Ocean internal wave data acquisition method, medium and system

Abstract

The present invention provides a method, medium and system for collecting ocean internal wave data, which belongs to the field of ocean hydrological observation technology. The present invention collects water body movement data through an acoustic Doppler current profiler and seawater parameters through a temperature-salinity-depth profiler, and combines satellite altimeter and drifting buoy data. Time-frequency analysis is performed on the water body movement data and bandpass filtering is used to obtain internal wave candidate data. A set of internal wave mathematical equations including basic fluid mechanics equations, buoyancy frequency, dispersion relationship, etc. is constructed to verify and filter out valid data. The valid data is input into a pre-trained classification model to determine the internal wave type, and finally the key waveform segment is extracted through a preset extraction equation set to generate and output internal wave characteristic waveform data. The present invention combines multi-source observation data and mathematical and physical models to realize the identification, classification and feature extraction of ocean internal waves, solving the current problem of the lack of a data acquisition method that can effectively extract effective ocean internal wave signals and accurately identify internal wave characteristics.

Description

Technical Field

[0001] The present invention belongs to the technical field of ocean hydrological observation, and in particular relates to an ocean internal wave data acquisition method, medium and system. Background Art

[0002] Internal waves are a crucial physical phenomenon in ocean dynamics. Generated by seawater density stratification and buoyancy, they propagate within the ocean as waves. These waves not only influence the transport of matter and energy within the ocean but also have profound impacts on marine ecosystems, marine engineering facilities, and underwater acoustics. In recent years, with global climate change and increased marine development, the need for accurate monitoring and analysis of internal waves has become increasingly urgent. Modern ocean internal wave monitoring technology has developed a multi-layered and multi-dimensional observation system. In field observations, acoustic Doppler current profilers, with their high temporal and spatial resolution and continuous observation capabilities, have become a core instrument for studying the dynamic characteristics of internal waves. They can simultaneously obtain horizontal and vertical velocity profiles, providing critical data for studying internal wave propagation characteristics. Temperature-salinity-depth profilers, on the other hand, measure seawater temperature, salinity, and pressure to obtain information on the density stratification structure of seawater. This data is the basis for calculating buoyancy frequency and analyzing internal wave propagation conditions.

[0003] However, existing ocean internal wave monitoring technology faces a prominent technical problem in practical applications: the collected raw data contains a large amount of irrelevant information and noise interference, resulting in insufficient accuracy in internal wave signal identification and feature extraction. This problem mainly manifests itself in the following aspects: First, there are multi-scale motion phenomena in the ocean environment, such as tides, eddies, and turbulence, and their signals are often mixed with internal wave signals, increasing the difficulty of separating effective signals. Secondly, the measurement errors and environmental noise of various observation instruments themselves can also affect data quality. Thirdly, existing data processing methods often analyze a single data source and lack a mechanism for the comprehensive utilization of multi-source data, which results in the acquired internal wave feature information being incomplete and inaccurate.

[0004] These issues severely restrict the in-depth study of ocean internal waves. In practical applications, inaccurate identification of internal wave characteristics can lead to deviations in marine engineering design parameters, affecting the safe operation of offshore platforms. In scientific research, this can lead to misunderstandings about internal wave propagation mechanisms, hindering the construction and verification of relevant theoretical models. Therefore, establishing a data acquisition and processing method that can effectively extract effective ocean internal wave signals and accurately identify internal wave characteristics is of great practical significance. Summary of the Invention

[0005] In view of this, the present invention provides an ocean internal wave data acquisition method, medium and system, which can solve the technical problem that the existing technology lacks a data acquisition method that can effectively extract effective ocean internal wave signals and accurately identify internal wave characteristics.

