Fishing boat safety early warning method based on comprehensive data analysis

Abstract

The invention discloses a fishing boat safety early warning method based on comprehensive data analysis, and particularly relates to the technical field of data analysis. Ship data, meteorological data and sea condition data are obtained in real time, preprocessing is carried out, key abnormal features are extracted, dynamic prediction is carried out on security threats in combination with a machine learning model, security threat levels are divided according to prediction results, and the system classifies and filters early warning information according to threat priorities, so that information overload is reduced, and the security threat prediction efficiency is improved. Action suggestions conforming to the operation habits of the sailors are generated based on historical data and real-time environment changes, so that the sailors are helped to make quick decisions; besides, by continuously learning and analyzing the reaction mode of the sailors and automatically adjusting the early warning and suggestion mode, the decision support function is improved, and the navigation safety and the emergency response capability of the ship are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and in particular to a fishing vessel safety early warning method based on comprehensive data analysis. Background Art

[0002] Fishing vessel safety warning systems monitor and analyze various safety risks that fishing vessels may encounter during navigation or operations at sea, issuing advance warnings to facilitate preventive or emergency measures and reduce the risk of accidents. These risks may include severe weather, changing sea conditions, equipment failure, and vessel collisions. Fishing vessel safety warning systems typically rely on modern technologies, such as satellite navigation, meteorological data, and sensors, to collect relevant information in real time and provide timely and effective warnings to ensure the safety of fishermen and their property.

[0003] The existing technology has the following shortcomings:

[0004] As data volumes increase, especially when integrating data from multiple dimensions, crew members may face information overload. While the system may provide a wealth of warnings and recommendations, it can be difficult for crew members to quickly digest and make decisions in the complex and stressful maritime environment. Furthermore, insufficient decision-making support and the inability to provide clear action recommendations can lead to crew members failing to implement optimal responses. Summary of the Invention

[0005] The purpose of the present invention is to provide a fishing vessel safety early warning method based on comprehensive data analysis to solve the shortcomings in the background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solutions: a fishing vessel safety early warning method based on comprehensive data analysis, comprising:

[0007] Acquire ship data, weather data and sea condition data in real time and perform pre-processing;

[0008] Extract several abnormal features from pre-processed real-time data, combine them with machine learning models to dynamically predict security threats, and classify security threats according to the prediction results;

[0009] Classify and filter warning information according to threat priority;

[0010] Make intelligent decisions based on historical data and real-time environmental change data, automatically generate action suggestions that are consistent with the crew's operating habits, and guide the crew to take the best response measures;

[0011] Continuously learn and analyze crew members' reactions and decision-making patterns in different situations, and automatically adjust warning and advice methods.

[0012] Preferably, the ship data is acquired in real time through various sensors and ship equipment, including GPS, AIS, ship monitoring equipment, steering gear and speed control systems and meteorological sensors.

[0013] Preferably, a number of abnormal features are extracted from the pre-processed real-time data, including calculating the deviation between the current position and the planned route based on the historical position trajectory and navigation path of the ship, and generating a route deviation value after analyzing the route deviation. The generation method is:

[0014] Set the longitude and latitude of the current position of the ship to (Lat current ,Lon current );

[0015] A scheduled route is a path consisting of multiple longitude and latitude points, indicating that the ship is sailing along the path. The scheduled route is composed of multiple consecutive navigation points, namely (Lat i ,Lon i ) and (Lat i+1 ,Lon i+1 ), where i=1, 2, ..., n-1 represents a point in the path;

[0016] For each route segment (Lat i ,Lon i ) to (Lat i+1 ,Lon i+1 ), calculate the current ship position (Lat current ,Lon current ) to the shortest distance from the current position to the route segment, which is completed by calculating the perpendicular distance from the point to the line segment; for all route segments on the scheduled route, the shortest distance from the current position to each segment is calculated, and the smallest distance is selected as the route deviation value.

[0017] Preferably, the pressure change rate anomaly value is generated after analyzing the pressure change rate within a fixed time period. The generation method is: obtain the pressure data within a period of time, calculate the pressure change rate, and the pressure change rate refers to the degree of change of the pressure per unit time. The calculation formula is: P(t) is the air pressure value at the current time t, P(t-Δt) is the air pressure value at the previous time point, Δt is the time difference, and V(t) is the rate of change of air pressure. The average and standard deviation of the rate of change of air pressure over a period of time are calculated to determine the normal range of change. The average value of the rate of change of air pressure is: Among them, V(t i ) is the time point t i The air pressure change rate at the moment, m is the total number of data points in the time period, and the standard deviation of the air pressure change rate is: Among them, σ Vis the standard deviation of the rate of change of air pressure; if V(t) exceeds μ V +k·σ V or μ V -k·σ V , where k is a constant, the pressure change rate is considered abnormal and recorded as the pressure change rate abnormal value.

[0018] Preferably, by comparing with historical ocean current data, identifying abnormal ocean current patterns, analyzing the abnormal ocean current patterns and generating ocean current abnormal pattern values, the generation method is:

[0019] Collect the ocean current data of the current time period. The ocean current data is the ocean current speed and direction at each time point, expressed as (V current ,D current ), where V current is the ocean current speed, D current Is the direction of the ocean current. Since the ocean current direction is an angle value, the statistical method of angle is needed to calculate its mean and standard deviation. The vector representation is used to calculate the mean value μ of the ocean current direction. D and standard deviation σ D The deviation of the ocean current speed ΔV is calculated by comparing the current ocean current speed with the average value of the historical ocean current speed. The deviation of the ocean current direction ΔD is measured by calculating the deviation between the current direction and the historical direction. The ocean current anomaly mode value EF is calculated as follows: Among them, α and β are weight factors, which represent the influence weight of ocean current speed and direction on the overall abnormal pattern value.

