Crude oil spectrum-based marine oil spill monitoring method and system

Abstract

The invention relates to the technical field of marine environment monitoring, in particular to a marine oil spill monitoring method and system based on a crude oil spectrum. Comprising the following steps: S1, identifying N marine oil spill positions by using a fuzzy reasoning model, and distributing monitoring points for the N marine oil spill positions to form a first monitoring sequence; s2, K important oil spill positions and monitoring points corresponding to the important oil spill positions are optimized through a genetic algorithm, a second monitoring sequence is formed, and K is smaller than N and is a positive integer larger than or equal to 1; s3, calculating the time when the crude oil reaches the sea surface and marking a path; a high-resolution spectrometer is used for collecting spectral data in visible light, near-infrared and middle-infrared wave bands, and features are screened after preprocessing. According to the method, the ocean oil spill monitoring path is optimized through fuzzy reasoning and a genetic algorithm, spectral data are predicted in combination with a Transform model, the path and the model are dynamically corrected, and the monitoring precision and the response efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of marine environment monitoring, and in particular to a marine oil spill monitoring method and system based on crude oil spectrum. Background Art

[0002] Crude oil is a vital component of the global energy system and plays an indispensable role in economic development, industrial production, transportation, and daily life. However, marine oil spills can cause severe damage to marine ecosystems, pollute water quality, poison marine life, affect fisheries and tourism, and lead to economic losses and ecological disasters. Existing marine oil spill monitoring methods, such as satellite remote sensing, drone monitoring, and surface vessel inspections, while offering advantages such as wide coverage and strong real-time performance, also suffer from high costs, significant weather impacts, and limited resolution. To improve the accuracy and reliability of monitoring, spectral technology plays an important role in marine oil spill monitoring, providing high-precision oil identification and classification, while also enabling rapid detection of oil composition and concentration, thereby improving the efficiency of emergency response. However, despite the introduction of spectral technology, existing marine oil spill monitoring methods still have deficiencies in resolution, false alarm rate, and the ability to distinguish oil from other substances, resulting in a need for further improvement in the accuracy and reliability of monitoring results. Summary of the Invention

[0003] In order to overcome the shortcoming of low accuracy in marine oil spill monitoring, the present invention provides a marine oil spill monitoring method and system based on crude oil spectrum.

[0004] The technical implementation scheme of the present invention is: a method for monitoring marine oil spills based on crude oil spectrum, comprising the following steps: S1: Using the fuzzy inference model to identify N marine oil spill locations, and assigning monitoring points to the N marine oil spill locations to form the first monitoring sequence; S2: Optimizing K important oil spill locations and the monitoring points corresponding to the important oil spill locations by genetic algorithm to form a second monitoring sequence, where K is less than N and is a positive integer greater than or equal to 1; S3: Calculate the time it takes for the crude oil to reach the sea surface and mark the path; use a high-resolution spectrometer to collect spectral data in the visible, near-infrared, and mid-infrared bands, pre-process, and filter features; use a Transformer prediction model to predict the spectral data, and fuse the first monitoring sequence and the second monitoring sequence to form a first fused monitoring path; S4: Considering the influence of energy loss and solar spectrum, the first fusion monitoring path is adjusted to form a second fusion monitoring path; the actual value of the crude oil spectrum data monitored is compared with the predicted value of the crude oil spectrum data, the spectrum difference data is obtained, and the Transformer prediction model and the second fusion monitoring path are corrected; S5: Generate monitoring priority markers for oil spill locations based on spectral difference data and identify key monitoring areas; continuously iterate and optimize the monitoring path and Transformer prediction model through a closed-loop feedback mechanism.

[0005] Preferably, the method of identifying N marine oil spill locations using a fuzzy inference model and allocating monitoring points to the N marine oil spill locations to form a first monitoring sequence includes: Collect and pre-process data on offshore platforms, submarine pipelines, and marine environments, define fuzzy set membership functions, and establish a fuzzy rule base; Perform fuzzy reasoning, including data fuzzification, rule matching and reasoning results, and finally defuzzify and convert the results into specific ocean oil spill locations; Taking the sea surface projection point of the marine oil spill location as the monitoring location center, the farthest sea surface monitoring range is allocated to N marine monitoring locations, and then the N marine monitoring locations and the farthest sea surface monitoring ranges allocated to the N marine monitoring locations constitute the first marine oil spill monitoring sequence.

[0006] Preferably, the optimization of K important oil spill locations and the monitoring points corresponding to the important oil spill locations by a genetic algorithm to form a second monitoring sequence, where K is less than N and is a positive integer greater than or equal to 1, comprises: Generate an initial set of potential ocean oil spill locations, evaluate the fitness of each location, and select locations with high fitness to enter the next generation; Perform crossover and mutation operations to generate a new set of ocean oil spill locations; Merge the parent and offspring generations, select the ocean oil spill locations with high fitness to enter the next generation, until the maximum fitness threshold is reached, and the optimal K ocean oil spill locations are selected; K ocean monitoring locations are allocated to the optimal K ocean oil spill locations. Each monitoring location is centered on the sea surface projection point of the oil spill point and is allocated the farthest sea surface monitoring range to form the second ocean oil spill monitoring sequence.

