Marine ecological big data analysis method based on marine environment information

Abstract

The invention discloses a marine ecological big data analysis method based on marine environment information, and relates to the technical field of marine data processing, and the method comprises the steps: constructing a diffusion path model based on marine environment data; generating diffusion path weights based on the spatial relationship between the diffusion path model and the data acquisition points; calculating a first priority value of the data based on the diffusion path weight and obtaining a spatial risk propagation feature; obtaining time dimension data based on diffusion path weight time serialization processing, constructing a space-time combined weight model based on the diffusion path weight and the time dimension data, obtaining a space-time weight matrix based on the space-time combined weight model, and calculating a data second priority value based on the space-time weight matrix to obtain space-time risk propagation features; feature extraction is carried out based on the marine environment data to obtain feature vectors, and diffusion model library subsets are obtained based on the feature vectors; and selecting an optimal diffusion model from the diffusion model library subset according to space-time risk propagation characteristics.

Description

Technical Field

[0001] The present invention relates to the technical field of marine data processing, and in particular, to a method for analyzing marine ecological big data based on marine environmental information. Background Art

[0002] The marine ecosystem is one of the most complex and important ecosystems on the earth, and its health status is closely related to global climate change, biodiversity conservation, sustainable utilization of fishery resources, and the human living environment. In recent years, with the rapid development of marine observation technologies, the ability to obtain marine environmental information and ecological data has grown explosively, forming a veritable marine ecological big data. However, how to effectively utilize this massive, multi-source, and heterogeneous marine environmental information and ecological big data, deeply understand complex marine ecological processes, and achieve accurate ecological assessment, prediction, and management still faces a series of severe technical challenges.

[0003] The prior art CN119046018B discloses a method for processing and analyzing marine ecological data affected by the marine environment, including: Step 1, based on historical collection data, determining data collection points in a marine area; Step 2, performing data analysis on the historically collected marine ecological data to determine a single-body interval, and then calculating the environmental normal value of the marine ecological data according to the data frequency in the single-body interval. Then, combining the real-time collected marine ecological data with the environmental normal value for calculation to obtain the ecological risk value of the data collection point; Step 3: Determining abnormal collection points and marginal collection points and the straight-line distance between them, and combining with the ecological risk value of the abnormal collection points for calculation to obtain a data priority value. Arranging the data priority values in terms of position to obtain a priority sequence, and then processing the real-time collected marine ecological data according to the priority sequence, so as to improve the response efficiency of risk data and enhance the overall risk prevention and control ability. However, it only relies on the straight-line distance between abnormal collection points and marginal collection points to calculate the data priority value and construct the priority sequence, lacking consideration of the dynamic changes in the time dimension. Summary of the Invention

[0004] The present application solves the key problems in the prior art such as over-simplified spatial relationship modeling, lack of time dynamics, insufficient analysis of risk propagation characteristics, poor model adaptability, and passive response lag by providing a method for analyzing marine ecological big data based on marine environmental information, and achieves the technical effects of high-precision, dynamic, proactive prediction, and prevention and control of marine ecological risk management.

[0005] The present application provides a method for analyzing marine ecological big data based on marine environmental information, including: S1: Obtain marine environmental data, construct a diffusion path model based on the marine environmental data; generate diffusion path weights based on the spatial relationship between the diffusion path model and the data collection points; calculate the first priority value of the data based on the diffusion path weights and obtain the spatial risk propagation characteristics; S2: Obtain time - dimension data through time - serialization processing of the diffusion path weights, construct a space - time joint weight model based on the diffusion path weights and the time - dimension data, obtain a space - time weight matrix based on the space - time joint weight model, and calculate the second priority value of the data based on the space - time weight matrix to obtain the space - time risk propagation characteristics; S3: Extract feature vectors based on the marine environmental data, and obtain a subset of the diffusion model library based on the feature vectors; S4: Select the optimal diffusion model from the subset of the diffusion model library according to the space - time risk propagation characteristics, perform space - time weight correction through the optimal diffusion model and the space - time weight matrix, and output the optimal decision.