[0006] The present invention is achieved in that:

[0007] A first aspect of the present invention provides a method for collecting ocean internal wave data, comprising the following steps:

[0008] S10, continuously and in real time using an acoustic Doppler current profiler to collect ocean water movement data, the ocean water movement data including horizontal current velocity data and vertical current velocity data; simultaneously using a temperature-salinity-depth profiler to collect seawater temperature, salinity, and depth data, and calculating density gradient data and background density data based on the seawater state equation; inverting satellite altimeter data to obtain sea surface wave number direction data; and measuring group velocity data using a deployed drifting buoy;

[0009] S20, performing time-frequency analysis on the ocean water movement data to obtain ocean internal wave frequency range data;

[0010] S30, filtering the ocean water motion data using a bandpass filter according to the ocean internal wave frequency range data to obtain ocean internal wave candidate data;

[0011] S40, constructing a mathematical equation group for ocean internal waves, wherein the mathematical equation group for ocean internal waves includes a basic equation of fluid mechanics, a buoyancy frequency equation, a dispersion relation equation, a density perturbation equation, an energy propagation equation, and a phase velocity equation;

[0012] S50, substituting the candidate ocean internal wave data into the ocean internal wave mathematical equation group for verification, and screening out valid ocean internal wave data that satisfies the ocean internal wave mathematical equation group;

[0013] S60, inputting the ocean internal wave valid data into a pre-trained ocean internal wave classification model for classification to obtain ocean internal wave type data;

[0014] S70. Utilizing a preset extraction equation group, extract key waveform segments from the ocean internal wave valid data according to the ocean internal wave type data, generate ocean internal wave characteristic waveform data, and output the data.

[0015] On the basis of the above technical solution, the ocean internal wave data acquisition method of the present invention can also be improved as follows:

[0016] The basic fluid mechanics equation is used to calculate the motion state of seawater particles. The input includes horizontal velocity data and vertical velocity data, and the output is the motion trajectory data of seawater particles.

[0017] The buoyancy frequency equation is used to calculate the seawater stratification intensity. The input includes density gradient data and gravity acceleration data, and the output is the buoyancy frequency value.

[0018] The dispersion relation equation is used to verify the internal wave propagation characteristics. The input includes angular frequency data and wave number data, and the output is the internal wave dispersion relation verification result.

[0019] The density perturbation equation is used to calculate density field changes. The input includes background density data and perturbation amplitude data, and the output is density perturbation distribution data.

[0020] The energy propagation equation is used to calculate the energy propagation direction. The input includes group velocity data and wave number direction data, and the output is energy propagation vector data.

[0021] The phase velocity equation is used to calculate the wave propagation velocity. The input includes angular frequency data and wave number data, and the output is the phase velocity value.

[0022] Furthermore, in the step of screening out valid ocean internal wave data that meets the ocean internal wave mathematical equation group, the screening criteria are that the internal wave dispersion relationship verification results meet the theoretical prediction value, the density disturbance distribution data is within the preset threshold range, and the energy propagation vector data points reasonably.

[0023] Among them, the ocean internal wave classification model adopts a convolutional neural network structure, specifically including four convolution layers, two pooling layers and three fully connected layers.

[0024] Furthermore, the ocean internal wave classification model uses labeled historical ocean internal wave data to train the deep neural network model, specifically using historical ocean internal wave data with labeled information for supervised learning training.

[0025] Furthermore, the preset extraction equation group includes Fourier transform equation, wavelet transform equation, Hilbert transform equation, and empirical mode decomposition equation.

[0026] Furthermore, the Fourier transform equation is used to extract the main frequency waveform segment, the input is the effective data of the ocean internal wave, and the output is the main frequency waveform data;

[0027] The wavelet transform equation is used to extract energy concentrated waveform segments, the input is effective ocean internal wave data, and the output is energy concentrated segment waveform data;

[0028] The Hilbert transform equation is used to extract the envelope mutation waveform segment, the input is the effective data of the ocean internal wave, and the output is the envelope mutation waveform data;

[0029] The empirical mode decomposition equation is used to extract characteristic mode waveform segments, the input is effective ocean internal wave data, and the output is characteristic mode segment waveform data.

[0030] Furthermore, the preset threshold range of the density disturbance distribution data is ±30% of the background density gradient.