[0020] Preferably, the route deviation value, the air pressure change rate anomaly value and the ocean current anomaly pattern value are converted into a comprehensive feature vector, and the comprehensive feature vector is used as the input of the machine learning model. The machine learning model uses each set of comprehensive feature vectors to predict the ship's security threat risk score value label as the prediction target, and takes minimizing the sum of the prediction errors of all security threat risk score value labels as the training target. The machine learning model is trained until the sum of the prediction errors reaches convergence, and the model training is stopped. The security threat risk score value is determined according to the model output result, wherein the machine learning model is a polynomial regression model.

[0021] Preferably, the obtained security threat risk score of the ship is compared with a gradient risk threshold, the gradient risk threshold includes a first risk threshold and a second risk threshold, and the first risk threshold is less than the second risk threshold, and the security threat risk score is compared with the first risk threshold and the second risk threshold respectively;

[0022] If the security threat risk score is greater than the second risk threshold, it is classified as a high-risk level, indicating that the ship is in great danger and requires immediate emergency measures;

[0023] If the security threat risk score is greater than or equal to the first risk threshold and less than or equal to the second risk threshold, it is classified as a medium risk level and the route needs to be adjusted or preventive measures need to be taken;

[0024] If the security threat risk score value is less than the first risk threshold, it is classified as a low risk level. There is a certain risk, but no immediate measures are required and the ship continues to sail.

[0025] Preferably, high risk level, medium risk level and low risk level are converted into corresponding threat priorities, wherein high risk level corresponds to high priority warning, medium risk level corresponds to medium priority warning, and low risk level corresponds to low priority warning; filtering rules include: high priority warning: directly reminding the crew through sound and light alarms, text messages and voice to ensure that the crew can respond immediately; medium priority warning: reminding the crew through non-invasive means; low priority warning: reminding the crew through reports or regular inspections, without the need for immediate processing.

[0026] Optimally, the system continuously learns and analyzes crew members’ reactions and decision-making patterns in different situations, and automatically adjusts warnings and recommendations, including:

[0027] Define the action space, the ship's action options, and generate warning information;

[0028] Design a reward and penalty mechanism to reward crew members if their actions improve navigation efficiency; conversely, punish them if they ignore warnings or take inappropriate actions that lead to risks;

[0029] Through Q-learning or deep reinforcement learning, initialize the Q table or policy network, record the Q values ​​under different states, and train through continuous exploration and utilization to optimize decision-making behavior.

[0030] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0031] 1. This invention, through a comprehensive data analysis method for fishing vessel safety warnings, effectively addresses the information overload problem in traditional systems. By acquiring real-time vessel, meteorological, and sea condition data and performing efficient preprocessing, the system extracts key anomaly features. Combined with machine learning models, it performs dynamic predictions and accurately classifies vessel safety threat levels. This method not only categorizes and filters warning information based on threat priority, reducing unnecessary interference, but also automatically generates personalized action recommendations for crew members based on historical data and real-time environmental changes, helping them make optimal decisions in complex environments and improving navigation safety and efficiency.

[0032] 2. This invention utilizes reinforcement learning technology to continuously learn and analyze crew members' decision-making patterns and automatically adjust warning and advisory methods. By defining an appropriate action space and designing reward and penalty mechanisms, the system can optimize decision-making behavior and adapt to the response requirements in different situations. Through training using Q-learning or deep reinforcement learning, the system gradually optimizes warning strategies, ensuring that crew members always receive timely and effective decision-making support in a changing environment. This technical solution significantly improves the system's intelligence level, reduces crew stress during complex navigation, and ensures the safety and efficiency of fishing vessels. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0034] Figure 1 This is a mind map of the method of the present invention. DETAILED DESCRIPTION

[0035] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0036] For examples, see Figure 1 As shown, the fishing vessel safety early warning method based on comprehensive data analysis described in this embodiment includes:

[0037] Acquire ship data, weather data and sea condition data in real time and perform pre-processing;

[0038] Extract several abnormal features from pre-processed real-time data, combine them with machine learning models to dynamically predict security threats, and classify security threats according to the prediction results;

[0039] Classify and filter warning information according to threat priority;

[0040] Make intelligent decisions based on historical data and real-time environmental change data, automatically generate action suggestions that are consistent with the crew's operating habits, and guide the crew to take the best response measures;

[0041] Continuously learn and analyze crew members' reactions and decision-making patterns in different situations, and automatically adjust warning and advice methods.

[0042] Real-time acquisition and preprocessing of vessel, weather, and sea condition data is a crucial step in the fishing vessel safety early warning system. This process not only ensures the timeliness and accuracy of the data but also provides a solid foundation for subsequent data analysis and safety prediction.

[0043] Ship data is usually acquired in real time through various sensors and ship equipment, including but not limited to:

[0044] GPS (Global Positioning System): Provides real-time position, speed and heading information of the ship.

[0045] AIS (Automatic Identification System): Obtains basic information about the ship, such as its name, type, onboard status, navigation status, etc., as well as its distance and relative position to other ships.

[0046] Ship monitoring equipment: such as engine monitoring systems, fuel consumption monitoring systems, hull status sensors, temperature sensors, etc., providing real-time working status information of internal equipment on the ship.

[0047] Steering gear and speed control system: Provides real-time data on ship control systems, such as rudder angle and speed changes, to help assess the stability of ship operations.

[0048] Meteorological sensors (shipborne): Some ships are equipped with shipborne meteorological instruments to collect real-time and close-range environmental data such as wind speed, wind direction, temperature, and humidity.