[0006] Furthermore, the marine environmental data includes: ocean current data, terrain data, pollution source data, and meteorological data; The diffusion path model includes: adopting a finite - element hydrodynamic model to simulate the diffusion trajectory of pollutants under the action of ocean currents and terrain, and obtaining a diffusion path probability map.

[0007] Furthermore, the diffusion path weights include: determining abnormal collection points, predicting the probability and time delay of pollutants reaching adjacent collection points according to the diffusion path model, and taking the prediction results as the diffusion path weights; The spatial risk propagation characteristics include: correcting the data priority value formula according to the diffusion path weights to obtain the first priority value of the data, and analyzing the risk propagation direction, intensity, and key node characteristics from the distribution of the first priority value of the data as the spatial risk propagation characteristics.

[0008] Furthermore, the space - time joint weight model includes: generating diffusion path weights of pollutant diffusion paths through a hydrodynamic model, and at the same time combining time - series analysis to extract tidal cycles and storm events to construct a space - time joint weight model; The space - time weight matrix includes: performing space - time convolution fusion on the diffusion path weights and the time - dimension data, and optimizing the weight coefficients of each dimension through reinforcement learning, and finally forming a space - time weight matrix that changes in real - time with the environment to drive the calculation of the second priority value of the data.

[0009] Furthermore, the space - time risk propagation characteristics include: The first priority value formula of the data is corrected according to the spatio-temporal weight matrix to obtain the second priority value of the data. The propagation direction, risk intensity, and key node characteristics are parsed from the distribution of the second priority value of the data as spatio-temporal risk propagation characteristics. The diffusion modes of steady state, turbulence, period, and compound are identified according to the change law of the second priority value of the data.

[0010] Further, the eigenvector includes: vertical flow velocity variance, tidal phase index, tidal range asymmetry coefficient, sea surface height anomaly, and flow velocity mutation flag. The subset of the diffusion model library includes: internal wave diffusion model, tidal asymmetry model, and circulation mutation model.

[0011] Further, the optimal diffusion model is to evaluate the explanatory ability and prediction fitting degree of each model in the subset of the diffusion model library for the current spatio-temporal risk characteristics, and select the optimal diffusion model.

[0012] Further, the spatio-temporal weight correction includes: Using the diffusion model to predict the probability distribution of the pollutant diffusion path based on the current environmental eigenvector, comparing the prediction result with the diffusion path weight in the spatio-temporal weight matrix to generate a correction gradient, and calculating the third priority value of the data with the corrected weight matrix.

[0013] Further, the optimal decision includes: Establish a collaborative model for the pollutant diffusion speed and the resource deployment speed. This model generates a resource scheduling benchmark value through a dynamic matching algorithm, implements decision correction according to the environmental dynamics, realizes bidirectional coupling through spatio-temporal weight matrix correction, and drives the resource pre-deployment strategy.

[0014] Further, the resource pre-deployment strategy includes: Calculating the virtual priority of the key node based on the resource scheduling benchmark value and the spatio-temporal weight matrix; scheduling the protection resources to the node with the highest virtual priority before the pollutant arrives; dynamically adjusting the spatio-temporal weight correction gradient through reinforcement learning to realize the dynamic environment response.

[0015] One or more technical solutions provided in this application have at least the following technical effects or advantages: By adopting a hydrodynamic model to construct a diffusion path probability map, combining the diffusion path weights to correct and calculate the first priority value of the data, the prediction accuracy of spatial risk propagation characteristics is improved; by using spatio-temporal convolution to fuse the diffusion path weights and time dimension data, generating an optimized second priority value of the data with a spatio-temporal weight matrix, the dynamic analysis of spatio-temporal risk propagation characteristics is realized; by using a random forest classifier to output a subset of the diffusion model library, matching the optimal diffusion model to generate a correction gradient, the dynamic calibration of the spatio-temporal weight matrix is realized; by using a pollutant diffusion-resource deployment collaborative model, calculating virtual priorities based on the spatio-temporal weight matrix and implementing a resource pre-deployment strategy, the automated decision-making output is realized. Brief Description of the Drawings

[0016] Figure 1 This is a flowchart of a method for analyzing marine ecological big data based on marine environmental information in an embodiment of the present invention. Detailed Embodiments

[0017] To facilitate the understanding of the present invention, the present application will be described more comprehensively with reference to the relevant drawings; the drawings show preferred embodiments of the present invention. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein; on the contrary, these embodiments are provided to make the disclosure of the present invention more thorough and comprehensive.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs; the terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention; the term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0019] Example 1: As Figure 1 shown, a method for analyzing marine ecological big data based on marine environmental information.