[0031] The seawater state equation is specifically expressed as follows:

[0032] ρ=ρ0[1-α(T-T0)+β(S-S0)-γ(P-P0)];

[0033]

[0034] Where ρ is the density of seawater, kg / m 3 ; ρ0 is the reference density, which is 1025 kg / m 3 ; P is temperature, °C; T0 is reference temperature, taken as 15 °C; S is salinity; S0 is reference salinity, taken as 35; P is pressure, Pa; P0 is reference pressure, taken as 1.013×10 5 Pa; α is the thermal expansion coefficient, which is 2.1×10 -4 / ℃; β is the salinity contraction coefficient, which is 7.6×10 -4 ;γ is the compression coefficient, which is 4.5×10 -10 / Pa; z is the depth, m.

[0035] The parameter acquisition method is:

[0036] T, S, and P are directly measured by the temperature-salinity-depth profiler;

[0037] Background density ρ b Obtained by time averaging the calculated density:

[0038]

[0039] Where N is the number of measurements; ρ i The density value calculated for the i-th measurement.

[0040] The equation for retrieving the direction of sea surface wavenumbers using satellite altimeter data is specifically expressed as follows:

[0041]

[0042] Where, is the wave number vector; η is the sea surface height anomaly, m; k is the wave number size, m -1 ; g is the acceleration due to gravity, which is 9.81 m / s 2 ; c is the phase velocity, m / s; h is the water depth, m.

[0043] The parameter acquisition method is:

[0044] η is obtained by direct measurement of satellite altimeter;

[0045] h is obtained by sounding.

[0046] The equations of the ocean internal wave mathematical equations are specifically described as follows:

[0047] (1) Basic equations of fluid mechanics:

[0048]

[0049] Where u is the horizontal velocity, m / s; w is the vertical velocity, m / s; p is the pressure, Pa; ρ′ is the density disturbance, kg / m 3 ν is the kinematic viscosity coefficient, which is 1.01×10 -6 m 2 / s.

[0050] (2) Buoyancy frequency equation:

[0051]

[0052] Where N is the buoyancy frequency, rad / s.

[0053] (3) Dispersion relation equation:

[0054]

[0055] Where ω is the angular frequency, rad / s; k x is the horizontal wave number, m -1 ;k z is the vertical wave number, m -1 .

[0056] (4) Density perturbation equation:

[0057]

[0058] Where A is the amplitude, m; φ(z) is the vertical mode function.

[0059] (5) Energy propagation equation:

[0060]

[0061] Where, is the group velocity vector, m / s.

[0062] (6) Phase velocity equation:

[0063]

[0064] Where, cp is the phase velocity, m / s.

[0065] The equations of the extraction equation group are specifically described as follows:

[0066] (1) Fourier transform equation:

[0067] F=W F X;

[0068] W F =[ω mn ],ω mn =e -2πi(m-1)(n-1) / N ;

[0069] P=diag(|F| 2 );

[0070]

[0071] Where, is the input data matrix, each column represents an observation time series; is the Fourier transform coefficient matrix; is the Fourier transform matrix; is the power spectrum matrix; The main frequency band selection matrix is based on the power spectrum threshold λ F Sure; It is the output main frequency band waveform data matrix.

[0072] (2) Wavelet transform equation:

[0073] W = ΨX;

[0074]

[0075] E=|W| 2 ;

[0076]

[0077] Where, is the wavelet coefficient matrix; is the wavelet basis matrix, L is the scale number; is the wavelet energy matrix; The matrix is selected for the energy concentration segment and is determined according to the energy threshold λW; It is the output energy concentration segment waveform data matrix.

[0078] (3) Hilbert transform equation:

[0079] H=W H X;

[0080]

[0081] Where, is the Hilbert transform matrix; is the Hilbert transform kernel matrix; is the envelope matrix; Select the matrix for the envelope mutation segment, according to the envelope change rate threshold λ H Sure; It is the output envelope mutation segment waveform data matrix.

[0082] (4) Empirical mode decomposition equation:

[0083] C k =S k X;

[0084]

[0085] Where, is the kth internal model function matrix; is the screening operator matrix; is the envelope mean matrix; are the upper and lower envelope matrices respectively; Select the matrix for the characteristic mode segment according to the modal energy ratio threshold λ E Sure; It is the output characteristic mode segment waveform data matrix.

