Offshore wind plant abnormal ship behavior identification method and system

Abstract

The invention discloses a method for identifying abnormal ship behaviors of an offshore wind plant. The method comprises the following steps: S1, acquiring a windmill interpretation mark library of historical offshore windmill remote sensing images; performing marine target marking on the pre-collected target area remote sensing image; identifying and acquiring an offshore windmill in the initial sample data set based on ship satellite AIS data of a target area remote sensing image according to an interpretation mark of a windmill interpretation mark library, and acquiring an optimized data set; s2, acquiring an offshore windmill detection model used for identifying and extracting position information of an offshore windmill, and converting the position information of the offshore windmill into a wind scene vector in a vector layer form; s3, converting the motion trail of the ship satellite AIS data into a trajectory vector based on a preset MMSI number; s4, taking the wind scene vector as a target area, and obtaining a relative position relation between two end points of the trajectory vector and the target area to obtain a topological relation graph; s5, a topological relation detection model used for recognizing the topological relation between the offshore windmill and the moving ship is obtained; and S6, according to the topological relation detection model, identifying abnormal ship behaviors of the offshore wind plant to obtain ship warning data so as to realize safety monitoring of the offshore wind plant. The problem that an existing offshore wind plant risk assessment method mainly depends on a prediction model and expert experience and cannot accurately reflect complex conditions in actual operation is solved.

Description

Technical Field

[0001] The present invention relates to the field of smart maritime technology, and in particular to a method and system for identifying abnormal ship behavior in an offshore wind farm. Background Art

[0002] In the context of global energy structure transformation and sustainable development, offshore wind farms, as an important part of clean energy, are of great significance in promoting energy transformation through their construction and operation. However, the existence of offshore wind farms has brought new challenges to the navigation safety of ships in the surrounding waters. The construction and operation of offshore wind farms require reasonable planning and management of ship routes to avoid potential navigation risks and conflicts.

[0003] Navigation risk assessments for offshore wind farms are usually conducted before the construction of wind farms. These assessments are mainly based on predictions and a priori analysis to identify risks and develop corresponding preventive measures. However, with the actual operation of wind farms, actual navigation behaviors and patterns differ from predictions, which requires a posteriori analysis, that is, to evaluate the actual impact of wind farms on ship routes based on data collected after the actual operation of wind farms, and to verify or revise previous risk assessments accordingly.

[0004] Existing offshore wind farm risk assessment methods mainly rely on prediction models and expert experience, which cannot accurately reflect the complex situations in actual operations. In addition, existing methods also have shortcomings in real-time monitoring, dynamic risk assessment and emergency response. For example, the integrated application of remote sensing images and AIS data is not common in existing technologies, resulting in the identification and analysis of abnormal ship behaviors not being timely and accurate enough. Moreover, existing technologies have limited capabilities in processing large amounts of real-time data and providing decision support, making it difficult to meet the needs of offshore wind farm safety management. Therefore, in order to solve the above-mentioned technical problems, it is urgent to construct a method and system for identifying abnormal ship behaviors in offshore wind farms. Summary of the invention

[0005] The present invention provides a method and system for identifying abnormal ship behavior in an offshore wind farm to overcome the above technical problems.

[0006] In order to achieve the above object, the technical solution of the present invention is:

[0007] A method for identifying abnormal ship behavior in an offshore wind farm comprises the following steps:

[0008] S1: Obtain the windmill interpretation symbol library of historical offshore windmill remote sensing images;

[0009] Mark the marine targets on the pre-collected remote sensing images of the target area to obtain the initial sample data set;

[0010] The offshore targets include offshore windmill targets and mobile ship targets;

[0011] Based on the ship satellite AIS data of the target area remote sensing image, and according to the interpretation marks of the windmill interpretation mark library, the offshore windmills in the initial sample data set are identified and obtained to obtain the optimized data set;

[0012] S2: training a pre-established Oriented RCNN network model according to the optimized data set to obtain an offshore wind turbine detection model for identifying and extracting offshore wind turbine location information, and converting the offshore wind turbine location information into a wind field vector in the form of a vector layer;

[0013] S3: Obtain ship satellite AIS data through the ship automatic identification system;

[0014] And the ship satellite AIS data includes ship position and motion trajectory data;

[0015] Based on the preset MMSI number, the motion trajectory of the ship satellite AIS data is converted into a trajectory vector;