[0020] S1: Obtain marine environmental data, construct a diffusion path model based on the marine environmental data; generate diffusion path weights based on the spatial relationship between the diffusion path model and the data collection points; calculate the first priority value of the data based on the diffusion path weights and obtain the spatial risk propagation characteristics; The marine environmental data includes: ocean current data, terrain data, pollution source data, and meteorological data; Specifically, real-time or historical ocean current speed and direction are obtained through buoy monitoring and satellite remote sensing as ocean current data; bathymetric elevation maps are obtained through a multibeam sounding system and a public terrain database, and global seabed elevation data is integrated as terrain data; pollutant types, release amounts, and densities are obtained through on-site monitoring platforms and remote sensing inversion as pollution source data; real-time wind speed and direction are monitored and storm event paths are tracked through meteorological buoys and meteorological satellites as meteorological data.

[0021] The diffusion path model includes: using a finite element hydrodynamic model to simulate the diffusion trajectory of pollutants under the action of ocean currents and terrain, and obtaining a diffusion path probability map.

[0022] Specifically, a diffusion path model is constructed using hydrodynamic equations, and the formula is:

[0023] where is the change rate of pollutant concentration over time, is the advective transport of pollutants driven by ocean currents, is the mixing of pollutants caused by turbulent diffusion, is the release intensity of the pollution source, is the pollutant concentration, is the ocean current velocity field, is the diffusion coefficient.

[0024] The bathymetric elevation map is converted into a terrain boundary constraint function, and a flow velocity attenuation coefficient is added to the ridge area:

[0025] where is the terrain resistance parameter, is the relative height of the ridge, is the flow velocity attenuation coefficient; A diffusion suppression coefficient is added to the shoal area:

[0026] where is the actual water depth, is the critical shallow water threshold, is the suppression intensity coefficient, is the diffusion suppression coefficient.

[0027] The flow velocity attenuation coefficient and the diffusion suppression coefficient of the terrain boundary constraint function are embedded into the hydrodynamic equation, and the modified formula is:

[0028] where is , For ..

[0029] Use Monte Carlo simulation to generate a diffusion path probability map, and correct the particle movement trajectory according to the flow velocity decay coefficient and the diffusion inhibition coefficient.

[0030] The diffusion path weight includes: determining abnormal sampling points, predicting the probability and time delay of pollutant arrival at adjacent sampling points according to the diffusion path model, and using the prediction results as the diffusion path weight; Specifically, taking the average value of the marine ecological data as the environmental normal value, the abnormal sampling point refers to the area with marine environmental risks or problems calculated from the real-time marine ecological data and the environmental normal value. For each pollutant sampling point, perform Monte Carlo simulation based on the diffusion path model; taking the pollution source data, ocean current velocity field and the diffusion coefficient as input data, obtaining the probability and time delay of reaching adjacent sampling points, normalizing the obtained probability and time delay so that they are in the same dimension [0,1], and performing calculations to obtain the diffusion path weight:

[0031] wherein, is the diffusion path weight.

[0032] The spatial risk propagation characteristics include: correcting the data priority value formula according to the diffusion path weight to obtain the first data priority value, and analyzing the risk propagation direction, intensity, and key node characteristics from the data priority value distribution as the spatial risk propagation characteristics.

[0033] Specifically, optimize the calculation of the data priority value based on the diffusion path weight:

[0034] wherein, is the first data priority value, is the ecological risk value, is the diffusion path weight.