[0007] Preferably, the step of calculating the time it takes for the crude oil to reach the sea surface and marking the path includes: Based on the depth data of the oil spill point, the ocean water body is divided into the current-dominated layer, the wind-dominated layer and the water temperature-dominated layer; When a certain depth layer is dominated by a single factor, the depth value of the layer is divided by the corresponding rising rate coefficient to obtain the rising time of the crude oil in the layer; Based on the superposition results of the three layers, the most influential factors, layer depth and layer time of the crude oil floating path are marked; The path mark sequence is used as the monitoring point attribute, and exclusive sequences are assigned to the first monitoring position and the second monitoring position.

[0008] Preferably, the method of using the path mark sequence as the monitoring point attribute and assigning exclusive sequences to the first monitoring position and the second monitoring position includes: The layered time consumption data in the exclusive sequence is used as the operation object to extract the crude oil rising time consumption in the ocean current dominant layer, the wind dominant layer and the water temperature dominant layer; According to the sum of the three-layer time consumption, all oil spill locations are sorted from longest to shortest according to the total time consumption to obtain the total time consumption sorting result; The time consumption of each dominant layer is independently sorted from long to short to obtain the layered time consumption sorting result.

[0009] Preferably, the predicting spectral data using the Transformer prediction model includes: extracting a path marker feature vector according to the first path marker sequence and the second path marker sequence; extracting spectral feature labels based on the crude oil spectral data collected by the high-resolution spectrometer; Constructing a Transformer prediction model, taking the path label feature vector as input and the spectral feature label as output, and training the model; Using the trained Transformer prediction model, the new path label feature vector is input and the corresponding crude oil spectral data prediction value is output.

[0010] Preferably, the fusing of the first monitoring sequence and the second monitoring sequence to form a first fused monitoring path includes: Based on the predicted values ​​of the spectral data, the monitoring priority is determined by sorting the values ​​from large to small; When it is necessary to integrate multiple monitoring sequence paths, the total time consumption sorting results and the layered time consumption sorting results are obtained; the oil spill location is selected based on the principle of priority based on the longest total time consumption; if the total time consumption is the same, the time consumption is compared step by step in the order of ocean current dominant layer, wind dominant layer, and water temperature dominant layer.

[0011] Preferably, the first fusion monitoring path is adjusted to form a second fusion monitoring path by taking into account the influence of energy loss and solar spectrum; the actual value of the crude oil spectrum data monitored is compared with the predicted value of the crude oil spectrum data, the spectrum difference data is obtained, and the Transformer prediction model and the second fusion monitoring path are corrected, including: Taking the first fused monitoring path as a benchmark, evaluate the negative impact of energy loss and solar spectrum in each monitoring point path on field monitoring to obtain a negative impact value; If there is a multipath negative impact, sort it by the negative impact value from large to small; The monitoring point access order is replanned according to the sorting results to form a second fusion monitoring path.

[0012] Preferably, generating a monitoring priority mark for the oil spill location based on the spectral difference data and determining a key monitoring area includes: Taking the spectrum difference data as input, recording the number of corrections of the crude oil spectrum data prediction value and the absolute value of the spectrum difference data difference; The number of corrections of the crude oil spectrum data prediction value is used as a first weight, and the absolute value of the spectrum difference data difference is used as a second weight; According to the sum of the two weights, the monitoring locations are sorted from largest to smallest according to the total weight value; The location with the highest total weight is selected as the priority monitoring point; The monitoring path, the first weight and the second weight are updated regularly using a closed-loop mechanism.

[0013] A marine oil spill monitoring system based on crude oil spectrum, comprising: Fuzzy reasoning and genetic algorithm module: used to collect and preprocess ocean-related data, determine the location of the oil spill through fuzzy reasoning models, and use genetic algorithms to optimize the location of monitoring points; Path analysis module: used to calculate the time it takes for crude oil to reach the sea surface based on depth data, mark the oil spill path, and assign path marker sequences according to monitoring points; Spectral data acquisition and processing module: used to collect ocean surface spectral data using a high-resolution spectrometer, screen key features after preprocessing, and predict spectral data based on the Transformer model; Fusion monitoring path module: used to combine the spectral data prediction values ​​of different monitoring sequences to form a comprehensive monitoring path connecting all oil spill locations; Energy loss and solar spectrum impact module: used to evaluate the negative impact of energy loss and solar spectrum, and adjust the monitoring path to optimize the monitoring effect; Monitoring optimization and feedback module: used to compare the difference between predicted values ​​and actual monitored values, correct the Transformer prediction model and monitoring path, mark key monitoring locations, and continuously optimize through a closed-loop mechanism.

