safety of ocean ranches.

[0102] Through the description of the above embodiments, those skilled in the art can clearly understand that the present invention can be implemented through hardware or by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), including several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in various implementation scenarios of the present invention.

[0103] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that they can still modify or equivalently replace the technical solutions of the present invention, and these modifications or equivalent replacements cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A marine ranch early warning method based on knowledge graph, characterized in that: include: Step S100, determining the major management keywords of marine ranch management, and setting several levels of sub-category management keywords under each major management keyword, constructing a progressive connection between the major management keywords and the sub-category management keywords based on the progressive relationship between the major management keywords and the sub-category management keywords, forming a marine ranch knowledge graph, setting several groups of preset acquisition parameter groups for the sub-category management keywords at the end of the marine ranch knowledge graph, the preset acquisition parameter groups including several types of acquisition parameter intervals, and configuring a state description for each acquisition parameter group; Step S200, using the GIS map information system to determine the marine ranch plan, determine the equipment locations in the marine ranch and the detection point locations of the parameter detection points, and configure them on the marine ranch plan in the form of equipment mapping points and detection mapping points respectively; Step S300, when the ocean ranch is in normal operation, the status performance of different devices and the collection parameters corresponding to different detection points are collected, and the combination of the status performance of the devices and the collection parameters that appear simultaneously is recorded as a normal state parameter group, and the normal state parameter group is mapped and associated with the corresponding ocean ranch plan to obtain a normal ocean ranch plan performance model; Step S400, based on the preset abnormal situation, the normal state parameter group in the normal ocean ranch plan representation model is adjusted to obtain the abnormal ocean ranch plan representation model, and the state description corresponding to the abnormal situation is associated with the abnormal ocean ranch plan representation model, and based on the abnormal state parameter group of the abnormal ocean ranch plan representation model, relative to the preset acquisition parameter group mapped in the ocean ranch knowledge graph, the subcategory keyword to which the abnormal ocean ranch plan representation model belongs is determined; Step S500, when real-time supervision of the ocean ranch is carried out, a real-time ocean ranch floor plan representation model is constructed based on the real-time status performance of the equipment and the real-time collection parameters, and the real-time ocean ranch floor plan representation model is compared with different abnormal ocean ranch floor plan representation models. If there is a matching abnormal ocean ranch floor plan representation model, an abnormal alarm is issued, and based on the subcategory management keywords corresponding to the abnormal ocean ranch floor plan representation model, it is marked and displayed on the ocean ranch knowledge graph.

2. The marine ranch early warning method based on knowledge graph according to claim 1 is characterized in that: Major management keywords include: water quality management, equipment management, and biological resource management.

3. The marine ranch early warning method based on knowledge graph according to claim 2 is characterized in that: Subcategory management keywords include: The sub-category management keywords corresponding to water quality management include dissolved oxygen content, pH value, and water temperature; The sub-category management keywords corresponding to equipment management include aquaculture cage monitoring, feeding system monitoring, and water quality testing equipment calibration; The sub-category management keywords corresponding to biological resource management include, fish farming technology, shellfish harvesting plan, and seaweed culture area planning; Among them, the sub-category management keywords corresponding to the monitoring of aquaculture nets also include structural integrity and position stability; the sub-category management keywords corresponding to the monitoring of feeding systems also include daily maintenance and troubleshooting; the sub-category management keywords corresponding to the calibration of water quality monitoring equipment also include accuracy and reliability; the sub-category management keywords corresponding to fish farming technology also include seed selection, hatching, feeding and disease prevention and control; the sub-category management keywords corresponding to shellfish harvesting plans also include harvesting timing, harvesting methods and harvesting quantity; the sub-category management keywords corresponding to seaweed farming area planning also include growth requirements, water flow conditions and light intensity.

4. The marine ranch early warning method based on knowledge graph according to claim 1 is characterized in that: The method for determining each acquisition parameter interval in the preset acquisition parameter group includes: Step S101, analyzing the historical operation records of the ocean ranch, determining several groups of historical status parameter groups, and classifying the historical status parameter groups into normal and abnormal categories to form normal historical status parameter sets and abnormal historical status parameter sets, and based on the status description in the historical operation records of the ocean ranch, reclassifying the normal historical status parameter sets and the abnormal historical status parameter sets respectively; Step S102, constructing a vertical acquisition parameter axis for the position corresponding to each detection point in the ocean ranch plan, mapping the historical acquisition parameters corresponding to each detection point in the historical state parameter group to the acquisition parameter axis in the form of historical acquisition parameter mapping points, and connecting the historical acquisition parameter mapping points in pairs to obtain a mapping point connection diagram; Step S103, comparing the mapping point connection diagrams corresponding to the historical state parameter sets after the secondary classification for mutual equality, and based on the comparison results, classifying the normal historical state parameter sets and the abnormal historical state parameter sets three times to obtain the normal historical state parameter subsets and the abnormal historical state parameter subsets; Step S104: performing collection parameter range analysis on the collection parameter groups in each historical state parameter subset to obtain a preset collection parameter group corresponding to the historical state parameter subset.