[0035] Calculate the spatial gradient vector of the value of each sampling point. The high gradient direction indicates the main flow direction of pollutant diffusion. Identify the risk propagation path through the spatial gradient distribution of the first data priority value as the risk propagation direction feature; through quantify the risk level by the numerical size of the value, the higher the value, the stronger the probability and urgency of being affected by pollutants at this location, as the intensity feature; identify the extreme points in the value distribution as the key node characteristics, wherein, The value local peak points are the pollution source nodes, The regions with sudden value increases are the sensitive nodes, and the transfer nodes are located at the intersections of multiple diffusion paths, The value change rate is significant.

[0036] The technical solutions in the embodiments of the present application at least have the following technical effects or advantages: This application quantifies the impact of complex marine environments on pollutant diffusion through hydrodynamic equations; Monte Carlo simulation generates probabilistic diffusion paths to improve prediction accuracy. Define the diffusion path weights, fuse the pollutant arrival probability and time delay, and dynamically correct the data first priority value to make the risk assessment more in line with the actual propagation process; analyze the direction, intensity, and key nodes from the distribution of the data first priority value to form a structured spatial risk profile to guide precise prevention and control.

[0037] Embodiment 2: In Embodiment 1, the impact of the marine environment on pollutant diffusion is quantified through a hydrodynamic model, probabilistic diffusion paths are generated in combination with Monte Carlo simulation, and spatial risk assessment is corrected based on dynamic weights to achieve precise analysis of the risk propagation direction, intensity, and key nodes. However, only the diffusion path weights at the current moment are used to calculate the data first priority value, without considering the time-varying characteristics of the marine environment, and the data first priority value does not incorporate time series data into the weight calculation, making it impossible to predict the evolution trend of the risk over time. This embodiment further improves Embodiment 1.

[0038] S2: Obtain time-dimensional data based on the time serialization process of the diffusion path weights, construct a spatio-temporal joint weight model based on the diffusion path weights and the time-dimensional data, obtain a spatio-temporal weight matrix based on the spatio-temporal joint weight model, and calculate the data second priority value based on the spatio-temporal weight matrix to obtain the spatio-temporal risk propagation characteristics; Specifically, based on the dynamic characteristics of the marine environment, the time serialization process generates time-dimensional data by tracking wind speed and tidal data at the minute level in real time. The time-dimensional data includes a time correction coefficient and a terrain obstacle coefficient; the time correction coefficient linearly increases with the increase in wind speed to capture the acceleration effect of storms on pollutant diffusion. Its formula is:

[0039] Where, is the time correction coefficient, 1 is the basic diffusion coefficient benchmark, 0.1 is the diffusion acceleration slope, indicating that for every 1 m / s increase in wind speed, the pollutant diffusion speed increases by 10%, F is the wind speed input in real-time meteorology, 10 is the critical value of marine hydrodynamics, and when the wind speed > 10 m / s, the sea surface turbulence significantly increases, is the effective wind speed increment, and only the wind speed part exceeding the threshold (10 m / s) is calculated to avoid ineffective disturbances in the low wind speed section.

[0040] The terrain obstruction coefficient increases exponentially with the decrease of the tide level, quantifying the enhanced diffusion resistance caused by the exposure of the seabed terrain during the low tide period. Its formula is: , where, is the terrain obstruction coefficient, is the terrain complexity index, with a value range greater than 0. For example, it is 0 in the flat area and 2 in the coral reef area. k is the terrain sensitivity coefficient, and the default value is 0.8.

[0041] The spatio-temporal joint weight model includes: generating the diffusion path weight of the pollutant diffusion path through a hydrodynamic model, and at the same time combining time series analysis to extract the tidal cycle and storm events to construct the spatio-temporal joint weight model; Specifically, adding the time dimension data to the original diffusion path weight to establish the spatio-temporal joint weight model: , where, is the spatio-temporal joint weight.

[0042] The spatio-temporal weight matrix includes: performing spatio-temporal convolution fusion on the diffusion path weight and the time dimension data, and optimizing the weight coefficients of each dimension through reinforcement learning, and finally forming a spatio-temporal weight matrix that changes in real time with the environment to drive the calculation of the second priority value of the data.