[0049] Data collection frequency: To ensure real-time data, ship data collection is typically updated every second or every minute, depending on the ship's operational needs and system requirements. All data is transmitted in real time to a shore-based platform or a processing unit on the ship via wireless communication methods (such as satellite communication or wireless networks).

[0050] Get weather data in real time:

[0051] Meteorological satellites: These provide a wide range of weather information, including wind speed, temperature, air pressure, precipitation, cloud cover, and other data. Detailed weather forecasts for specific sea areas can be obtained through satellite systems.

[0052] Weather stations: Real-time data collected by marine weather stations, shore-based meteorological stations, or ocean buoys. They provide real-time weather data for a specific area, including sea surface wind speed, wind direction, temperature, humidity, etc.

[0053] Weather forecast data: Obtain weather forecasts for the next few hours to days through weather forecast models, including early warning information for extreme weather events such as storms, hurricanes, and precipitation.

[0054] Data collection frequency: Meteorological data is usually updated once an hour, but in the event of extreme weather changes, warning information may be updated more frequently (for example, every 30 minutes or 15 minutes).

[0055] Get real-time sea condition data:

[0056] Ocean buoys: Ocean buoys are installed on the sea surface and can monitor sea conditions in real time, including wave height, wave period, current speed, tidal information, salinity, temperature, etc.

[0057] Radar and sonar systems: A ship's own radar and sonar systems provide data about surrounding sea conditions, such as the height and shape of waves on the surface. Radar data can also be used to detect other nearby ships or obstacles.

[0058] Tide and tidal current forecast: Through professional tide and tidal current prediction models, we can obtain tidal information of specific sea areas and understand the impact of tidal changes on ship navigation.

[0059] Ocean model data: Forecast data obtained from global or regional ocean models can provide forecasts of sea conditions over the next few days (e.g., sea surface fluctuations, temperature changes, current direction, etc.).

[0060] Data collection frequency: Sea state data is generally updated once an hour, and some key areas (such as near ports or storm fronts) may be updated more frequently.

[0061] Real-time ship, weather, and sea condition data are often affected by various factors (such as equipment transmission delays, sensor errors, data loss, etc.), so preprocessing is necessary to ensure data quality and prepare for subsequent analysis. Specific preprocessing steps include:

[0062] Remove outliers: Eliminate abnormal data caused by sensor failure or signal loss (such as negative wind speed or sea surface wave height exceeding a reasonable range).

[0063] Handling missing data: If some data is missing, use interpolation methods or forward filling, backward filling and other methods to fill the gaps to ensure data continuity.

[0064] Data format standardization: Convert data collected by different sensors and devices into a unified format to ensure data compatibility in subsequent analysis.

[0065] Time synchronization: Ensure that timestamps are aligned across all data sources. This is especially important as vessel position, weather, and sea condition data may come from different devices or services. Aligning data using a unified time standard (such as UTC) ensures timeliness and consistency across different data sources.

[0066] Data denoising and filtering: Removing noise from the data through filters (such as Kalman filters or median filters) helps improve the smoothness and accuracy of the data, especially when there may be interference during sensor acquisition.

[0067] Data normalization and standardization: Data of different dimensions (such as wind speed and temperature) are normalized to allow for unified analysis and integration of different types of data. This ensures that each type of data has equal influence in the calculation, preventing certain data from excessively influencing the analysis results.

[0068] Ship, weather, and sea condition data originate from multiple different data sources and sensors. After preprocessing, they are fused. Data fusion algorithms (such as weighted averaging, Kalman filtering, and multi-sensor data fusion) integrate these disparate data sources into a unified data stream, generating a comprehensive and accurate real-time view of the environment.

[0069] The processed data is uploaded to the cloud or local server for data visualization and trend analysis on the real-time monitoring platform. This data can provide a basis for fishing vessels' navigation decisions and support subsequent safety warnings and risk analysis.

[0070] In this invention, the real-time acquisition and preprocessing of vessel, meteorological, and sea condition data is fundamental to the effective operation of the fishing vessel safety warning system. Through efficient data collection, cleaning, synchronization, and integration, it provides an accurate and reliable basis for subsequent risk prediction, warning information generation, and decision support, significantly improving the safety and response capabilities of fishing vessels in complex maritime environments.

[0071] Abnormal feature extraction of ship data includes:

[0072] Abnormal speed: Speed ​​is an important indicator for safe navigation of a ship. Too fast or too slow speed may indicate potential faults or abnormal conditions.

[0073] Abnormal detection: If the ship's speed suddenly exceeds the preset range (such as ±20% of the planned speed), or suddenly slows down or speeds up without obvious external reasons (such as strong winds or bad sea conditions), it may mean mechanical failure or crew error.

[0074] Extraction method: Identify anomalies by performing statistical analysis on the ship's speed (such as mean, standard deviation, and historical speed change trends).

[0075] Abnormal vessel position: The vessel's position data can reveal its navigation trajectory. Deviation from the normal course or set navigation path may indicate an abnormality during the navigation process.

[0076] Anomaly detection: If a vessel's position deviates significantly from its planned route, this could indicate a navigation system malfunction, a maneuvering error, or the vessel encountering strong winds.

[0077] Extraction method: Based on the ship's historical position trajectory and navigation path, the deviation between the current position and the planned route is calculated. If it exceeds a certain threshold, an abnormal alarm is triggered.

[0078] Abnormal status of ship machinery: The status of a ship's mechanical equipment directly affects its safety, such as abnormalities in engine temperature, fuel consumption, transmission system, etc.

[0079] Abnormality detection: If data such as ship engine temperature, fuel consumption rate, and mechanical failure suddenly change or exceed the normal operating range, there may be mechanical failure or equipment abnormality.