[0086] Parameter acquisition method:

[0087] 1.λ F Determined by power spectrum distribution, generally 10%-30% of the maximum power;

[0088] 2.λ W Determined by wavelet energy distribution, generally 60%-80% of the total energy;

[0089] 3.λ H Determined by envelope change rate statistics, usually taking 2 times the standard deviation;

[0090] 4.λ E It is determined by the modal energy ratio, generally 30%-50%.

[0091] A second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are run in a computer, they are used to execute the above-mentioned ocean internal wave data collection method.

[0092] A third aspect of the present invention provides an ocean internal wave data acquisition system, which includes the above-mentioned computer-readable storage medium.

[0093] Compared with the prior art, the method, medium and system for collecting ocean internal wave data provided by the present invention have the following beneficial effects:

[0094] 1. This method achieves multi-sensor data fusion and analysis. This method fully utilizes data collected by various sensor devices, such as acoustic Doppler current profilers, temperature-salinity-depth profilers, and satellite remote sensing equipment. Through mathematical modeling and intelligent algorithms, this data is organically combined, providing a solid data foundation for the comprehensive analysis of internal wave parameters.

[0095] 2. A method for extracting internal wave characteristics based on a set of mathematical equations is proposed. This method constructs a set of mathematical equations describing the physical properties of internal waves and verifies them using monitoring data. Key parameters such as the frequency, amplitude, and propagation velocity of the internal waves are extracted from these equations, providing a reliable mathematical basis for internal wave characteristic analysis.

[0096] 3. Deep learning is used to achieve intelligent identification of internal wave types. This method uses a pre-trained convolutional neural network model to automatically classify the filtered internal wave valid data, achieving rapid identification of internal wave types and significantly improving the efficiency of internal wave research.

[0097] 4. Comprehensively apply multiple signal processing algorithms to generate internal wave characteristic waveforms. This method uses methods such as Fourier transform, wavelet transform, Hilbert transform, and empirical mode decomposition to extract key waveform features such as the main frequency band, energy concentration segment, envelope mutation segment, and characteristic mode segment from the effective internal wave data, providing rich information for internal wave dynamic analysis.

[0098] Therefore, the present invention solves the technical problem that the prior art lacks a data acquisition method that can effectively extract effective ocean internal wave signals and accurately identify internal wave characteristics. BRIEF DESCRIPTION OF THE DRAWINGS

[0099] Figure 1 A flow chart of the method provided by the present invention;

[0100] Figure 2 This is a Fourier transform main frequency band waveform analysis diagram in the embodiment;

[0101] Figure 3 Graphs showing typical wave characteristics of three different types of internal waves in the embodiment;

[0102] Figure 4 2 is a performance comparison chart of the method of the present invention and the traditional method in the embodiment. DETAILED DESCRIPTION

[0103] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0104] like Figure 1 FIG. 1 is a flow chart of a method for collecting ocean internal wave data provided by the present invention. The method includes the following steps:

[0105] S10. Continuously and in real time, collect ocean water movement data using an acoustic Doppler current profiler, which includes horizontal and vertical current velocity data. Simultaneously, collect seawater temperature, salinity, and depth data using a temperature-salinity-depth profiler, and calculate density gradient data and background density data based on the seawater state equation. Obtain sea surface wave direction data by inverting satellite altimeter data. Obtain group velocity data by measuring with a deployed drifting buoy.

[0106] S20, performing time-frequency analysis on the ocean water movement data to obtain ocean internal wave frequency range data;

[0107] S30, filtering the ocean water motion data using a bandpass filter based on the ocean internal wave frequency range data to obtain ocean internal wave candidate data;

[0108] S40, constructing a mathematical equation group for ocean internal waves, the mathematical equation group for ocean internal waves including basic equations of fluid mechanics, buoyancy frequency equation, dispersion relation equation, density perturbation equation, energy propagation equation, and phase velocity equation;

[0109] S50, substituting the candidate ocean internal wave data into the ocean internal wave mathematical equations for verification, and screening out valid ocean internal wave data that satisfies the ocean internal wave mathematical equations;

[0110] S60, inputting the ocean internal wave valid data into a pre-trained ocean internal wave classification model for classification to obtain ocean internal wave type data;

[0111] S70. Utilizing a preset extraction equation group, extract key waveform segments from the effective ocean internal wave data according to the ocean internal wave type data, generate and output ocean internal wave characteristic waveform data.