[0016] S4: Taking the wind field vector as the target area, obtaining the relative position relationship between the two end points of the trajectory vector and the target area, so as to obtain a topological relationship diagram between the offshore wind turbine and the moving ship;

[0017] S5: training a pre-established CNN-LSTM network model according to the topological relationship graph to obtain a topological relationship detection model for identifying the topological relationship between the offshore wind turbine and the moving ship;

[0018] S6: Based on the topological relationship detection model, abnormal ship behavior in offshore wind farms can be identified to obtain ship warning data;

[0019] The ship warning data includes ship type, frequent time period and high-risk sea area;

[0020] By setting up offshore monitoring and early warning equipment based on ship warning data, safe monitoring of offshore wind farms can be achieved.

[0021] Furthermore, the S1 specifically includes the following steps:

[0022] S11: Obtain historical offshore windmill remote sensing images, and call corresponding ship satellite AIS data from the existing database according to the shooting time and shooting range of the remote sensing images;

[0023] S12: acquiring remote sensing frame images frame by frame according to the remote sensing images of the offshore wind turbines, and acquiring characteristic data of the offshore wind turbines in the remote sensing frame images;

[0024] The characteristic data of the offshore windmill includes geometric profile characteristics, spatial characteristics and spectral characteristics;

[0025] And define and obtain windmill interpretation signs according to geometric contour features, spatial features and spectral features, and then obtain a windmill interpretation sign library;

[0026] S13: marking marine targets on the pre-collected remote sensing images of the target area to obtain an initial sample data set; and the marine targets include offshore windmill targets and mobile ship targets;

[0027] S14: Using the satellite AIS data of ships in the same period as the remote sensing image of the corresponding target area, the moving ship targets are removed from the initial sample data set;

[0028] And according to the interpretation marks of the windmill interpretation mark library, the offshore windmill targets in the initial sample data set are identified and acquired, and then an optimized data set including only the offshore windmill targets is obtained.

[0029] Further, in S2, the pre-established Oriented RCNN network model is trained according to the optimized data set, and the training process is as follows:

[0030] S21: randomly dividing the optimized data set into a training set and a test set according to a preset ratio;

[0031] S22: Input the training set into a pre-created Oriented RCNN network model for model training;

[0032] And based on the constructed model loss function L, confirm whether the output of the trained HRU-Net model converges;

[0033] If the output of the trained OrientedRCNN network model converges, execute step S23;

[0034] S23: Evaluate the model accuracy of the trained OrientedRCNN network model based on the test set;

[0035] The indicators for evaluating the accuracy of the model are the recall rate and accuracy of the model;

[0036] If it is confirmed that the model accuracy evaluation meets the preset model accuracy threshold, then the trained OrientedRCNN network model is confirmed to be an offshore wind turbine detection model for identifying and extracting offshore wind turbine location information;

[0037] If it is confirmed that the model accuracy evaluation does not meet the preset model accuracy threshold, the weight parameters of the OrientedRCNN network model are adaptively adjusted based on the back propagation method, and step S22 is repeated.

[0038] Furthermore, the expression of the model loss function L constructed in S22 is

[0039] L=L cls +λ reg L reg

[0040] Where: L cls represents the classification loss; L reg Represents the loss of bounding box regression and angle regression; λ reg Represents the weight coefficient used to balance the regression loss.

[0041] Furthermore, the method for obtaining the topological relationship diagram of the offshore wind turbine and the mobile ship in S4 is as follows:

[0042] Taking the wind field vector as the target area, the relative position relationship between the two end points of the trajectory vector and the target area is obtained;

[0043] The relative position relationship includes an intersection relationship, a containment relationship, an adjacency relationship, a non-intersection relationship and a crossing relationship;

[0044] The intersection relationship is that the ship's trajectory line passes through the surface area of ​​the windmill, and the starting point and the end point of the trajectory line are both the same boundary of the surface area;

[0045] The inclusion relationship is that the starting point or the end point of the ship's trajectory is located in the windmill area;

[0046] The adjacency relationship is that the ship's trajectory touches the boundary of the target area but does not cross it;

[0047] The non-intersecting relationship means that the ship's trajectory line has no intersection with the target area;

[0048] The crossing relationship is that the ship trajectory crosses from one boundary of the target area to another boundary;

[0049] And obtain the topological relationship diagram of offshore wind turbines and mobile ships based on the relative position relationship.