[0043] Specifically, fusing the diffusion path weight and the time dimension data into the spatio-temporal weight, and the fusion formula is:

[0044] where, is the spatio-temporal weight, representing the final risk weight of the position (x, y) at time t, is the spatial diffusion weight, is the spatio-temporal convolution operator, is the wind speed influence weight, is the terrain influence weight, and The sum of the values is 1, is the longitude and latitude position of the ocean data collection point, and t is the time stamp of the data collection.

[0045] Taking , ,..., as the spatio-temporal weight matrix , further optimizing the first priority value of the data: , where, is the optimized data priority value, is a dynamic adjustment coefficient, and its value range is [1, 2].

[0046] The spatio-temporal risk propagation characteristics include: correcting the data first priority value formula according to the spatio-temporal weight matrix to obtain the data second priority value, and analyzing the propagation direction, risk intensity and key node characteristics from the data second priority value distribution as the spatio-temporal risk propagation characteristics; identifying steady state, turbulence, periodic and composite diffusion patterns according to the change law of the data second priority value.

[0047] Specifically, through the data second priority value spatial gradient distribution to identify the main direction of pollutant diffusion as the propagation direction feature; through the value size to quantify the risk level and urgency as the risk intensity feature, according to the value is divided into three risk levels, <5 is low risk, 5 ≤ <8 is medium risk, ≥8 is high risk; the pollution source node ( local maximum value), sensitive node ( sudden increase rate > 20% / hour), transfer node ( change rate > 1.5 / km) are used as key node characteristics.

[0048] By analyzing the spatial distribution form and time dimension fluctuation law of, matching steady state, turbulence, periodic and composite diffusion patterns; after identifying the pattern, dynamically adjust the resource pre-deployment strategy, and reverse optimize the parameters of the spatio-temporal weight matrix to improve the subsequent prediction accuracy.

[0049] The technical solutions in the embodiments of the present application at least have the following technical effects or advantages: The present application generates time dimension data in real time by tracking wind speed and tidal data at the minute level, constructs a space-time joint weight model by combining the diffusion path weight; performs spatio-temporal convolution fusion on the diffusion path weight and time dimension data, and optimizes the weight coefficients of each dimension through reinforcement learning to form a spatio-temporal weight matrix that changes in real time with the environment; optimizes the calculation of the data second priority value based on the spatio-temporal weight matrix to obtain spatio-temporal risk propagation characteristics, analyzes the propagation direction, risk intensity and key node characteristics, and identifies steady state, turbulence, periodic and composite diffusion patterns according to the change law of the data second priority value, realizing the improvement of dynamic prediction accuracy. Embodiment 3: Embodiment 2 constructs a dynamic weight matrix through spatio-temporal convolution fusion , realizing the prediction of risk spatio-temporal evolution. However, the weight matrix depends on historical data training, cannot adapt to sudden environmental changes, and is not associated with physical mechanisms, resulting in prediction deviation and decision lag. This embodiment further improves Embodiment 2.

[0050] S3: Extract features based on marine environmental data to obtain feature vectors, and obtain a subset of the diffusion model library based on the feature vectors; The characteristic vectors include: vertical velocity variance, tidal phase index, tidal range asymmetry coefficient, sea level height anomaly and velocity mutation sign; Specifically, the vertical velocity variance is a statistic that characterizes the vertical mixing intensity of the water body and reflects the diffusion capacity of pollutants in the vertical direction; the tidal phase index quantifies the time position in the tidal cycle and is related to the periodic influence of tidal dynamics on diffusion; the tidal range asymmetry coefficient characterizes the nonlinear difference in the duration and intensity of high and low tides and drives the net transport of pollutants; the sea surface height anomaly is the deviation of the sea surface height inverted by satellite remote sensing from the climatological mean; the velocity mutation mark is used to identify the instantaneous sudden change in ocean current velocity.