[0014] Beneficial effects: The present invention uses a fuzzy inference model to identify the location of marine oil spills and assign monitoring points to form a first monitoring sequence; uses a genetic algorithm to optimize important monitoring points based on oil spill risk and environmental sensitivity to generate a second monitoring sequence; calculates the crude oil floating time and marks the layered path characteristics to ensure comprehensive coverage of the oil spill path; collects spectral data through a high-resolution spectrometer, trains a Transformer prediction model based on the path marking feature vector, and outputs spectral prediction values; fuses monitoring sequences to form an optimized path to reduce energy loss and solar spectrum interference; dynamically corrects the model and path based on spectral difference, and uses closed-loop feedback to improve monitoring accuracy and response efficiency, thereby achieving efficient and accurate monitoring of marine oil spills. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a flow chart of the marine oil spill monitoring method based on crude oil spectrum of the present invention; Figure 2 The diagram is a structural diagram of the marine oil spill monitoring system based on crude oil spectrum of the present invention. DETAILED DESCRIPTION

[0016] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0017] Example 1: A method for monitoring marine oil spills based on crude oil spectra, such as Figure 1 As shown, the following steps are included: S1: Using the fuzzy inference model to identify N marine oil spill locations, and assigning monitoring points to the N marine oil spill locations to form the first monitoring sequence; S2: Optimizing K important oil spill locations and the monitoring points corresponding to the important oil spill locations by genetic algorithm to form a second monitoring sequence, where K is less than N and is a positive integer greater than or equal to 1; S3: Calculate the time it takes for the crude oil to reach the sea surface and mark the path; use a high-resolution spectrometer to collect spectral data in the visible, near-infrared, and mid-infrared bands, pre-process, and filter features; use a Transformer prediction model to predict the spectral data, and fuse the first monitoring sequence and the second monitoring sequence to form a first fused monitoring path; S4: Considering the influence of energy loss and solar spectrum, the first fusion monitoring path is adjusted to form a second fusion monitoring path; the actual value of the crude oil spectrum data monitored is compared with the predicted value of the crude oil spectrum data, the spectrum difference data is obtained, and the Transformer prediction model and the second fusion monitoring path are corrected; S5: Generate monitoring priority markers for oil spill locations based on spectral difference data and identify key monitoring areas; continuously iterate and optimize the monitoring path and Transformer prediction model through a closed-loop feedback mechanism.

[0018] The fuzzy inference model is used to identify N marine oil spill locations, and monitoring points are allocated to the N marine oil spill locations to form a first monitoring sequence, including: Collect and pre-process data on offshore platforms, submarine pipelines, and marine environments, define fuzzy set membership functions, and establish a fuzzy rule base; Perform fuzzy reasoning, including data fuzzification, rule matching and reasoning results, and finally defuzzify and convert the results into specific ocean oil spill locations; Taking the sea surface projection point of the marine oil spill location as the monitoring location center, the farthest sea surface monitoring range is allocated to N marine monitoring locations, and then the N marine monitoring locations and the farthest sea surface monitoring ranges allocated to the N marine monitoring locations constitute the first marine oil spill monitoring sequence.

[0019] Further explanation is, fuzzy reasoning model: a method based on fuzzy logic to deal with uncertainty and imprecise data, and to infer the location of marine oil spills by defining membership functions and rule bases; Assuming that the coordinates of the oil spill location are (10, 20), the farthest monitoring position on the sea surface with a radius of 5 kilometers is allocated with this coordinate as the center; the specific monitoring positions include (10, 25), (15, 20), (5, 20) and (10, 15), which constitute the first marine oil spill monitoring sequence: [(10, 25, 5km), (15, 20, 5km), (5, 20, 5km), (10, 15, 5km)], and the sequence includes the center coordinates and the farthest monitoring range of each monitoring position.

[0020] K important oil spill locations and the monitoring points corresponding to the important oil spill locations are optimized by a genetic algorithm to form a second monitoring sequence, where K is less than N and is a positive integer greater than or equal to 1, including: Generate an initial set of potential ocean oil spill locations, evaluate the fitness of each location, and select locations with high fitness to enter the next generation; Perform crossover and mutation operations to generate a new set of ocean oil spill locations; Merge the parent and offspring generations, select the ocean oil spill locations with high fitness to enter the next generation, until the maximum fitness threshold is reached, and the optimal K ocean oil spill locations are selected; K ocean monitoring locations are allocated to the optimal K ocean oil spill locations. Each monitoring location is centered on the sea surface projection point of the oil spill point and is allocated the farthest sea surface monitoring range to form the second ocean oil spill monitoring sequence.

[0021] Further explanation is as follows: Genetic algorithm: an optimization technology that imitates the natural selection process and finds the optimal solution through selection, crossover and mutation operations to optimize the location of monitoring points; Input data preparation: The first monitoring sequence generated by the fuzzy inference model is used as the initial input data. The first monitoring sequence contains N marine oil spill locations and their monitoring point coordinates, and the farthest monitoring range of the sea surface; Fitness function definition: With oil spill risk probability, environmental sensitivity, and monitoring cost as optimization objectives, the fitness function is constructed: ;in: The probability of oil spill (calculated based on historical leakage data and pipeline corrosion); is the environmental sensitivity (quantitatively graded according to the distribution of nearby coral reefs and fishing grounds); is the monitoring cost (positively correlated with the offshore distance and water depth); , , is the weight coefficient (default =0.6, =0.3, =0.1$).

[0022] Optimize the execution process: Initialization: Encode the N positions of the first monitoring sequence into chromosomes (each gene represents the coordinates and monitoring range of a position); Selection: Calculate the fitness value of each chromosome, and retain the top 30% of individuals with the highest fitness; Crossover: Randomly pair the remaining individuals and exchange some genes (crossover rate = 0.8) to generate new chromosomes; Mutation: Randomly adjust the monitoring range or coordinates in the chromosome with a 5% probability; Iteration termination: When the optimal fitness value changes less than 1% for 10 consecutive generations or reaches 100 generations, select the K positions with the highest fitness; Output: Assign monitoring points to the optimized K positions, and form the second monitoring sequence according to the rules of the first monitoring sequence (K < N and K is greater than or equal to 1).