[0050] An abnormal ship behavior identification system for an offshore wind farm comprises at least

[0051] Remote sensing image real-time acquisition module, remote sensing image processing module, offshore windmill detection module, vector construction module, topological relationship acquisition module, topological relationship detection module and emergency response module;

[0052] The remote sensing image real-time acquisition module is used to retrieve historical offshore windmill remote sensing images;

[0053] Or collect and store remote sensing images of the target area in real time;

[0054] The remote sensing image processing module is used to obtain a windmill interpretation symbol library of historical offshore windmill remote sensing images;

[0055] Or the pre-collected remote sensing images of the target area are annotated with offshore targets to obtain an initial sample data set, and based on the ship satellite AIS data of the remote sensing images of the target area, combined with the interpretation marks of the windmill interpretation mark library, the offshore windmills in the initial sample data set are identified and obtained to obtain an optimized data set;

[0056] The offshore windmill detection module is used to establish an OrientedRCNN network model.

[0057] And train the OrientedRCNN network model based on the optimized data set to obtain an offshore wind turbine detection model for identifying and extracting offshore wind turbine location information;

[0058] The vector construction module is used to convert the offshore windmill position information detected and identified by the offshore windmill detection model into a wind field vector in the form of a vector layer;

[0059] Or the movement trajectory of the ship satellite AIS data obtained through the ship automatic identification system can be converted into a trajectory vector based on the preset MMSI number;

[0060] The topological relationship acquisition module is used to obtain a topological relationship diagram of the offshore wind turbine and the mobile ship according to the wind field vector and the trajectory line vector;

[0061] The topological relationship detection module is used to establish a CNN-LSTM network model.

[0062] And train the CNN-LSTM network model according to the topological relationship graph to obtain a topological relationship detection model for identifying the topological relationship between offshore wind turbines and mobile ships;

[0063] The emergency response module is used to identify abnormal ship behaviors in offshore wind farms based on the topological relationship detection model and obtain ship warning data;

[0064] By setting up offshore monitoring and early warning equipment based on ship warning data, safe monitoring of offshore wind farms can be achieved.

[0065] Beneficial effects: The present invention provides a method and system for identifying abnormal ship behavior in an offshore wind farm, which obtains a windmill interpretation mark library of historical offshore windmill remote sensing images; identifies and obtains offshore windmills in an initial sample data set based on ship satellite AIS data of remote sensing images of target areas, combined with interpretation marks of the windmill interpretation mark library, so as to judge ships through ship satellite AIS data, thereby filtering out ships mistakenly identified as offshore windmills, so as to improve the accuracy of target extraction of offshore windmills; trains the Oriented RCNN network model, and realizes real-time monitoring and intelligent analysis of abnormal ship behaviors in the offshore wind farm area through an automated iterative process, so as to obtain an offshore windmill detection model, thereby improving the recognition and extraction of offshore windmill position information; obtains a topological relationship diagram of offshore windmills and mobile ships according to the relative position relationship between wind field vectors and trajectory line vectors, and realizes the recognition of the topological relationship between offshore windmills and mobile ships according to the topological relationship detection model, thereby greatly improving the efficiency of identifying abnormal ship behaviors in offshore wind farms, and accurately reflects the judgment of complex situations in actual operations, and then obtains ship warning data through an emergency response module, thereby realizing safety monitoring of offshore wind farms. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0067] Figure 1 This is a flow chart of the method for identifying abnormal ship behavior in an offshore wind farm according to the present invention;

[0068] Figure 2 This is a flow chart of the abnormal ship behavior identification system for offshore wind farms of the present invention;

[0069] Figure 3 Schematic diagram of the topological relationship between the offshore wind turbine and the mobile ship in this embodiment. DETAILED DESCRIPTION

[0070] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0071] This embodiment provides a method for identifying abnormal ship behavior in an offshore wind farm. Figure 1 As shown, the specific steps include:

[0072] S1: Obtain the windmill interpretation symbol library of historical offshore windmill remote sensing images;

[0073] Mark the marine targets on the pre-collected remote sensing images of the target area to obtain the initial sample data set;

[0074] The offshore targets include offshore windmill targets and mobile ship targets;

[0075] Based on the ship satellite AIS data of the target area remote sensing image, and according to the interpretation marks of the windmill interpretation mark library, the offshore windmills in the initial sample data set are identified and obtained to obtain the optimized data set;