[0051] S4: Select the optimal diffusion model from the diffusion model library subset based on the spatiotemporal risk propagation characteristics, perform spatiotemporal weight correction through the optimal diffusion model and the spatiotemporal weight matrix, and output the optimal decision.

[0052] The diffusion model library subset includes: internal wave diffusion model, tidal asymmetry model and circulation mutation model.

[0053] Specifically, the diffusion model library subset is a feature vector that is dynamically filtered through a random forest classifier to identify three types of pollutant diffusion physical models. The output is the applicable probability of the three diffusion models. When the probability of a model exceeds a threshold (>0.7), the model is activated and added to the subset to modify the spatiotemporal weight matrix and optimize decision making. If the probabilities of all models are below the threshold, the basic fluid dynamics model is used by default.

[0054] The optimal diffusion model is to evaluate the explanatory power and predictive fit of each model in the diffusion model library subset for the current spatiotemporal risk characteristics, and select the optimal diffusion model.

[0055] Specifically, when the vertical flow velocity variance is significant (>0.05m² / s²) and the sea level height anomaly is positive (>0.1m), the internal wave diffusion model is adopted; when the absolute value of the tidal range asymmetry coefficient is greater than 0.3, the tidal asymmetry model is adopted; when the flow velocity mutation flag is 1, the circulation mutation model is adopted; the circulation mutation model has the highest priority because it involves emergencies, followed by the internal wave diffusion model and the tidal asymmetry model, to ensure that the system responds to the most urgent risks first.

[0056] The spatiotemporal weight correction includes: using a diffusion model to predict the probability distribution of pollutant diffusion paths based on the current environmental characteristic vector, comparing the prediction results with the diffusion path weights in the spatiotemporal weight matrix, generating a correction gradient, and using the corrected weight matrix to calculate the third priority value of the data.

[0057] Specifically, using the optimal diffusion model that matches the current environment, a probability distribution map of pollutant diffusion paths is predicted based on real-time feature vectors, and the formula is:

[0058] Among them, is the arrival probability of pollutants predicted by the model, is the diffusion model, is the feature vector.

[0059] Compare the model prediction result with the diffusion path weight in the spatio-temporal weight matrix : Perform a difference comparison:

[0060] Among them, reflects the deviation between the model prediction and the actual weight, and generates a correction gradient.

[0061] Update the diffusion path weight through the correction gradient:

[0062] Among them, is the updated diffusion path weight, is the learning rate, with a value range of [0.05, 0.2], which controls the correction amplitude. Substitute the updated back into the spatio-temporal weight matrix to generate a corrected matrix .

[0063] Based on the modified calculate the third priority value of the data:

[0064] Among them, is the third priority value of the data, which integrates the prediction results of the physical model, making the analyzed spatio-temporal risk propagation characteristics closer to the real ocean dynamics process.

[0065] The technical solutions in the embodiments of the present application described above have at least the following technical effects or advantages: In the present application, a subset of the diffusion model library including the internal wave diffusion model, the tidal asymmetry model, and the circulation mutation model is generated through a random forest classifier, and the optimal diffusion model is dynamically matched based on physical thresholds, where the circulation mutation model has the highest priority; Using the optimal diffusion model to predict the probability of pollutant diffusion paths, comparing with the diffusion path weights in the spatio-temporal weight matrix to generate a correction gradient, and through the learning rate Update the weights, reconstruct the spatio-temporal weight matrix and calculate the third priority value of the data; through the coupling of the physical model and real-time data, significantly improve the risk prediction accuracy and decision-making timeliness in the sudden marine environment.

[0066] Example 4: In Example 3, a correction gradient is generated through the difference between the model prediction result and the spatio-temporal weight matrix, and the diffusion path weight is updated in real time through a dynamic learning rate, enabling the risk prediction to adapt to sudden environmental changes and reducing the prediction error in the scenario of historical data invalidation by 30%-50%. However, its optimization only stays at the level of risk feature analysis, and no association is established between the prediction result and the resource scheduling decision, resulting in the inability to convert high-precision prediction into proactive prevention and control actions. This example further improves Example 3.