[0112] The specific implementation of the above steps is described in detail below:

[0113] The specific implementation of step S10 is as follows: First, an acoustic Doppler current profiler is used to continuously and in real time collect ocean water motion data, including horizontal and vertical velocity data. At the same time, a temperature-salinity-depth profiler is used to collect seawater temperature, salinity, and depth data. Then, density gradient data and background density data are calculated based on the seawater state equation. In addition, sea surface wave number direction data are inverted using satellite altimeter data. Finally, group velocity data of ocean internal waves are measured using a deployed drifting buoy. The purpose of this step is to obtain various basic data required to describe the motion and propagation characteristics of ocean internal waves.

[0114] Among them, the specific expression of the seawater state equation is:

[0115] ρ=ρ0[1-α(T-T0)+β(S-S0)-γ(P-P0)];

[0116]

[0117] Where ρ is the density of seawater, in kg / m 3 ; ρ0 is the reference density, which is 1025kg / m 3 ; T is temperature, unit is ℃; T0 is reference temperature, value is 15℃; S is salinity; S0 is reference salinity, value is 35; P is pressure, unit is Pa; P0 is reference pressure, value is 1.013×10 5 Pa; α is the thermal expansion coefficient, which is 2.1×10 -4 / ℃; β is the salinity contraction coefficient, which is 7.6×10 -4 ;γ is the compression coefficient, which is 4.5×10 -10 / Pa; z is the depth, in meters.

[0118] Among these parameters, T, S and P can be directly measured by temperature-salinity-depth profiler. b It is obtained by time averaging the calculated density values:

[0119]

[0120] Where N is the number of measurements, ρ i The density value calculated for the i-th measurement.

[0121] The equation for retrieving the direction of sea surface wavenumbers using satellite altimeter data is as follows:

[0122]

[0123]

[0124] Where, is the wave number vector; η is the sea surface height anomaly, unit is m; k is the wave number size, unit is m -1 ; g is the acceleration due to gravity, which is 9.81m / s 2 c is the phase velocity in m / s, and h is the water depth in m. η can be directly measured by satellite altimeters, and h can be obtained by sounding.

[0125] The specific implementation of step S20 is: performing time-frequency analysis on the ocean water motion data obtained in step S10 to obtain frequency range data of ocean internal waves. The purpose of this step is to determine the frequency characteristics of ocean internal waves.

[0126] Step S30 is implemented by filtering the ocean water motion data acquired in step S10 using a bandpass filter based on the ocean internal wave frequency range data acquired in step S20 to obtain candidate ocean internal wave data. This step aims to extract the portion of the raw data that matches the internal wave frequency as candidate internal wave data.

[0127] The specific implementation of step S40 is: constructing a set of mathematical equations describing the motion and propagation characteristics of ocean internal waves, including the following equations:

[0128] (1) Basic equations of fluid mechanics:

[0129]

[0130] Where n is the horizontal velocity in m / s; w is the vertical velocity in m / s; p is the pressure in Pa; ρ′ is the density disturbance in kg / m 3 ; v is the kinematic viscosity coefficient, which is 1.01×10 -6 m 2 / s. These equations describe the motion state of seawater particles.

[0131] (2) Buoyancy frequency equation:

[0132]

[0133] Where N is the buoyancy frequency in rad / s. This equation is used to calculate the intensity of seawater stratification.

[0134] (3) Dispersion relation equation:

[0135]

[0136] Where ω is the angular frequency in rad / s; k x is the horizontal wave number, in m -1 ;k z is the vertical wave number, in m -1 This equation is used to verify the propagation characteristics of internal waves.

[0137] (4) Density perturbation equation:

[0138]

[0139] Where A is the internal wave amplitude in meters and φ(z) is the vertical mode function. This equation describes the density field changes caused by internal waves.

[0140] (5) Energy propagation equation:

[0141]

[0142] Where, is the group velocity vector in m / s. This equation is used to calculate the direction of propagation of internal wave energy.

[0143] (6) Phase velocity equation:

[0144]

[0145] Where c p is the phase velocity in m / s. This equation is used to calculate the propagation speed of internal waves.