[0023] Example of genetic algorithm optimization. Assume that the first monitoring sequence outputs 4 positions: [(10, 25, 5km), (15, 20, 5km), (5, 20, 5km), (10, 15, 5km)]. Step 1, input data encoding: Chromosome 1: [10, 25, 5, 15, 20, 5, 5, 20, 5, 10, 15, 5] (complete sequence); Step 2, fitness calculation (take chromosome 1 as an example): : The pipeline corrosion rate at position (10, 25) = 0.7 → =0.7, : There is a fishing ground near (10, 25) → =0.9, : 20 km offshore → =0.4, =0.6×0.7 + 0.3×0.9 - 0.1×0.4 = 0.77. Step 3, optimization result: After 100 generations of iteration, retain the K = 2 positions with the highest fitness: [(10, 25, 5km), (15, 20, 5km)] → the second monitoring sequence.

[0024] Calculate the time for the crude oil to reach the sea surface and mark the path, including: Based on the depth data of the oil spill point, divide the ocean water body into the sea current dominant layer, the wind dominant layer, and the water temperature dominant layer; When a certain depth layer is dominated by a single factor, divide the depth value of this layer by the corresponding upward rate coefficient to obtain the upward time of the crude oil in this layer; Based on the superposition results of the three layers, the most influential factors, layer depth and layer time of the crude oil floating path are marked; The path mark sequence is used as the monitoring point attribute, and exclusive sequences are assigned to the first monitoring position and the second monitoring position.

[0025] Further explanation is that the oil spill point depth data is: the seabed depth of the source of the crude oil leak (unit: meter), which comes from marine platform sensors or historical databases; current dominant layer: the depth range in the ocean water body where the current is the main influencing factor of the crude oil rising speed (such as 0-30 meters); wind dominant layer: the depth range where the wind dominates the crude oil rising speed (such as 30-70 meters); water temperature dominant layer: the depth range where the water temperature dominates the crude oil rising speed (such as 70-100 meters); rising rate coefficient: the average rising speed of crude oil in a specific factor layer (unit: meter / hour), such as the current layer coefficient is 0.5; stratification time: the time it takes for crude oil to pass through a single layer, calculated as the layer depth divided by the rising rate coefficient; total time: the sum of the three layer times, which represents the total time it takes for the crude oil to reach the sea surface; path marker sequence: a data structure (such as a list) that records the maximum influencing factor, stratification depth and segmentation time; exclusive sequence: the path marker sequence assigned to the monitoring point as its attribute.

[0026] Data acquisition method: Oil spill point depth data: obtained through real-time monitoring of seabed pressure sensors or drilling platforms; rise rate coefficient: based on the ocean physics experiment database (such as the current layer coefficient 0.5m / h, wind layer 0.8m / h, water temperature layer 0.6m / h).

[0027] The ocean water is divided vertically into three layers (dominated by currents, wind, and water temperature). Each layer is determined by a single factor, determining the rate of crude oil ascent. The time taken for each layer is calculated (depth divided by rate coefficient), and the total time is added up. Key path information (such as the most influential factor) is also noted. This sequence serves as a monitoring point attribute for subsequent prioritization. The principle is based on the dynamics of crude oil ascent: environmental factors (currents, wind, and water temperature) at different depths have varying weights on crude oil movement.

[0028] Example: Location A, 100 meters deep: Current-dominated layer (0-30 meters): Depth 30 meters ÷ coefficient 0.5 = 60 hours; Wind-dominated layer (30-70 meters): Depth 40 meters ÷ coefficient 0.8 = 50 hours; Water temperature-dominated layer (70-100 meters): Depth 30 meters ÷ coefficient 0.6 = 50 hours; Total time = 60 + 50 + 50 = 160 hours. The most influential factor is the current (which takes the longest time); Path marker sequence: [(current, 30m, 60h), (wind, 40m, 50h), (water temperature, 30m, 50h)], assigned to monitoring point A (10, 25); Location B, 80 meters deep: Total time 130 hours, sequence assigned to B (15, 20).

[0029] Using the path marker sequence as the monitoring point attribute, the first monitoring location and the second monitoring location are assigned exclusive sequences, including: The layered time consumption data in the exclusive sequence is used as the operation object to extract the crude oil rising time consumption in the ocean current dominant layer, the wind dominant layer and the water temperature dominant layer; According to the sum of the three-layer time consumption, all oil spill locations are sorted from longest to shortest according to the total time consumption to obtain the total time consumption sorting result; The time consumption of each dominant layer is independently sorted from long to short to obtain the layered time consumption sorting result.

[0030] Further explanation is that the layered time consumption data: the time consumption value of each layer extracted from the exclusive sequence (such as the time consumption of the ocean current layer); the total time consumption sorting result: a list of all oil spill locations sorted from large to small by total time consumption (such as A(160h)>B(130h)); the layered time consumption sorting result: a list of the time consumption of each dominant layer (ocean current, wind speed, water temperature) independently sorted (such as the time consumption sorting of the ocean current layer).