[0076] The specific steps include:

[0077] S11: Obtain historical offshore windmill remote sensing images, and call corresponding ship satellite AIS data from the existing database according to the shooting time and shooting range of the remote sensing images;

[0078] S12: acquiring remote sensing frame images frame by frame according to the remote sensing images of the offshore wind turbines, and acquiring characteristic data of the offshore wind turbines in the remote sensing frame images;

[0079] The characteristic data of the offshore windmill includes geometric profile characteristics, spatial characteristics and spectral characteristics;

[0080] Specifically, the geometric characteristics of offshore windmills, that is, offshore windmills usually have a clear outline, including a round base and three blades; spatial characteristics, that is, offshore windmills usually have a regular spatial layout, most of which are arranged in an array at sea, and their spacing is relatively fixed; spectral characteristics, the materials of windmills are mainly metal and composite materials, which have spectral reflection characteristics different from ships, seawater, etc.;

[0081] And define and obtain windmill interpretation signs according to geometric contour features, spatial features and spectral features, and then obtain a windmill interpretation sign library;

[0082] In this embodiment, the offshore windmills seen on the Internet are generally photographed by reporters and others, and can be understood as the main view or side view. Remote sensing images can be understood as overhead views of offshore windmills taken by satellites from space; interpretation marks are professional terms in the remote sensing discipline, and an interpretation mark is generally a row. The first column of this row is the name of the feature (here, the offshore windmill), the second column is a photo of the offshore windmill taken in real life, and the third column is a screenshot of the offshore windmill in the remote sensing image. In this embodiment, the windmill interpretation mark is defined as [A, B, C], that is, A represents the geometric outline features of the offshore windmill; B represents the spatial features of the offshore windmill; C represents the spectral features of the offshore windmill; this interpretation mark is used to allow other personnel to better identify offshore windmills from remote sensing images;

[0083] S13: marking marine targets on the pre-collected remote sensing images of the target area to obtain an initial sample data set; and the marine targets include offshore windmill targets and mobile ship targets;

[0084] The marking tool for marking the marine target is a known technical means, such as QGIS software, etc., which will not be described in detail here;

[0085] S14: Using the satellite AIS data of ships in the same period as the remote sensing image of the corresponding target area, the moving ship targets are removed from the initial sample data set;

[0086] And according to the interpretation marks of the windmill interpretation mark library, the offshore windmill targets in the initial sample data set are identified and acquired, and then an optimized data set including only the offshore windmill targets is obtained.

[0087] In this embodiment, the ship position filtering is to obtain the position of the ship at sea through AIS data, compare the ship position with the target position in the image, and filter out the ship target confused with the windmill. That is, if a certain offshore windmill target has no corresponding AIS ship information and meets the windmill interpretation mark, it is confirmed to be an offshore windmill;

[0088] S2: training a pre-established OrientedRCNN network model according to the optimized data set to obtain an offshore wind turbine detection model for identifying and extracting offshore wind turbine location information, and converting the offshore wind turbine location information into a wind field vector in the form of a vector layer;

[0089] Specifically, the pre-established OrientedRCNN network model is trained according to the optimized dataset, and the training process is as follows:

[0090] S21: randomly dividing the optimized data set into a training set and a test set according to a preset ratio; wherein the preset ratio is preferably 7:3;

[0091] S22: Input the training set into the pre-created OrientedRCNN network model for model training;

[0092] Based on the constructed model loss function L, confirm whether the output of the trained HRU-Net model converges; the expression of the constructed model loss function L is

[0093] L=L cls +λ reg L reg

[0094] Where: L cls represents the classification loss; L reg represents the loss of bounding box regression and angle regression; reg Represents the weight coefficient used to balance the regression loss;

[0095] If the output of the trained OrientedRCNN network model converges, execute step S23;

[0096] S23: Evaluate the model accuracy of the trained OrientedRCNN network model based on the test set;

[0097] The indicators for evaluating the accuracy of the model are the recall rate and accuracy of the model;

[0098] If it is confirmed that the model accuracy evaluation meets the preset model accuracy threshold, then the trained OrientedRCNN network model is confirmed to be an offshore wind turbine detection model for identifying and extracting offshore wind turbine location information;