[0080] Extraction method: Use the data from the equipment monitoring system to perform threshold detection to determine whether it exceeds the normal operating range. At the same time, combine historical fault data to establish a characteristic pattern to determine whether it is an abnormality.

[0081] Based on the historical position trajectory and navigation path of the ship, the deviation between the current position and the planned route is calculated. The route deviation is analyzed and the route deviation value is generated. The specific generation method is as follows:

[0082] Set the longitude and latitude of the current position of the ship to (Lat current ,Lon current ).

[0083] A scheduled route is usually a path consisting of multiple longitude and latitude points, indicating that the ship should sail along this path. i ,Lon i ) and (Lat i+1 ,Lon i+1 ), where i=1, 2, ..., n-1 represents a point in the path.

[0084] For each route segment (Lat i ,Lon i ) to (Lat i+1 ,Lon i+1 ), calculate the current ship position (Lat current ,Lon current ) to the shortest distance of the route segment, which is completed by calculating the perpendicular distance between the point and the line segment. The formula for the shortest distance from a point to a line segment is: D is the shortest distance from the current position of the ship to the route segment.

[0085] For all segments on the planned route, the shortest distance from the current position to each segment is calculated, and the minimum distance is selected as the route deviation value. In other words, the minimum deviation between the current position of the ship and the planned route is selected.

[0086] Abnormal feature extraction of meteorological data includes:

[0087] Abnormal wind speed: Wind speed is crucial to the safety of ship navigation. Extreme wind speeds (such as sudden storms or hurricanes) may threaten the safety of ships.

[0088] Anomaly detection: If the wind speed increases sharply or changes direction suddenly (such as the formation of a storm), it will have a serious impact on the navigation of the ship.

[0089] Extraction method: By analyzing historical wind speed data, the normal range and trend of wind speed changes are set, and any sudden change in wind speed that exceeds the threshold can be regarded as an anomaly.

[0090] Abnormal air pressure: Air pressure is an important indicator for predicting severe weather (such as storms, typhoons, cyclones, etc.). A sudden drop in air pressure is often accompanied by storms or other extreme weather.

[0091] Anomaly Detection: A sharp drop in air pressure may indicate an approaching storm or typhoon and is a key feature for early warning.

[0092] Extraction method: By detecting the rate of change of air pressure in a short period of time (such as when the air pressure drops by more than a certain value), the arrival of extreme weather can be identified in advance.

[0093] Abnormal temperature and humidity: Abnormal temperature and humidity fluctuations are often closely related to changes in weather systems. For example, extremely low temperatures may cause freezing, while extremely high temperatures may increase the risk of equipment failure on board ships.

[0094] Anomaly detection: If the temperature or humidity changes too quickly (e.g., exceeding a certain rate change threshold), it may indicate extreme weather or climate change.

[0095] Extraction method: By setting temperature and humidity change thresholds, calculating the gap between the actual change rate and the predetermined range, sudden weather changes can be quickly identified.

[0096] The pressure change rate anomaly value is generated by analyzing the pressure change rate within a fixed time period. The generation method is as follows:

[0097] Obtain air pressure data over a period of time. Air pressure data comes from sensors or weather stations and is typically collected in hourly units. P(t) represents the air pressure value at time t. Time t is expressed in hours (h) or can be adjusted to finer granularity such as minutes or seconds depending on the actual situation. Select a fixed time period, such as the pressure change every hour or every 10 minutes.

[0098] The rate of change of air pressure refers to the degree of change of air pressure per unit time, and the calculation formula is: P(t) is the air pressure at the current time t, and P(t-Δt) is the air pressure at the previous time (e.g., the previous hour or 10 minutes). Δt is the time difference, expressed in hours (e.g., 1 hour or 10 minutes, depending on the specific requirements). V(t) is the rate of change of air pressure, expressed in air pressure units / time (e.g., hPa / h, indicating the change in air pressure per hour).

[0099] Choosing the right time period Δt is crucial. If the time period is too long, pressure changes may be smoothed out; if it is too short, the sensor may be affected by noise. Typically, 1 hour or 10 minutes is used.

[0100] To further assess the degree of abnormality in air pressure changes, a statistical analysis of the air pressure change rate can be performed. The normal range of changes can be determined by calculating the average and standard deviation of the air pressure change rate over a period of time. The average value of the air pressure change rate is: Among them, V(t i ) is the time point t i The rate of change of air pressure at the moment, m is the total number of data points in the time period. The standard deviation of the rate of change of air pressure is: Among them, σ V It is the standard deviation of the rate of change of air pressure, which reflects the fluctuation amplitude of the rate of change of air pressure.

[0101] By calculating the average and standard deviation of the pressure change rate, a threshold can be set to determine whether an abnormality occurs. V +k·σ V or μ V -k·σ V (where k is a constant, usually 2 or 3), the air pressure change rate can be considered abnormal and recorded as an abnormal value of the air pressure change rate.

[0102] Abnormal feature extraction of sea state data includes:

[0103] Abnormal wave height: Wave height is a key factor affecting ship stability. Excessively high waves may cause the ship to capsize or be seriously damaged.

[0104] Anomaly detection: If the wave height exceeds the safety limit set by the ship, or the wave height changes suddenly (such as rising sharply in a short period of time), it will cause danger to the ship's navigation.

[0105] Extraction method: By comparing the wave height with historical data, the warning threshold is set, and an abnormal warning is triggered when the wave height exceeds the range.

[0106] Abnormal ocean currents: The speed and direction of ocean currents directly affect the ship's navigation route. Abnormal ocean currents may make it impossible for the ship to maintain its planned route or increase fuel consumption.

[0107] Anomaly detection: Severe changes in ocean current speed or direction may cause ships to deviate from their course or increase navigation risks.