[0031] Data acquisition method: Hierarchical time-consuming data: extract values ​​directly from the exclusive sequence (such as "60h" and "50h" in the sequence); sorting results: generated by processing time-consuming data through algorithms (such as quick sort).

[0032] Based on the time consumption data in the dedicated sequence, a multi-level sorting process is performed: the total time consumption for all locations is calculated and sorted from longest to shortest (prioritizing urgent locations with long latency). The time consumption for each layer (ocean current layer, wind layer, and water temperature layer) is independently sorted to resolve decision conflicts when the total time consumption is the same. This embodies the monitoring principle of "longer time consumption, higher risk" and ensures that resources are focused on high-latency areas.

[0033] Example: Extract data from the exclusive sequence of paragraph 1: Position A: current layer 60 hours, wind layer 50 hours, water temperature layer 50 hours; total time consumption 160 hours; Position B: current layer 40 hours, wind layer 60 hours, water temperature layer 30 hours; total time consumption 130 hours; total time consumption sorting result: [A(160h), B(130h)] (A has higher priority); layered time consumption sorting result: current layer: [A(60h), B(40h)] (from long to short); wind layer: [B(60h), A(50h)]; water temperature layer: [A(50h), B(30h)]; when the total consumption is the same (for example, the total consumption of the newly added position C is 160 hours), it is more time-consuming to sort them in the order of current layer > wind layer > water temperature layer.

[0034] Use the Transformer prediction model to predict spectral data, including: extracting a path marker feature vector according to the first path marker sequence and the second path marker sequence; extracting spectral feature labels based on the crude oil spectral data collected by the high-resolution spectrometer; Constructing a Transformer prediction model, taking the path label feature vector as input and the spectral feature label as output, and training the model; Using the trained Transformer prediction model, the new path label feature vector is input and the corresponding crude oil spectral data prediction value is output.

[0035] Further explanation: Processing complex spectral data through deep learning models can capture subtle changes in spectral characteristics, improve the accuracy and reliability of predictions, and provide a scientific basis for subsequent monitoring and analysis. Assume that the first path label sequence is [(ocean current, 30m, 1h), (wind force, 70m, 2h)], and the second path label sequence is [(water temperature, 100m, 1.5h), (ocean current, 20m, 0.5h)]. The path label sequences are converted into numerical feature vectors, and spectral data collected by a high-resolution spectrometer is independently obtained as labels to construct and train a Transformer model. The trained model takes the path label feature vectors as input and outputs corresponding crude oil spectral data predictions, such as [0.8, 0.6] and [0.7, 0.5]. These predictions help accurately assess the crude oil composition and state. The specific implementation steps are as follows: Step 1: Extract the path marker feature vector. The path marker sequence is structured data calculated based on the oil spill location, including: the maximum influencing factor (such as ocean current = 1, wind force = 2, water temperature = 3), depth (meters), and rise time (hours); Example: First path marker sequence → feature vector: [[1,30,1], [2,70,2]], Second path marker sequence → feature vector: [[3,100,1.5], [1,20,0.5]], Implementation method: The path analysis module converts the depth data into a numerical feature vector; Step 2: Construct a spectral training dataset, input feature source: the numerical vector of the path marker sequence, output label source: reflectance data collected by the spectrometer in the visible light / near infrared / mid-infrared bands (normalized), dataset structure: Input feature (X): [[influencing factor code, depth, time],...], output label Signature (Y): [spectral value 1, spectral value 2, ...], example alignment: X=[[1,30,1],[3,100,1.5]]→Y=[0.8,0.7] (true spectral values ​​at position 1 and position 2); Step 3: Train the Transformer model, model architecture: encoder-decoder structure, input dimension: 3 (corresponding to three elements of the feature vector), output dimension: 1 (single-value spectral prediction), training process: input X→model learning X→Y mapping→loss function (MSE) optimization, hyperparameters: number of layers = 6, number of attention heads = 8, learning rate = 0.001 (Adam optimizer); Step 4: Predict spectral data, prediction input: new path label feature vector (no spectral data required), prediction output: spectral value; example: input [[1,30,1]]→output 0.8, input [[3,100,1.5]]→output 0.7.

[0036] Fusion of the first monitoring sequence and the second monitoring sequence to form a first fused monitoring path includes: Based on the predicted values ​​of the spectral data, the monitoring priority is determined by sorting the values ​​from large to small; When it is necessary to integrate multiple monitoring sequence paths, the total time consumption sorting results and the layered time consumption sorting results are obtained; the oil spill location is selected based on the principle of priority based on the longest total time consumption; if the total time consumption is the same, the time consumption is compared step by step in the order of ocean current dominant layer, wind dominant layer, and water temperature dominant layer.

[0037] Further explanation is as follows: spectral data prediction value: crude oil spectral characteristic value (such as oil pollution concentration) output by the Transformer model, the larger the value, the higher the risk; monitoring priority: the order of location monitoring based on prediction value or time consumption; priority principle of longest total time consumption: priority is given to location monitoring with the longest total time consumption; level-by-level time consumption comparison: if the total time consumption is the same, the time consumption is compared in the order of ocean current layer → wind layer → water temperature layer.