5. The marine ranch early warning method based on knowledge graph according to claim 4 is characterized in that: The method for comparing the mapping point connection graphs corresponding to the historical state parameter sets after secondary classification for mutual equivalence includes: Step S1031, setting different key weights for different parameter mapping points, comparing the mapping point height difference of each parameter mapping point between the mapping point connection graphs, and constructing a first equivalence operator in combination with the key weights; Step S1032, comparing the coincidence scale of the mapping point connection lines of the mapping point connection diagram, and based on the coincidence scale, modifying the first equivalence operator to obtain the degree of equivalence between the mapping point connection diagrams; The expression for calculating the degree of equality is: Among them, T is the degree of equality, H is the overlap scale, and K i is the key weight corresponding to the i-th parameter mapping point, h max is the preset maximum height difference, Δh i is the height difference of the i-th parameter mapping point, and n is the number of all parameter mapping points.

6. The marine ranch early warning method based on knowledge graph according to claim 5 is characterized in that: The method for determining the coincidence scale of the mapping point connection lines of the mapping point connection graph includes: Step S1033, determining the corresponding mapping point lines in the mapping point line diagram, and setting a number of coincidence detection points for the mapping point lines according to a preset interval, determining the coincidence detection distances between the corresponding coincidence detection points, and calculating the average of the coincidence detection distances to obtain an average coincidence detection distance; Step S1034, setting an overlap detection distance judgment interval for the average overlap detection distance, setting an overlap parameter corresponding to each overlap detection distance judgment interval, and determining the overlap parameter corresponding to the mapping point connection line based on the overlap detection distance judgment interval to which each mapping point connection line belongs; Step S1035, determining the overlap scale between the mapping point connection diagrams based on the overlap parameter corresponding to each mapping point connection line; Among them, the expression for calculating the coincidence scale is: Among them, H is the overlap scale, R is the overlap parameter influence adjustment coefficient, c is the overlap parameter influence adjustment constant, s max is the preset maximum coincidence parameter, s x is the coincidence parameter corresponding to the xth mapping point connection line, and N is the number of mapping point connections.

7. The marine ranch early warning method based on knowledge graph according to claim 1 is characterized in that: Based on the preset abnormal conditions, the methods for adjusting the normal marine ranch floor plan representation model include: Step S401, analyzing the historical operation records of the ocean ranch, determining a number of abnormal historical state parameter groups, and determining corresponding state descriptions, traversing and comparing the preset preset conditions and the determined state descriptions, determining a number of matching state descriptions, and filtering out the preset acquisition parameter groups corresponding to the state descriptions; Step S402, based on the several acquisition parameter intervals corresponding to the screened preset acquisition parameter groups, determine the adjustment range of each abnormal historical state parameter group, obtain several adjusted abnormal state parameter groups after adjustment, and adjust the normal state parameter group with the adjusted abnormal state parameter groups as reference objects.

8. The marine ranch early warning method based on knowledge graph according to claim 1 is characterized in that: Methods for determining the consistency between the real-time ocean ranch plan representation model and the abnormal ocean ranch plan representation model include: Step S501, compare the real-time ocean ranch plan view representation model and the abnormal ocean ranch plan view representation model, including the comparison of the status representation of the corresponding equipment and the comparison of the collection parameters of the corresponding parameter detection points. If the performance differences between the status representations are within the preset range and the parameter differences between the collection parameters are within the preset difference amounts, then it is determined that the real-time ocean ranch plan view representation model and the abnormal ocean ranch plan view representation model are consistent.

9. A marine ranch early warning system based on knowledge graph, characterized in that: The marine ranch early warning method for executing any one of claims 1 to 8 comprises: The first module is used to determine the major management keywords of marine ranch management, and set several levels of sub-category management keywords under each major management keyword. Based on the progressive relationship between the major management keywords and the sub-category management keywords, a progressive connection between the major management keywords and the sub-category management keywords is constructed to form a marine ranch knowledge graph. For the sub-category management keywords at the end of the marine ranch knowledge graph, several groups of preset collection parameter groups are set. The preset collection parameter groups include several types of collection parameter intervals, and a state description is configured for each collection parameter group; The second module is used to determine the marine ranch plan using the GIS map information system, determine the equipment locations in the marine ranch and the detection points of the parameter detection points, and configure them on the marine ranch plan in the form of equipment mapping points and detection mapping points respectively; The third module is used to collect the status performance of different equipment and the collection parameters corresponding to different detection points when the ocean ranch is in normal operation, record the combination of the status performance of the equipment and the collection parameters that appear at the same time as a normal state parameter group, and map the normal state parameter group with the corresponding ocean ranch plan to obtain a normal ocean ranch plan performance model; The fourth module is used to adjust the normal state parameter group in the normal ocean ranch plan representation model based on the preset abnormal situation to obtain the abnormal ocean ranch plan representation model, and associate the state description corresponding to the abnormal situation with the abnormal ocean ranch plan representation model, and determine the subcategory keyword to which the abnormal ocean ranch plan representation model belongs based on the abnormal state parameter group of the abnormal ocean ranch plan representation model relative to the preset acquisition parameter group mapped in the ocean ranch knowledge graph; The fifth module is used to build a real-time ocean ranch floor plan representation model based on the real-time status performance of the equipment and the real-time collection parameters when conducting real-time supervision on the ocean ranch, and compare the real-time ocean ranch floor plan representation model with different abnormal ocean ranch floor plan representation models. If there is a matching abnormal ocean ranch floor plan representation model, an abnormal alarm will be issued, and based on the subcategory management keywords corresponding to the abnormal ocean ranch floor plan representation model, it will be marked and displayed on the ocean ranch knowledge graph.

Images