[0146] This set of mathematical equations comprehensively describes the motion and propagation characteristics of ocean internal waves, laying the foundation for subsequent internal wave identification and feature extraction.

[0147] The specific implementation of step S50 is to substitute the candidate ocean internal wave data obtained in step S30 into the mathematical equations constructed in step S40 for verification, thereby screening out valid ocean internal wave data that satisfies the equations. Specific screening criteria include: the internal wave dispersion relationship verification results meet theoretical predictions, the density perturbation distribution data falls within a preset threshold range, and the energy propagation vector data points in a reasonable direction. The purpose of this step is to screen out valid data from the candidate data that meets the theoretical characteristics of internal waves. The preset threshold range for the density perturbation distribution data is generally set to ±30% of the background density gradient.

[0148] Step S60 is implemented by inputting the valid ocean internal wave data screened in step S50 into a pre-trained deep neural network model for classification, thereby obtaining ocean wave type data. This deep neural network model utilizes a convolutional neural network architecture, comprising four convolutional layers, two pooling layers, and three fully connected layers. During model training, supervised learning is performed using labeled historical ocean internal wave data. The goal of this step is to automatically identify and classify internal waves using deep learning techniques.

[0149] The specific implementation of step S70 is to use a preset set of extraction equations, based on the internal wave type data obtained in step S60, to extract key waveform segments from the effective ocean internal wave data screened in step S50, generate characteristic waveform data for the ocean internal waves, and output them. These extraction equations include:

[0150] (1) Fourier transform equation:

[0151] F=W F X;

[0152] W F =[ω mn ],ω mn =e -2πi(m-1)(n-1) / N ;

[0153] P=diag(|F| 2 );

[0154]

[0155] in, is the input data matrix; is the Fourier transform coefficient matrix; is the Fourier transform matrix; is the power spectrum matrix; The main frequency band selection matrix is based on the power spectrum threshold λ F (Usually 10%-30% of the maximum power) It is the output main frequency band waveform data matrix.

[0156] (2) Wavelet transform equation:

[0157] W = ΨX;

[0158]

[0159] E=|W| 2 ;

[0160]

[0161] in, is the wavelet coefficient matrix; is the wavelet basis matrix, L is the scale number; is the wavelet energy matrix; The matrix is selected for the energy concentration segment according to the energy threshold λ W (Usually 60%-80% of the total energy) It is the output energy concentration segment waveform data matrix.

[0162] (3) Hilbert transform equation:

[0163] H=W H X;

[0164]

[0165] in, is the Hilbert transform matrix; is the Hilbert transform kernel matrix; is the envelope matrix; Select the matrix for the envelope mutation segment, according to the envelope change rate threshold λ H (Usually take 2 times the standard deviation) to determine; It is the output envelope mutation segment waveform data matrix.

[0166] (4) Empirical mode decomposition equation:

[0167] C k =S k X;

[0168]

[0169] in, is the kth internal model function matrix; is the screening operator matrix; is the envelope mean matrix; are the upper and lower envelope matrices respectively; The matrix is selected for the characteristic mode segment according to the modal energy ratio threshold λ E (Usually 30%-50%) to determine; It is the output characteristic mode segment waveform data matrix.

[0170] A second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are run in a computer, they are used to execute the above-mentioned ocean internal wave data collection method.

[0171] A third aspect of the present invention provides an ocean internal wave data acquisition system, which includes the above-mentioned computer-readable storage medium.

[0172] Specifically, the principle of the present invention is:

[0173] First, from the perspective of physical mechanism, ocean internal waves, as a special wave phenomenon, have their own unique physical characteristics and propagation laws. The present invention comprehensively describes the physical characteristics of internal waves by constructing a complete set of mathematical equations for ocean internal waves. Among them, the basic equations of fluid mechanics describe the motion state of water particles under the action of internal waves, the buoyancy frequency equation reflects the influence of seawater stratification on the propagation of internal waves, the dispersion relation equation characterizes the relationship between the frequency and wave number of internal waves, the density perturbation equation describes the density field changes caused by internal waves, the energy propagation equation describes the propagation characteristics of internal wave energy, and the phase velocity equation reflects the propagation speed of the wave. These equations constitute a complete physical framework, which provides a theoretical basis for identifying real internal wave signals.