[0038] Data acquisition method: Spectral data Prediction value: Transformer model based on path labeling feature vector prediction (input such as [influencing factor encoding, depth, time]).

[0039] The first monitoring sequence (generated by fuzzy inference) and the second monitoring sequence (optimized by a genetic algorithm) are integrated to form a unified path. Spectral data are prioritized from highest to lowest predicted value (highest value first). To resolve multiple sequence conflicts, a time-consuming sorting approach is introduced: the location with the longest total time is prioritized. If the time consumption is the same, the time consumption of the current layer, wind layer, and water temperature layer is compared in order. This principle ensures that high-risk and long-delay locations are covered first.

[0040] Example: The spectral prediction value of position A is 0.8, and the value of position B is 0.7 → Sort by predicted value: [A(0.8), B(0.7)]; if the predicted value of position C is 0.8 (the same as A), but the total time consumed A(160h)>C(150h): the total time consumption sorting result is: [A(160h), C(150h), B(130h)]; priority: A>C>B (the longest total time consumption is given priority); if the total time consumption of A and C is both 160h: compare the ocean current layer consumption: A(60h)>C(55h) → give priority to A; if the ocean current layers are the same (such as both 60h), compare the wind layer consumption; the first fusion monitoring path: connect positions by priority, such as A→C→B.

[0041] Taking into account the influence of energy loss and solar spectrum, the first fusion monitoring path is adjusted to form a second fusion monitoring path. The actual value of the crude oil spectrum data monitored is compared with the predicted value of the crude oil spectrum data to obtain spectral difference data and correct the Transformer prediction model and the second fusion monitoring path, including: Taking the first fused monitoring path as a benchmark, evaluate the negative impact of energy loss and solar spectrum in each monitoring point path on field monitoring to obtain a negative impact value; If there is a multipath negative impact, sort it by the negative impact value from large to small; The monitoring point access order is replanned according to the sorting results to form a second fusion monitoring path.

[0042] Further explanations include: Energy loss: the energy consumption of the monitoring equipment (such as battery degradation), the numerical negative impact; Solar spectrum impact: the degree to which sunlight interferes with the spectrometer measurement (such as high values ​​at noon); Negative impact value: a comprehensive score that quantifies energy loss and solar spectrum impact (a large value indicates a serious negative impact); Spectral difference data: the absolute value of the difference between the actual monitored value and the predicted value (such as |actual-predicted|).

[0043] Data acquisition method: Actual spectral data: On-site collection by high-resolution spectrometer (visible light / near infrared / mid-infrared band); Negative impact value: Calculated by formula: ,in, : Energy loss weight (default 0.6), : Solar spectrum impact weight (default 0.4), energy loss: estimated based on device distance and power consumption model, solar spectrum impact: based on time (e.g. noon = 0.9, dusk = 0.2).

[0044] Using the first fused monitoring path as a benchmark, the system evaluates the negative impact of each path (energy loss and sunlight interference). It then sorts the paths by negative impact from highest to lowest, re-planning the access sequence (prioritizing paths with the lowest negative impact). The system then compares actual and predicted spectral data, using the spectral differences to modify the Transformer model and paths. This minimizes external interference and improves monitoring accuracy.

[0045] Example: First fusion path: A→B; Negative impact assessment: Point A has high energy loss (0.8), and the solar impact of noon monitoring is 0.7 → Negative impact value = 0.6×0.8+0.4×0.7=0.76; Point B has energy loss of 0.5, and the solar impact of dusk monitoring is 0.2 → Negative impact value = 0.6×0.5+0.4×0.2=0.38; Sorting negative impact values: [A(0.76), B(0.38)] → Replanning to B→A (visiting low negative impact points first); Actual spectrum: A actual value 0.75 vs. predicted 0.8 → Difference 0.05, used to correct model parameters; Second fusion monitoring path: B→A.

[0046] Generate monitoring priority marks for oil spill locations based on spectral difference data and identify key monitoring areas, including: Taking the spectrum difference data as input, recording the number of corrections of the crude oil spectrum data prediction value and the absolute value of the spectrum difference data difference; The number of corrections of the crude oil spectrum data prediction value is used as a first weight, and the absolute value of the spectrum difference data difference is used as a second weight; According to the sum of the two weights, the monitoring locations are sorted from largest to smallest according to the total weight value; The location with the highest total weight is selected as the priority monitoring point; The monitoring path, the first weight and the second weight are updated regularly using a closed-loop mechanism.

[0047] A further explanation is that the spectral difference data is the absolute value of the difference between the actual and predicted spectral values ​​(such as |0.75-0.8|=0.05); the number of corrections is the number of times the Transformer model corrects the predicted value of a certain position (historical records); the first weight is the weight of the number of corrections, which reflects the instability of the model; the second weight is the weight of the absolute value of the spectral difference, which reflects the current error size; the total weight is the first weight + the second weight, which is used to sort the priority; the closed-loop mechanism is an iterative system that regularly feeds back data, updates weights and paths.

[0048] Data acquisition method: Number of corrections: Count the number of historical corrections from the system log; Absolute value of difference: Real-time calculation | Actual monitoring value - predicted value |.