[0108] Extraction method: Identify abnormal current patterns by comparing them with historical current data. In particular, changes in the magnitude, duration, or direction of currents exceeding a certain threshold are considered abnormal.

[0109] Abnormal tides: Tidal changes have a significant impact on berthing and ship operations. Abnormal tides may cause navigation difficulties or make the port unusable.

[0110] Abnormality detection: If the tide changes abnormally (such as the tide is too large or too small, or the tidal fluctuation amplitude exceeds the normal range), it will affect the operating efficiency and safety of the ship.

[0111] Extraction method: By comparing the tide prediction data with the actual tide, the abnormal amplitude and change trend of the tide are calculated.

[0112] By comparing with historical ocean current data, abnormal ocean current patterns are identified and analyzed to generate ocean current abnormal pattern values. The generation method is as follows:

[0113] Collect and pre-process ocean current data for the current time period. Ocean current data includes the speed and direction of the current and is usually obtained through equipment such as ocean buoys, radar, or satellites.

[0114] The ocean current data is the ocean current speed and direction at each time point, expressed as (V current ,D current ), where V current is the ocean current speed, D current The direction of the ocean current. Use historical records to obtain the standard ocean current pattern in a certain area (such as the average ocean current speed and direction over the past few days, weeks, or months).

[0115] To identify anomalous current patterns, it's first necessary to establish a standard historical current model. This model is typically constructed by calculating the mean and standard deviation of historical current data. The mean and standard deviation of historical current velocities are calculated. Since current direction is an angular value, angular statistical methods are used to calculate its mean and standard deviation. Vector representation can be used to calculate the mean of current direction.

[0116] Average value of current direction: Among them, D historical (t i ) is the historical current direction (in degrees). The angle is converted into a vector using the sin and cos functions, and the average is calculated. The standard deviation is calculated as: Among them, σ D is the standard deviation of the direction, and g is the total number of points of historical data.

[0117] The deviation of the current speed can be calculated by comparing the current current speed with the average of the historical current speed: ΔV = |V current -μ V |; If ΔV exceeds a certain multiple of the historical current velocity standard deviation, the current velocity is considered abnormal. The deviation of the current direction is measured by calculating the deviation between the current direction and the historical direction. Since the direction is an angle, the angle difference needs to be calculated: ΔD = min(|D current -μ D |,360°-|D current -μ D |); if ΔD exceeds a certain threshold (e.g., 2 or 3 times the standard deviation), the current direction is considered abnormal.

[0118] In order to combine the abnormal values ​​of speed and direction, a comprehensive ocean current abnormal mode value EF can be generated by weighted average method, which is expressed as: Where α and β are weight factors, representing the weight of the influence of ocean current speed and direction on the overall anomaly pattern value. Generally, if speed changes are more important than direction changes, α>β can be set.

[0119] The route deviation value, the abnormal value of the air pressure change rate and the abnormal ocean current pattern value are converted into a comprehensive feature vector, and the comprehensive feature vector is used as the input of the machine learning model. The machine learning model uses each set of comprehensive feature vectors to predict the ship's security threat risk score value label as the prediction target, and takes minimizing the sum of the prediction errors of all security threat risk score value labels as the training target. The machine learning model is trained until the sum of the prediction errors reaches convergence, and the model training is stopped. The security threat risk score value is determined according to the model output results. Among them, the machine learning model is a polynomial regression model.

[0120] Comparing the obtained ship security threat risk score with a gradient risk threshold, where the gradient risk threshold includes a first risk threshold and a second risk threshold, and the first risk threshold is less than the second risk threshold, and comparing the security threat risk score with the first risk threshold and the second risk threshold respectively;

[0121] If the security threat risk score is greater than the second risk threshold, it is classified as a high-risk level, indicating that the ship is in great danger and may need to take immediate emergency measures, such as avoiding navigation or returning to port;

[0122] If the security threat risk score is greater than or equal to the first risk threshold and less than or equal to the second risk threshold, it is classified as a medium risk level and requires increased vigilance, and may require adjusting the route or taking preventive measures;

[0123] If the security threat risk score value is less than the first risk threshold, it will be classified as a low risk level. There is a certain risk, but no immediate measures are required and the ship can continue sailing.

[0124] Warning information can come from a variety of sources, such as weather warnings, sea condition monitoring, equipment health monitoring, etc. According to the threat level of these warnings, the warning information is classified and filtered:

[0125] Convert high risk level, medium risk level and low risk level into corresponding threat priorities, where high risk level corresponds to high priority warning, medium risk level corresponds to medium priority warning, and low risk level corresponds to low priority warning;

[0126] High priority (urgent): These warnings represent a major safety threat to the ship and require immediate action. They include:

[0127] Extreme weather: such as storms, typhoons, thunderstorms, etc.

[0128] Abnormal sea conditions: such as high waves, large currents, ice areas, etc., may cause the ship to lose control or capsize.

[0129] Major equipment failures: such as main engine failure, navigation system failure, etc., which directly affect the controllability and safety of the ship.

[0130] Medium priority (warning): These warning messages indicate that there is a certain risk, but no immediate emergency measures are required. They include:

[0131] Strong wind: The wind speed is high, affecting the stability of navigation, but not posing an immediate threat.

[0132] Changes in ocean currents: Changes in the direction or strength of ocean currents may affect the course and require crew members to be alert.

[0133] Minor equipment failure: If there is a problem with the ship's auxiliary equipment, it may affect the performance of the ship but does not immediately threaten safety.

[0134] Low priority (alert): These warning messages represent potential minor risks that can be handled by the crew at their leisure. They include:

[0135] Minor weather changes: such as temperature changes, slight fluctuations in air pressure, etc.