[0099] Specifically, the model is evaluated for accuracy using the test set, and if the recall rate is greater than 90% and the accuracy rate is greater than 90%, the model is used to process the remote sensing image to extract offshore windmills;

[0100] If it is confirmed that the model accuracy evaluation does not meet the preset model accuracy threshold, the weight parameters of the OrientedRCNN network model are adaptively adjusted based on the back propagation method, and step S22 is repeated;

[0101] S3: Obtain ship satellite AIS data through the ship automatic identification system;

[0102] And the ship satellite AIS data includes ship position and motion trajectory data;

[0103] Based on the preset MMSI number, the motion trajectory of the ship satellite AIS data is converted into a trajectory vector; wherein the method for obtaining the trajectory vector is a known technical means, which will not be described in detail here;

[0104] S4: Taking the wind field vector as the target area, the relative position relationship between the two end points of the trajectory vector and the target area is obtained to obtain the topological relationship diagram between the offshore wind turbine and the moving ship to support subsequent spatial analysis and visualization operations;

[0105] Specifically, the method for obtaining the topological relationship diagram of offshore wind turbines and mobile ships is as follows:

[0106] Taking the wind field vector as the target area, the relative position relationship between the two end points of the trajectory vector and the target area is obtained;

[0107] The relative position relationship includes an intersection relationship, a containment relationship, an adjacency relationship, a non-intersection relationship and a crossing relationship;

[0108] The intersection relationship is that the ship's trajectory line passes through the surface area of ​​the windmill, and the starting point and the end point of the trajectory line are both the same boundary of the surface area;

[0109] The inclusion relationship is that the starting point or the end point of the ship's trajectory is located in the windmill area;

[0110] The adjacency relationship is that the ship's trajectory touches the boundary of the target area but does not cross it;

[0111] The non-intersecting relationship means that the ship's trajectory line has no intersection with the target area;

[0112] The crossing relationship is that the ship trajectory crosses from one boundary of the target area to another boundary;

[0113] And obtain the topological relationship diagram of offshore wind turbines and mobile ships based on the relative position relationship.

[0114] In this implementation, by defining the topological relationship type, such as Figure 3 As shown in , the intersection relationship means that when the ship's trajectory passes through the windmill's surface area and is on the same boundary of the surface area, the two are considered to intersect, such as Figure 3 (a), this relationship indicates that the ship is too close to the windmill and there is a risk of collision; the containment relationship refers to a special case, such as Figure 3 (b) The starting point or end point of the track is located in the windmill area; the adjacency relationship is when the ship track line touches the windmill surface boundary but does not cross it, it is called adjacency, such as Figure 3 (c) This relationship indicates that the ship is approaching the safety boundary of the windmill and needs attention; the non-intersecting relationship means that the ship's trajectory line and the windmill surface have no intersection, that is, the two are completely separated in space, such as Figure 3 (d) This is the normal relationship, but when analyzing potential risks, we also need to pay attention to those situations where non-intersection suddenly changes to intersection; the crossing relationship is similar to intersection, but it emphasizes the process of the trajectory line crossing from one boundary to another in the windmill area, such as Figure 3 (e) This relationship can better describe the movement of ships in the windmill area than simple intersection, and ensure that the vector data format is compatible to facilitate topological analysis;

[0115] S5: training a pre-established CNN-LSTM network model according to the topological relationship graph to obtain a topological relationship detection model for identifying the topological relationship between the offshore wind turbine and the moving ship;

[0116] Specifically, the pre-established CNN-LSTM network model uses a convolutional neural network and a long short-time neural network, and the output of the convolutional neural network is used as the input of the long short-time neural network. Both the CNN network and the LSTM use the existing standard structure. The evaluation index in the model training process uses the cross entropy loss function, and other common classification problem loss functions can also be used, which will not be described in detail here.

[0117] S6: Based on the topological relationship detection model, abnormal ship behavior in offshore wind farms can be identified to obtain ship warning data;

[0118] The ship warning data includes ship type, frequent time period and high-risk sea area;

[0119] By setting up offshore monitoring and early warning equipment based on ship warning data, safe monitoring of offshore wind farms can be achieved.