[0049] Using spectral difference data as input, the system calculates the first weight (number of corrections) and the second weight (absolute value of the difference). The total weight equals the first weight + the second weight, and the locations are ranked from highest to lowest by total weight. The point with the highest total weight is prioritized for monitoring. A closed-loop mechanism regularly updates the weights and paths (e.g., every 24 hours) to ensure adaptive optimization. The principle is to focus on areas with high errors or frequent corrections, improving monitoring efficiency.

[0050] Example: Location A: 3 corrections (first weight 3), absolute value of difference 0.05 (second weight 0.05) → total weight 3.05; Location B: 1 correction (weight 1), difference 0.1 (weight 0.1) → total weight 1.1; Sorting: [A(3.05), B(1.1)] → priority monitoring A; Closed-loop update: After 24 hours, the number of corrections for A increases to 4, and the difference decreases to 0.03 → new total weight 4.03, maintaining the highest priority; Key monitoring area: Location A(10,25).

[0051] Example 2: Based on Example 1, Figure 2 As shown, a marine oil spill monitoring system based on crude oil spectrum includes: Fuzzy reasoning and genetic algorithm module: used to collect and preprocess ocean-related data, determine the location of the oil spill through fuzzy reasoning models, and use genetic algorithms to optimize the location of monitoring points; Path analysis module: used to calculate the time it takes for crude oil to reach the sea surface based on depth data, mark the oil spill path, and assign path marker sequences according to monitoring points; Spectral data acquisition and processing module: used to collect ocean surface spectral data using a high-resolution spectrometer, screen key features after preprocessing, and predict spectral data based on the Transformer model; Fusion monitoring path module: used to combine the spectral data prediction values ​​of different monitoring sequences to form a comprehensive monitoring path connecting all oil spill locations; Energy loss and solar spectrum impact module: used to evaluate the negative impact of energy loss and solar spectrum, and adjust the monitoring path to optimize the monitoring effect; Monitoring optimization and feedback module: used to compare the difference between predicted values ​​and actual monitored values, correct the Transformer prediction model and monitoring path, mark key monitoring locations, and continuously optimize through a closed-loop mechanism.

[0052] The above is a detailed introduction to the present application. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core idea of ​​the present application. At the same time, for those skilled in the art, based on the idea of ​​the present application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for monitoring marine oil spills based on crude oil spectra, characterized in that: The following steps are involved: S1: Using the fuzzy inference model to identify N marine oil spill locations, and assigning monitoring points to the N marine oil spill locations to form the first monitoring sequence; S2: Optimizing K important oil spill locations and the monitoring points corresponding to the important oil spill locations by genetic algorithm to form a second monitoring sequence, where K is less than N and is a positive integer greater than or equal to 1; S3: Calculate the time it takes for the crude oil to reach the sea surface and mark its path; use a high-resolution spectrometer to collect spectral data in the visible, near-infrared, and mid-infrared bands, and filter features after preprocessing; Predicting spectral data using a Transformer prediction model, fusing the first monitoring sequence and the second monitoring sequence to form a first fused monitoring path; S4: Considering the influence of energy loss and solar spectrum, the first fusion monitoring path is adjusted to form a second fusion monitoring path; the actual value of the crude oil spectrum data monitored is compared with the predicted value of the crude oil spectrum data, the spectrum difference data is obtained, and the Transformer prediction model and the second fusion monitoring path are corrected; S5: Generate monitoring priority markers for oil spill locations based on spectral difference data and identify key monitoring areas; continuously iterate and optimize the monitoring path and Transformer prediction model through a closed-loop feedback mechanism.

2. A method for monitoring marine oil spills based on crude oil spectrum according to claim 1, characterized in that: The method of using a fuzzy inference model to identify N marine oil spill locations and allocating monitoring points to the N marine oil spill locations to form a first monitoring sequence includes: Collect and pre-process data on offshore platforms, submarine pipelines, and marine environments, define fuzzy set membership functions, and establish a fuzzy rule base; Perform fuzzy reasoning, including data fuzzification, rule matching and reasoning results, and finally defuzzify and convert the results into specific ocean oil spill locations; Taking the sea surface projection point of the marine oil spill location as the monitoring location center, the farthest sea surface monitoring range is allocated to N marine monitoring locations, and then the N marine monitoring locations and the farthest sea surface monitoring ranges allocated to the N marine monitoring locations constitute the first marine oil spill monitoring sequence.

3. A method for monitoring marine oil spills based on crude oil spectrum according to claim 1, characterized in that: The method of optimizing K important oil spill locations and the monitoring points corresponding to the important oil spill locations by a genetic algorithm to form a second monitoring sequence, where K is less than N and is a positive integer greater than or equal to 1, includes: Generate an initial set of potential ocean oil spill locations, evaluate the fitness of each location, and select locations with high fitness to enter the next generation; Perform crossover and mutation operations to generate a new set of ocean oil spill locations; Merge the parent and offspring generations, select the ocean oil spill locations with high fitness to enter the next generation, until the maximum fitness threshold is reached, and the optimal K ocean oil spill locations are selected; K ocean monitoring locations are allocated to the optimal K ocean oil spill locations. Each monitoring location is centered on the sea surface projection point of the oil spill point and is allocated the farthest sea surface monitoring range to form the second ocean oil spill monitoring sequence.