[0136] Equipment health status monitoring: The equipment status is normal, but there is a slight performance degradation or minor fault.

[0137] Filtering rules include:

[0138] High-priority warning: Alert the crew directly through sound and light alarms, text messages, voice, etc. to ensure that the crew can respond immediately.

[0139] Medium priority warning: The crew is reminded through non-intrusive means (such as interface pop-ups, non-emergency notifications, etc.), and it is recommended to be vigilant.

[0140] Low priority alerts: alert the crew through reports or regular inspections, and do not require immediate action.

[0141] Making intelligent decisions based on historical data and real-time environmental change data, and automatically generating action recommendations that align with crew operating habits, is a key component of modern intelligent navigation systems. Its purpose is to provide customized action recommendations to crew members by comprehensively analyzing historical and real-time data, thereby helping them make optimal decisions in complex navigation environments and improving navigation safety and efficiency.

[0142] Ship historical data: including past navigation tracks, ship operation logs (such as speed, heading, ship equipment status), weather and sea conditions information, equipment failure records, etc.

[0143] Crew operating habits data: This records crew operating behaviors, such as how they respond to adverse weather conditions or changing sea conditions in specific situations. This data helps analyze crew decision-making patterns and automatically adjust system recommendations to align with the crew's operating style.

[0144] Historical accident data: Analyze past ship accidents and similar environmental conditions to assess the conditions under which accidents occurred and the crew's response patterns when the accident occurred.

[0145] Meteorological data: Real-time weather information such as wind speed, air pressure, precipitation, etc., especially extreme weather that affects ship navigation (such as storms, thunderstorms, etc.).

[0146] Sea condition data: real-time wave height, tide, current and other data.

[0147] Ship status data: ship position, speed, heading, and real-time status of ship machinery and equipment (such as engine health, fuel consumption, equipment failure alarms, etc.).

[0148] Denoising and supplementing missing data: Real-time data may be affected by factors such as equipment failure and environmental interference, so data cleaning and interpolation processing are performed.

[0149] Data standardization and normalization: Standardize data from different sources and dimensions to ensure that different features have equal influence in the model.

[0150] Data synchronization: Ensure that historical data and real-time data are synchronized in time dimension.

[0151] Decision support systems rely on intelligent algorithms to combine historical data with real-time environmental data to generate action recommendations that the crew can understand and implement. The following are the key steps:

[0152] Use historical data to train machine learning models (such as regression analysis, support vector machines, random forests, etc.) so that the models can identify the decision-making behaviors commonly taken by crew members under similar environmental conditions.

[0153] By analyzing the crew's operating behavior in historical data and using cluster analysis, classification models and other technologies, the crew's decision-making patterns can be identified.

[0154] Build a model to link environmental changes with crew decisions. For example, under certain wind speed and wave height combinations, the crew may choose to slow down or change course.

[0155] Through real-time data analysis and combined with historical behavior patterns, action recommendations tailored to current environmental conditions are generated.

[0156] Analyze real-time weather, sea conditions and ship status to identify current navigation safety threats (such as strong winds, high waves, equipment failures, etc.).

[0157] Based on the crew's past decision-making behavior, the system can automatically adjust the action recommendations to make them more consistent with the crew's operating habits. For example, if a crew member is accustomed to changing course when wind speeds exceed a certain value, the system will provide this action recommendation based on the real-time wind speed.

[0158] Online learning and optimization through machine learning algorithms (such as reinforcement learning, deep learning, etc.) ensure that the system's recommendations continue to improve over time.

[0159] Reinforcement Learning: The system uses reinforcement learning algorithms to train models based on historical and real-time data, continuously optimizing recommendation strategies. The model uses a "reward" mechanism to learn the optimal decision path. For example, when the system's navigational recommendations avoid a potential accident, the system "rewards" that decision, gradually forming more accurate warnings and action recommendations.

[0160] Multi-objective optimization: Generate multiple action suggestions based on multiple objectives (such as shortest voyage, minimum fuel consumption, minimum risk, etc.), and optimize the selection based on the crew's preferences and actual operational needs.

[0161] Based on the intelligent decision-making model, the system generates the following types of action recommendations:

[0162] Course adjustment suggestion: In extreme weather or bad sea conditions, the system recommends adjusting the course to avoid dangerous areas.

[0163] Speed ​​adjustment suggestions: Provide suggestions for reducing or increasing speed based on wind speed, wave height, etc.

[0164] Emergency plan: When equipment failure or extreme weather is detected, the system generates an emergency plan, such as returning home or taking a safe harbor.

[0165] Equipment maintenance recommendations: Based on equipment health data, the system can remind crew members to perform necessary equipment inspections and maintenance.

[0166] The system can be personalized based on the crew's preferences. For example, a crew member may prefer to take a more conservative approach in some situations, but be more proactive in others. The system can generate personalized recommendations based on these preferences.

[0167] As crew members gradually become familiar with the system's suggestions, the system can continuously adjust the suggestions based on crew members' feedback and historical operating behaviors to make them more suitable for crew members' operating habits.

[0168] Presentation formats include:

[0169] Visual interface: Real-time environmental data and suggestions are displayed through graphical interfaces (such as navigation charts, real-time data panels, etc.), allowing crew members to intuitively understand navigation status and safety recommendations.

[0170] Voice prompts and alarms: Provide real-time advice to crew members through the voice system and issue voice alarms in emergency situations.

[0171] Push notifications: Specific action recommendations are pushed via mobile devices, ship displays, etc., ensuring that crew members have access to the latest information at all times.