[0120] Specifically, according to the abnormal ship behaviors classified and recorded in detail in step S6, their impact on the safety of offshore wind farms is evaluated, that is, by analyzing the topological relationship between the ship trajectory and the wind farm range, it can be found in this embodiment that the ship repeatedly crosses the 500-meter safety area around the pre-defined wind turbines. This behavior violates the relevant safety regulations and may seriously interfere with the normal operation of the wind farm, and pose a safety hazard to wind power facilities (such as wind turbines, cables, etc.), increasing the risk of collision and damage to the wind farm, which may affect the stable power supply and safe production of the wind farm. Therefore, this crossing behavior needs to be strictly supervised and intervened in time to ensure the safe and stable operation of the wind farm. Then, a warning range covering high-risk areas is defined (for example, if ships often enter from one of the boundaries of the wind farm field vector, the 500 meters around this boundary is set as the warning range); ships entering the warning range are monitored in real time, especially the types of ships whose topological relationships are identified as intersecting, containing and crossing, to ensure timely detection of potential risks; for illegal acts found in the monitoring, corresponding measures are taken to deal with them (corresponding measures: including the use of high-tech monitoring means to achieve all-weather monitoring, the establishment of an early warning mechanism for early intervention, the strengthening of publicity and education to enhance the safety awareness of crew members, the implementation of strict penalties and sanctions to curb violations, and the optimization of wind farm layout and identification to reduce the risk of crossing. Processing: First, contact the ship through the monitoring center and ask it to stop immediately and adjust the route. If the ship does not cooperate, law enforcement vessels will be dispatched to intercept and persuade it to leave on the spot. At the same time, the evidence of violations is recorded in detail and reported to the maritime department for subsequent legal processing. In addition, we will also follow up on the illegal ships to ensure that they do not commit similar acts again, and strengthen the inspection of the surrounding waters to maintain the safety and order of the wind farm.) to ensure the navigation safety and management efficiency of the offshore wind farm.

[0121] An abnormal ship behavior identification system for offshore wind farms, such as Figure 2 As shown, at least

[0122] Remote sensing image real-time acquisition module, remote sensing image processing module, offshore windmill detection module, vector construction module, topological relationship acquisition module, topological relationship detection module and emergency response module;

[0123] The remote sensing image real-time acquisition module is used to retrieve historical offshore windmill remote sensing images;

[0124] Or collect and store remote sensing images of the target area in real time;

[0125] The remote sensing image processing module is used to obtain a windmill interpretation symbol library of historical offshore windmill remote sensing images;

[0126] Or the pre-collected remote sensing images of the target area are annotated with offshore targets to obtain an initial sample data set, and based on the ship satellite AIS data of the remote sensing images of the target area, combined with the interpretation marks of the windmill interpretation mark library, the offshore windmills in the initial sample data set are identified and obtained to obtain an optimized data set;

[0127] The offshore windmill detection module is used for the pre-established Oriented RCNN network model.

[0128] And train the Oriented RCNN network model based on the optimized data set to obtain an offshore wind turbine detection model for identifying and extracting offshore wind turbine location information;

[0129] The vector construction module is used to convert the offshore windmill position information detected and identified by the offshore windmill detection model into a wind field vector in the form of a vector layer;

[0130] Or the movement trajectory of the ship satellite AIS data obtained through the ship automatic identification system can be converted into a trajectory vector based on the preset MMSI number;

[0131] The topological relationship acquisition module is used to obtain a topological relationship diagram of the offshore wind turbine and the mobile ship according to the wind field vector and the trajectory line vector;

[0132] The topology relationship detection module is used for the pre-established CNN-LSTM network model.

[0133] And according to the topological relationship graph, a pre-established CNN-LSTM network model is trained to obtain a topological relationship detection model for identifying the topological relationship between the offshore wind turbine and the mobile ship;

[0134] The emergency response module is used to identify abnormal ship behaviors in offshore wind farms based on the topological relationship detection model and obtain ship warning data;

[0135] To set up offshore monitoring and early warning equipment based on ship warning data to achieve safety monitoring of offshore wind farms;

[0136] An abnormal ship behavior identification system for an offshore wind farm, further comprising a data analysis and report generation module and an intelligent decision support module;

[0137] The data analysis and report generation module is used to provide in-depth analysis of historical data, including trend analysis of abnormal behavior, statistical information of risk events, and suggestions for improvement measures. By automatically generating detailed reports, it helps managers and decision makers understand the system operation status and achieve continuous improvement and security assurance;

[0138] The intelligent decision support module is used to pre-build an intelligent decision model based on the data and topological analysis results obtained by the data analysis and report generation module. The model can automatically generate decision suggestions or early warning information based on the input data and analysis results. The decision suggestions include adjusting the monitoring plan, optimizing the layout of the wind farm, and notifying relevant departments to intervene.