[0067] Step S4 also includes: outputting the optimal decision and providing automated technical support.

[0068] The optimal decision includes: establishing a collaborative model between the pollutant diffusion speed and the resource deployment speed. This model generates a resource scheduling benchmark value through a dynamic matching algorithm, and implements decision correction according to environmental dynamics, achieving two-way coupling through spatio-temporal weight matrix correction to drive the resource pre-deployment strategy.

[0069] Specifically, calculate the resource scheduling benchmark value based on the matching relationship between the pollutant diffusion speed and the resource deployment speed:

[0070] Among them, is the resource scheduling benchmark value, is the diffusion speed of spatio-temporal risk propagation characteristics dynamically predicted, is the deployment speed of the resource, is the environmental dynamics adjustment coefficient; and are both real-time outputs of the spatio-temporal weight matrix and is from the preset resource scheduling database.

[0071] The resource pre-deployment strategy includes: calculating the virtual priority of key nodes based on the resource scheduling benchmark value and the spatio-temporal weight matrix; scheduling the protection resources to the node with the highest virtual priority before the pollutant arrives; dynamically adjusting the spatio-temporal weight correction gradient through reinforcement learning to achieve dynamic environment response.

[0072] Specifically, use the resource scheduling benchmark value and the spatio-temporal weight matrix as input data for virtual priority calculation:

[0073] Among them, is the virtual priority value, is the node sensitivity coefficient, with the key node being 1.5 and the marginal node being 0.8. Extract the top 10% of the nodes as key nodes, and deploy resources to the nodes with the highest values before the pollutants arrive.

[0074] Update the learning rate through the feedback of the resource deployment effect :

[0075]

[0076] Among them, is the learning rate, is the actual pollution reduction rate, sourced from the detection data of the deployed sensor network, is the predicted reduction rate, sourced from the simulation prediction results of the optimal diffusion model.

[0077] Feedback the change of the local environment through resource deployment to the spatio-temporal weight matrix , triggering a new round of weight correction. An intelligent closed-loop of environmental dynamic perception - model real-time iteration - decision rolling optimization is realized, significantly improving the initiative and accuracy of marine pollution risk prevention and control.

[0078] The technical solutions in the embodiments of the present application described above have at least the following technical effects or advantages: The present application establishes a collaborative model for the pollutant diffusion speed and the resource deployment speed, generates a resource scheduling benchmark value, and calculates the virtual priority value of the key node based on the corrected data third priority value and the resource scheduling benchmark value to drive the resource pre-deployment strategy, and schedules the protection resources to the node with the highest virtual priority before the pollutants arrive; at the same time, dynamically optimize the system through the reinforcement learning mechanism: use the feedback comparison of the actual pollution reduction rate and the predicted reduction rate to adaptively adjust the learning rate of the spatio-temporal weight correction, form a closed-loop of environmental dynamic perception - model prediction - resource pre-deployment - effect feedback - weight iteration, transform the high-precision risk prediction into active prevention and control actions, shorten the resource response time by more than 50%, and increase the prevention and control coverage rate of the key nodes to 95%, significantly enhancing the adaptability and decision-making accuracy of the system to sudden marine environmental events.

[0079] The above is only the preferred embodiment of the present invention and is not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An ocean ecological big data analysis method based on ocean environmental information, characterized in that Including: S1: Obtain marine environmental data and construct a diffusion path model based on the marine environmental data; Generate diffusion path weights based on the spatial relationship between the diffusion path model and the data collection points; Calculate the first priority value of the data based on the diffusion path weights and obtain the spatial risk propagation characteristics; S2: Obtain time-dimensional data through time serialization processing of the diffusion path weights, construct a spatio-temporal joint weight model based on the diffusion path weights and the time-dimensional data, obtain a spatio-temporal weight matrix based on the spatio-temporal joint weight model, and calculate the second priority value of the data based on the spatio-temporal weight matrix to obtain spatio-temporal risk propagation characteristics; S3: Extract feature vectors based on the marine environmental data and obtain a subset of the diffusion model library based on the feature vectors; S4: Select the optimal diffusion model from the subset of the diffusion model library according to the spatio-temporal risk propagation characteristics, perform spatio-temporal weight correction through the optimal diffusion model and the spatio-temporal weight matrix, and output the optimal decision.