[0172] To continuously learn and analyze crew members' reactions and decision-making patterns in different situations and automatically adjust warnings and recommendations, reinforcement learning algorithms can be used. Reinforcement learning is a machine learning method based on a reward mechanism that continuously optimizes decision-making strategies through interaction with the environment, enabling the system to continuously improve itself.

[0173] In the ship safety warning system, the reinforcement learning agent is the warning and decision support system. Its environment includes real-time data of the ship (such as weather, sea conditions, ship status, etc.), and the behavior is the safety warning or action recommendations given based on this data.

[0174] In the ship safety warning system, the environment can be regarded as all external conditions of the ship, including: meteorological conditions: wind speed, air pressure, temperature, etc. Sea state data: wave height, tide, current, etc. Ship equipment status: speed, heading, fuel consumption, mechanical equipment status, etc. Crew operation history: the crew's response to different situations (such as the route and speed chosen when responding to a storm). These environmental conditions together constitute the state space of the system, which is a collection of all current information describing the system. For example, the current state of the ship may include: Wind speed: wind speed value per hour. Ship speed: current speed value. Current: direction and strength of the current. Crew behavior: the decision-making behavior of the crew in the current situation (such as whether to change the course).

[0175] In reinforcement learning, the action space defines all possible behaviors or decisions that an agent can choose. In a ship safety warning system, the action space includes: Course adjustment: changing the ship's course to avoid dangerous areas; Speed ​​adjustment: increasing or decreasing speed to avoid inclement weather or reduce fuel consumption; Equipment adjustment: prompting the crew to inspect or repair equipment based on real-time data; Alert generation: generating different levels of alerts based on current environmental conditions. Each action affects the system state, helping the crew take appropriate action to address environmental changes.

[0176] Rewards and penalties are the core of reinforcement learning. The agent optimizes its decision-making through rewards or penalties obtained from the environment. In the ship safety warning system, the reward and penalty mechanism can be designed as follows:

[0177] Rewards: Accident avoidance: If the actions taken by the crew (such as changing course, reducing speed, etc.) successfully avoid an accident or reduce risks, a reward will be given. Efficient operation: If the crew optimizes the navigation (such as saving fuel, shortening the voyage, etc.) by adjusting the speed and course, a reward will be given.

[0178] Penalties: Failure to avoid risk: If the crew ignores warning information and fails to change course or slow down in time, resulting in a potential danger, they will be penalized. Inappropriate response: If the crew's response (such as excessive speed reduction or inappropriate course) reduces navigation efficiency or fails to effectively avoid threats, they will be penalized.

[0179] To apply reinforcement learning, you first need to initialize a Q-table (or policy network), which records the Q-value (i.e., expected reward) of taking an action in different states. A higher Q-value indicates that taking that action in that state will bring greater long-term rewards.

[0180] If you use Q-learning, you can initialize a Q table to record the Q value of the state-action pair.

[0181] If deep reinforcement learning (Deep RL) is used, the policy or value function can be approximated by a neural network, avoiding the use of Q tables in large state spaces.

[0182] During training, the agent needs to continuously explore and exploit:

[0183] Exploration: Randomly select actions to explore the state space and action space and try different strategies.

[0184] Exploitation: Select the optimal action based on the current Q-table or policy network to maximize the expected reward in the current state.

[0185] Through continuous exploration and exploitation, the agent gradually learns which actions lead to better results in specific situations.

[0186] Each time a crew member executes the system's recommendations, the system receives feedback (Reward) from the environment and then updates the Q value or policy. For Q-learning, the update rule is: Q(s,a) is the Q value of selecting action a in the current state s, α is the learning rate, which controls the impact of new information on the Q value, R represents the immediate reward obtained from the environment, and γ is the discount factor, which represents the weight of future rewards. represents the maximum Q value of selecting the optimal action in the new state s′.

[0187] As training progresses, the system continuously updates its Q-values ​​or strategies, eventually converging to learn an optimal set of action strategies. Based on real-time data and historical experience, the system automatically generates action recommendations that align with the crew's operational habits and enables optimal decisions in similar situations.

[0188] In this invention, through reinforcement learning, the system automatically generates action recommendations that align with crew members' operational habits under varying environmental conditions, adjusting and optimizing them based on real-time data. This intelligent decision-making system based on reinforcement learning not only improves ship safety but also helps crew members better adapt to different navigation scenarios, reducing the risk of human error and ultimately improving navigation efficiency and safety.

[0189] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0190] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0191] It should be understood that the term "and / or" herein is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent three situations: A exists alone, A and B exist at the same time, and B exists alone, where A and B may be singular or plural. In addition, the character " / " herein generally indicates that the objects associated with each other are in an "or" relationship, but it may also indicate an "and / or" relationship, which can be understood by referring to the context. A person of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0192] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A fishing vessel safety early warning method based on comprehensive data analysis, characterized by: include: Acquire ship data, weather data and sea condition data in real time and perform pre-processing; Extract several abnormal features from pre-processed real-time data, combine them with machine learning models to dynamically predict security threats, and classify security threats according to the prediction results; Classify and filter warning information according to threat priority; Make intelligent decisions based on historical data and real-time environmental change data, automatically generate action suggestions that are consistent with the crew's operating habits, and guide the crew to take the best response measures; Continuously learn and analyze crew members' reactions and decision-making patterns in different situations, and automatically adjust warning and advice methods.

2. The fishing vessel safety early warning method based on comprehensive data analysis according to claim 1 is characterized in that: Ship data is acquired in real time through various sensors and ship equipment, including GPS, AIS, ship monitoring equipment, steering and speed control systems, and meteorological sensors.