4. A method for monitoring marine oil spills based on crude oil spectrum according to claim 1, characterized in that: The method of calculating the time when the crude oil reaches the sea surface and marking the path includes: Based on the depth data of the oil spill point, the ocean water body is divided into the current-dominated layer, the wind-dominated layer and the water temperature-dominated layer; When a certain depth layer is dominated by a single factor, the depth value of the layer is divided by the corresponding rising rate coefficient to obtain the rising time of the crude oil in the layer; Based on the superposition results of the three layers, the most influential factors, layer depth and layer time of the crude oil floating path are marked; The path mark sequence is used as the monitoring point attribute, and exclusive sequences are assigned to the first monitoring position and the second monitoring position.

5. A method for monitoring marine oil spills based on crude oil spectrum according to claim 4, characterized in that: The method of using the path mark sequence as the monitoring point attribute and allocating exclusive sequences to the first monitoring position and the second monitoring position includes: The layered time consumption data in the exclusive sequence is used as the operation object to extract the crude oil rising time consumption in the ocean current dominant layer, the wind dominant layer and the water temperature dominant layer; According to the sum of the three-layer time consumption, all oil spill locations are sorted from longest to shortest according to the total time consumption to obtain the total time consumption sorting result; The time consumption of each dominant layer is independently sorted from long to short to obtain the layered time consumption sorting result.

6. The method for monitoring marine oil spills based on crude oil spectrum according to claim 1, wherein: The method of predicting spectral data using the Transformer prediction model includes: extracting a path marker feature vector according to the first path marker sequence and the second path marker sequence; extracting spectral feature labels based on the crude oil spectral data collected by the high-resolution spectrometer; Constructing a Transformer prediction model, taking the path label feature vector as input and the spectral feature label as output, and training the model; Using the trained Transformer prediction model, the new path label feature vector is input and the corresponding crude oil spectral data prediction value is output.

7. The method for monitoring marine oil spills based on crude oil spectrum according to claim 1, wherein: The fusing the first monitoring sequence and the second monitoring sequence to form a first fused monitoring path includes: Based on the predicted values ​​of the spectral data, the monitoring priority is determined by sorting the values ​​from large to small; When it is necessary to integrate multiple monitoring sequence paths, the total time consumption sorting results and the layered time consumption sorting results are obtained; the oil spill location is selected based on the principle of priority based on the longest total time consumption; if the total time consumption is the same, the time consumption is compared step by step in the order of ocean current dominant layer, wind dominant layer, and water temperature dominant layer.

8. The method for monitoring marine oil spills based on crude oil spectrum according to claim 1, wherein: The first fusion monitoring path is adjusted to form a second fusion monitoring path by taking into account the influence of energy loss and solar spectrum; the actual value of the crude oil spectrum data monitored is compared with the predicted value of the crude oil spectrum data, the spectrum difference data is obtained, and the Transformer prediction model and the second fusion monitoring path are corrected, including: Taking the first fused monitoring path as a benchmark, evaluate the negative impact of energy loss and solar spectrum in each monitoring point path on field monitoring to obtain a negative impact value; If there is a multipath negative impact, sort it by the negative impact value from large to small; The monitoring point access order is replanned according to the sorting results to form a second fusion monitoring path.

9. The method for monitoring marine oil spills based on crude oil spectrum according to claim 1, wherein: The step of generating a monitoring priority mark for the oil spill location based on the spectral difference data and determining a key monitoring area includes: Taking the spectrum difference data as input, recording the number of corrections of the crude oil spectrum data prediction value and the absolute value of the spectrum difference data difference; The number of corrections of the crude oil spectrum data prediction value is used as a first weight, and the absolute value of the spectrum difference data difference is used as a second weight; According to the sum of the two weights, the monitoring locations are sorted from largest to smallest according to the total weight value; The location with the highest total weight is selected as the priority monitoring point; The monitoring path, the first weight and the second weight are updated regularly using a closed-loop mechanism.

10. A marine oil spill monitoring system based on crude oil spectrum, used in the marine oil spill monitoring method based on crude oil spectrum according to any one of claims 1 to 9, characterized in that: Includes: Fuzzy reasoning and genetic algorithm module: used to collect and preprocess ocean-related data, determine the location of the oil spill through fuzzy reasoning models, and use genetic algorithms to optimize the location of monitoring points; Path analysis module: used to calculate the time it takes for crude oil to reach the sea surface based on depth data, mark the oil spill path, and assign path marker sequences according to monitoring points; Spectral data acquisition and processing module: used to collect ocean surface spectral data using a high-resolution spectrometer, screen key features after preprocessing, and predict spectral data based on the Transformer model; Fusion monitoring path module: used to combine the spectral data prediction values ​​of different monitoring sequences to form a comprehensive monitoring path connecting all oil spill locations; Energy loss and solar spectrum impact module: used to evaluate the negative impact of energy loss and solar spectrum, and adjust the monitoring path to optimize the monitoring effect; Monitoring optimization and feedback module: used to compare the difference between predicted values ​​and actual monitored values, correct the Transformer prediction model and monitoring path, mark key monitoring locations, and continuously optimize through a closed-loop mechanism.

Images