2. The marine ecological big data analysis method based on marine environmental information according to claim 1, characterized in that The marine environmental data includes: ocean current data, terrain data, pollution source data, and meteorological data; The diffusion path model includes: adopting a finite element hydrodynamic model to simulate the diffusion trajectory of pollutants under the action of ocean currents and terrain, and obtaining a diffusion path probability map.

3. The marine ecological big data analysis method based on marine environmental information according to claim 1, wherein The diffusion path weights include: determining abnormal collection points, predicting the probability and time delay of pollutants reaching adjacent collection points according to the diffusion path model, and taking the prediction results as the diffusion path weights; The spatial risk propagation characteristics include: correcting the data priority value formula according to the diffusion path weights to obtain the first priority value of the data, and analyzing the risk propagation direction, intensity, and key node characteristics from the distribution of the first priority value of the data as the spatial risk propagation characteristics.

4. The marine ecological big data analysis method based on marine environmental information according to claim 1, characterized in that The spatio-temporal joint weight model includes: generating diffusion path weights of pollutant diffusion paths through a hydrodynamic model, and simultaneously combining time series analysis to extract tidal cycles and storm events to construct a spatio-temporal joint weight model; The spatio-temporal weight matrix includes: performing spatio-temporal convolution fusion on the diffusion path weights and the time-dimensional data, and optimizing the weight coefficients of each dimension through reinforcement learning to finally form a spatio-temporal weight matrix that changes in real time with the environment, driving the calculation of the second priority value of the data.

5. The method for analyzing marine ecological big data based on marine environmental information according to claim 1, wherein The spatio-temporal risk propagation characteristics include: Correcting the data first priority value formula according to the spatio-temporal weight matrix to obtain the second priority value of the data, and analyzing the propagation direction, risk intensity, and key node characteristics from the distribution of the second priority value of the data as the spatio-temporal risk propagation characteristics; identifying steady state, turbulent, periodic, and composite diffusion patterns according to the change law of the second priority value of the data.

6. The marine ecological big data analysis method based on marine environmental information according to claim 1, wherein, The feature vectors include: vertical velocity variance, tidal phase index, tidal range asymmetry coefficient, sea surface height anomaly, and flow velocity mutation flag; The subset of the diffusion model library includes: internal wave diffusion model, tidal asymmetry model, and circulation mutation model.

7. An ocean ecological big data analysis method based on ocean environment information according to claim 1, characterized in that The optimal diffusion model is to evaluate the explanatory ability and prediction fitting degree of each model in the subset of the diffusion model library for the current spatio-temporal risk characteristics, and select the optimal diffusion model.

8. The marine ecological big data analysis method based on marine environmental information according to claim 1, wherein, The spatio-temporal weight correction includes: Use a diffusion model to predict the probability distribution of the pollutant diffusion path based on the current environmental feature vector, compare the prediction result with the diffusion path weights in the spatio-temporal weight matrix to generate a correction gradient, and calculate the third priority value of the data using the corrected weight matrix.

9. An ocean ecological big data analysis method based on ocean environment information according to claim 1, characterized in that, The optimal decision includes: Establish a collaborative model for the pollutant diffusion speed and the resource deployment speed. This model generates a resource scheduling benchmark value through a dynamic matching algorithm, implements decision correction according to environmental dynamics, realizes two-way coupling through spatio-temporal weight matrix correction, and drives the resource pre-deployment strategy.

10. The marine ecological big data analysis method based on marine environmental information according to claim 9, wherein, The resource pre-deployment strategy includes: Calculate the virtual priority of the key nodes based on the resource scheduling benchmark value and the spatio-temporal weight matrix; schedule the protection resources to the node with the highest virtual priority before the pollutant arrives; dynamically adjust the spatio-temporal weight correction gradient through reinforcement learning to achieve dynamic environment response.

Images