[0139] Compared with the prior art, the beneficial effects of this embodiment are:

[0140] (1) Through an automated iterative process, real-time monitoring and intelligent analysis of abnormal ship behavior in the offshore wind farm area are achieved, significantly improving the efficiency of navigation safety and risk management.

[0141] (2) Based on the traditional offshore monitoring system, this embodiment adds a remote sensing image real-time acquisition module, a remote sensing image processing module, an offshore windmill detection module, a vector construction module, a topological relationship acquisition module, a topological relationship detection module, an emergency response module, a data analysis and report generation module, and an intelligent decision support module, thereby reducing the dependence on manual monitoring and the response time, and improving the system's automation level and the accuracy of decision support.

[0142] (3) This embodiment can be widely used in the identification of abnormal ship behavior, which is of great significance for improving the intelligent level of maritime traffic management and enhancing maritime safety measures.

[0143] (4) This embodiment uses data collected after the actual operation of the wind farm to evaluate the actual impact of the wind farm on the ship route, and verifies or corrects the previous risk assessment based on this data. This empirically based a posteriori analysis method not only improves the accuracy of risk assessment, but also provides a scientific basis for the continuous optimization and safety management of offshore wind farms.

[0144] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying abnormal ship behavior in an offshore wind farm, characterized in that: The specific steps include: S1: Obtain the windmill interpretation symbol library of historical offshore windmill remote sensing images; Mark the marine targets on the pre-collected remote sensing images of the target area to obtain the initial sample data set; The offshore targets include offshore windmill targets and mobile ship targets; Based on the ship satellite AIS data of the target area remote sensing image, and according to the interpretation marks of the windmill interpretation mark library, the offshore windmills in the initial sample data set are identified and obtained to obtain the optimized data set; S2: training a pre-established Oriented RCNN network model according to the optimized data set to obtain an offshore wind turbine detection model for identifying and extracting offshore wind turbine location information, and converting the offshore wind turbine location information into a wind field vector in the form of a vector layer; S3: Obtain ship satellite AIS data through the ship automatic identification system; And the ship satellite AIS data includes ship position and motion trajectory data; Based on the preset MMSI number, the motion trajectory of the ship satellite AIS data is converted into a trajectory vector; S4: Taking the wind field vector as the target area, obtaining the relative position relationship between the two end points of the trajectory vector and the target area, so as to obtain a topological relationship diagram between the offshore wind turbine and the moving ship; S5: training a pre-established CNN-LSTM network model according to the topological relationship graph to obtain a topological relationship detection model for identifying the topological relationship between the offshore wind turbine and the moving ship; S6: Based on the topological relationship detection model, abnormal ship behavior in offshore wind farms can be identified to obtain ship warning data; The ship warning data includes ship type, frequent time period and high-risk sea area; By setting up offshore monitoring and early warning equipment based on ship warning data, safe monitoring of offshore wind farms can be achieved.

2. The method for identifying abnormal ship behavior in an offshore wind farm according to claim 1, characterized in that: The S1 specifically includes the following steps: S11: Obtain historical offshore windmill remote sensing images, and call corresponding ship satellite AIS data from the existing database according to the shooting time and shooting range of the remote sensing images; S12: acquiring remote sensing frame images frame by frame according to the remote sensing images of the offshore wind turbines, and acquiring characteristic data of the offshore wind turbines in the remote sensing frame images; The characteristic data of the offshore windmill includes geometric profile characteristics, spatial characteristics and spectral characteristics; And define and obtain windmill interpretation signs according to geometric contour features, spatial features and spectral features, and then obtain a windmill interpretation sign library; S13: marking marine targets on the pre-collected remote sensing images of the target area to obtain an initial sample data set; and the marine targets include offshore windmill targets and mobile ship targets; S14: Using the satellite AIS data of ships in the same period as the remote sensing image of the corresponding target area, the moving ship targets are removed from the initial sample data set; And according to the interpretation marks of the windmill interpretation mark library, the offshore windmill targets in the initial sample data set are identified and acquired, and then an optimized data set including only the offshore windmill targets is obtained.