3. The fishing vessel safety early warning method based on comprehensive data analysis according to claim 1 is characterized in that: Extract several abnormal features from the pre-processed real-time data, including calculating the deviation between the current position and the planned route based on the ship's historical position trajectory and navigation path, and analyzing the route deviation to generate a route deviation value. The generation method is as follows: Set the longitude and latitude of the current position of the ship to (Lat current ,Lon current ); A scheduled route is a path consisting of multiple longitude and latitude points, indicating that the ship is sailing along the path. The scheduled route is composed of multiple consecutive navigation points, namely (Lat i ,Lon i ) and (Lat i+1 ,Lon i+1 ), where i=1, 2, ..., n-1 represents a point in the path; For each route segment (Lat i ,Lon i ) to (Lat i+1 ,Lon i+1 ), calculate the current ship position (Lat current ,Lon current ) to the shortest distance from the current position to the route segment, which is completed by calculating the perpendicular distance from the point to the line segment; for all route segments on the scheduled route, the shortest distance from the current position to each segment is calculated, and the smallest distance is selected as the route deviation value.

4. The fishing vessel safety early warning method based on comprehensive data analysis according to claim 3 is characterized by: The pressure change rate anomaly value is generated by analyzing the rate of change of air pressure within a fixed time period. The generation method is: obtain air pressure data within a period of time and calculate the air pressure change rate. The air pressure change rate refers to the degree of change of air pressure per unit time. The calculation formula is: P(t) is the air pressure value at the current time t, P(t-Δt) is the air pressure value at the previous time point, Δt is the time difference, and V(t) is the rate of change of air pressure. The average and standard deviation of the rate of change of air pressure over a period of time are calculated to determine the normal range of change. The average value of the rate of change of air pressure is: Among them, V(t i ) is the time point t i The air pressure change rate at the moment, m is the total number of data points in the time period, and the standard deviation of the air pressure change rate is: Among them, σ V is the standard deviation of the rate of change of air pressure; if V(t) exceeds μ V +k·σ V or μ V -k·σ V , where k is a constant, the pressure change rate is considered abnormal and recorded as the pressure change rate abnormal value.

5. The fishing vessel safety early warning method based on comprehensive data analysis according to claim 4 is characterized in that: By comparing with historical ocean current data, abnormal ocean current patterns are identified and analyzed to generate ocean current abnormal pattern values. The generation method is as follows: Collect the ocean current data of the current time period. The ocean current data is the ocean current speed and direction at each time point, expressed as (V current ,D current ), where V current is the ocean current speed, D current Is the direction of the ocean current. Since the ocean current direction is an angle value, the statistical method of angle is needed to calculate its mean and standard deviation. The vector representation is used to calculate the mean value μ of the ocean current direction. D and standard deviation σ D The deviation of the ocean current speed ΔV is calculated by comparing the current ocean current speed with the average value of the historical ocean current speed. The deviation of the ocean current direction ΔD is measured by calculating the deviation between the current direction and the historical direction. The ocean current anomaly mode value EF is calculated as follows: Among them, α and β are weight factors, which represent the influence weight of ocean current speed and direction on the overall abnormal pattern value.

6. The fishing vessel safety early warning method based on comprehensive data analysis according to claim 5 is characterized in that: The route deviation value, the abnormal value of the air pressure change rate and the abnormal ocean current pattern value are converted into a comprehensive feature vector, and the comprehensive feature vector is used as the input of the machine learning model. The machine learning model uses each set of comprehensive feature vectors to predict the ship's security threat risk score value label as the prediction target, and takes minimizing the sum of the prediction errors of all security threat risk score value labels as the training target. The machine learning model is trained until the sum of the prediction errors reaches convergence, and the model training is stopped. The security threat risk score value is determined according to the model output results. Among them, the machine learning model is a polynomial regression model.

7. The fishing vessel safety early warning method based on comprehensive data analysis according to claim 6 is characterized in that: Comparing the obtained ship security threat risk score with a gradient risk threshold, where the gradient risk threshold includes a first risk threshold and a second risk threshold, and the first risk threshold is less than the second risk threshold, and comparing the security threat risk score with the first risk threshold and the second risk threshold respectively; If the security threat risk score is greater than the second risk threshold, it is classified as a high-risk level, indicating that the ship is in great danger and requires immediate emergency measures; If the security threat risk score is greater than or equal to the first risk threshold and less than or equal to the second risk threshold, it is classified as a medium risk level and the route needs to be adjusted or preventive measures need to be taken; If the security threat risk score value is less than the first risk threshold, it is classified as a low risk level. There is a certain risk, but no immediate measures are required and the ship continues to sail.

8. The fishing vessel safety early warning method based on comprehensive data analysis according to claim 7 is characterized in that: The high-risk level, medium-risk level and low-risk level are converted into corresponding threat priorities, among which the high-risk level corresponds to a high-priority warning, the medium-risk level corresponds to a medium-priority warning, and the low-risk level corresponds to a low-priority warning; the filtering rules include: high-priority warning: directly reminding the crew through sound and light alarms, text messages and voice to ensure that the crew can respond immediately; medium-priority warning: reminding the crew through non-invasive means; low-priority warning: reminding the crew through reports or regular inspections, without the need for immediate processing.

9. The fishing vessel safety early warning method based on comprehensive data analysis according to claim 1 is characterized in that: Continuously learn and analyze crew reactions and decision-making patterns in different situations, and automatically adjust warnings and recommendations, including: Define the action space, the ship's action options, and generate warning information; Design a reward and penalty mechanism to reward crew members if their actions improve navigation efficiency; conversely, punish them if they ignore warnings or take inappropriate actions that lead to risks; Through Q-learning or deep reinforcement learning, initialize the Q table or policy network, record the Q values ​​under different states, and train through continuous exploration and utilization to optimize decision-making behavior.

Images