3. A method for identifying abnormal ship behavior in an offshore wind farm according to claim 2, characterized in that: In S2, the pre-established Oriented RCNN network model is trained according to the optimized data set, and the training process is as follows: S21: randomly dividing the optimized data set into a training set and a test set according to a preset ratio; S22: Input the training set into a pre-created Oriented RCNN network model for model training; And based on the constructed model loss function L, confirm whether the output of the trained HRU-Net model converges; If the output of the trained Oriented RCNN network model converges, execute step S23; S23: Evaluate the model accuracy of the trained Oriented RCNN network model based on the test set; The indicators for evaluating the accuracy of the model are the recall rate and accuracy of the model; If it is confirmed that the model accuracy evaluation meets the preset model accuracy threshold, then the trained Oriented RCNN network model is confirmed to be an offshore wind turbine detection model for identifying and extracting offshore wind turbine location information; If it is confirmed that the model accuracy evaluation does not meet the preset model accuracy threshold, the weight parameters of the Oriented RCNN network model are adaptively adjusted based on the back propagation method, and step S22 is repeated.

4. A method for identifying abnormal ship behavior in an offshore wind farm according to claim 3, characterized in that: The expression of the model loss function L constructed in S22 is L=L cls +λ reg L reg Where: L cls represents the classification loss; L reg represents the loss of bounding box regression and angle regression; reg Represents the weight coefficient used to balance the regression loss.

5. The method for identifying abnormal ship behavior in an offshore wind farm according to claim 3, characterized in that: The method for obtaining the topological relationship diagram of offshore wind turbines and mobile ships in S4 is as follows: Taking the wind field vector as the target area, the relative position relationship between the two end points of the trajectory vector and the target area is obtained; The relative position relationship includes an intersection relationship, a containment relationship, an adjacency relationship, a non-intersection relationship and a crossing relationship; The intersection relationship is that the ship's trajectory line passes through the surface area of ​​the windmill, and the starting point and the end point of the trajectory line are both the same boundary of the surface area; The inclusion relationship is that the starting point or the end point of the ship's trajectory is located in the windmill area; The adjacency relationship is that the ship's trajectory touches the boundary of the target area but does not cross it; The non-intersecting relationship means that the ship's trajectory line has no intersection with the target area; The crossing relationship is that the ship's trajectory crosses from one boundary of the target area to another boundary; And obtain the topological relationship diagram of offshore wind turbines and mobile ships based on the relative position relationship.

6. An offshore wind farm abnormal ship behavior identification system, based on any one of claims 1 to 4 offshore wind farm abnormal ship behavior identification method, characterized in that: At least include Remote sensing image real-time acquisition module, remote sensing image processing module, offshore windmill detection module, vector construction module, topological relationship acquisition module, topological relationship detection module and emergency response module; The remote sensing image real-time acquisition module is used to retrieve historical offshore windmill remote sensing images; or to collect and store remote sensing images of the target area in real time; The remote sensing image processing module is used to obtain a windmill interpretation symbol library of historical offshore windmill remote sensing images; Or the pre-collected remote sensing images of the target area are annotated with offshore targets to obtain an initial sample data set, and based on the ship satellite AIS data of the remote sensing images of the target area, combined with the interpretation marks of the windmill interpretation mark library, the offshore windmills in the initial sample data set are identified and obtained to obtain an optimized data set; The offshore windmill detection module is used to establish an Oriented RCNN network model. And train the Oriented RCNN network model based on the optimized data set to obtain an offshore wind turbine detection model for identifying and extracting offshore wind turbine location information; The vector construction module is used to convert the offshore windmill position information detected and identified by the offshore windmill detection model into a wind field vector in the form of a vector layer; Or the movement trajectory of the ship satellite AIS data obtained through the ship automatic identification system can be converted into a trajectory vector based on the preset MMSI number; The topological relationship acquisition module is used to obtain a topological relationship diagram of the offshore wind turbine and the mobile ship according to the wind field vector and the trajectory line vector; The topological relationship detection module is used to establish a CNN-LSTM network model. And train the CNN-LSTM network model according to the topological relationship graph to obtain a topological relationship detection model for identifying the topological relationship between offshore wind turbines and mobile ships; The emergency response module is used to identify abnormal ship behaviors in offshore wind farms based on the topological relationship detection model and obtain ship warning data; By setting up offshore monitoring and early warning equipment based on ship warning data, safe monitoring of offshore wind farms can be achieved.

Images