[0174] From a data processing perspective, this invention employs a progressive process of "coarse screening - fine verification - intelligent classification - feature extraction." First, time-frequency analysis is used to determine the frequency range in which internal waves may exist. A bandpass filter is then used to preliminarily screen candidate data, effectively removing obvious non-internal wave signals. Then, a rigorous verification system using mathematical equations is performed to ensure that the data meets the physical properties of internal waves. This physics-based verification method has a reliable theoretical basis and can effectively identify true internal wave signals.

[0175] In the intelligent classification phase, the present invention chooses a convolutional neural network as the core architecture of the classification model based on the following considerations: First, internal wave data has distinct spatiotemporal characteristics, and convolutional layers can effectively extract these features; second, the introduction of pooling layers can reduce the dimensionality of the data and extract more representative features; finally, multiple fully connected layers can establish nonlinear associations between features, achieving accurate classification. By training with labeled historical data, the model can learn the typical characteristics of internal waves, thereby accurately classifying new data.

[0176] To better understand and implement the present invention, the following provides an example of a specific application scenario: A marine research team conducted a month-long internal wave observation experiment in a certain area of the South China Sea. The experimental area was located in the central part of the South China Sea, at a depth of approximately 1,200 meters, with a relatively flat seabed. During the experiment, the team used equipment such as an acoustic Doppler current profiler, a temperature-salinity-depth profiler, satellite remote sensing, and drifting buoys to continuously observe and measure parameters such as ocean water movement, temperature, salinity, and density.

[0177] First, the team deployed four acoustic Doppler current profilers, stationed at depths of 100, 400, 800, and 1100 meters, to obtain three-dimensional velocity distribution data for the ocean water. These instruments took measurements every 10 minutes for 30 consecutive days, acquiring a total of 4032 sets of velocity data (4 instruments x 30 days x 24 hours / 0.17 hours). For example, the velocity data at a depth of 100 meters yields the time series of horizontal velocity u and vertical velocity w, as shown in Table 1:

[0178] Table 1. Horizontal and vertical velocity data at a depth of 100 meters

[0179] time u(m / s) w(m / s) 00:00 0.213 0.031 00:10 0.196 0.028 00:20 0.234 0.026 … … … 23:50 0.202 0.033

[0180] The team also deployed three temperature-salinity-depth profilers in the experimental area, at depths of 300, 700, and 1100 meters. These instruments took measurements every two hours for 30 consecutive days, acquiring a total of 360 sets of temperature-salinity-depth data (3 profilers x 30 days / 0.083 days). For example, the temperature, salinity, and pressure data at a depth of 700 meters are shown in Table 2.

[0181] Table 2. Temperature, salinity, and pressure data at a depth of 700 meters

[0182] time T(℃) S P(Pa) 00:00 5.12 34.52 7.0E+6 02:00 5.09 34.49 7.0E+6 04:00 5.06 34.46 7.0E+6 … … … … 22:00 5.18 34.57 7.0E+6

[0183] Based on these temperature, salinity and depth data, the density gradient can be calculated by the seawater state equation. and background density ρ b , as shown in Table 3:

[0184] Table 3. Density gradient and background density at a depth of 700 meters

[0185]

[0186] In addition, the team also used satellite altimeter data to invert the sea surface height anomaly η in the experimental area and obtained the sea surface wave number direction and wave number size k, as shown in Table 4:

[0187] Table 4. Sea surface wave direction and wave magnitude

[0188]

[0189] At the same time, the team deployed five drifting buoys in the experimental area and measured the group velocity of the ocean internal waves. As shown in Table 5:

[0190] Table 5. Group velocity of ocean internal waves

[0191]

[0192] With these basic data, the research team began to extract and analyze the characteristics of ocean internal waves.

[0193] First, the team performed time-frequency analysis on the flow velocity, temperature-salinity depth, and group velocity data obtained in step S10 and found that the frequency range of the internal wave was 0.001-0.01rad / s.

[0194] Then, according to this frequency range, the velocity and temperature-salinity-depth data are filtered using a bandpass filter to extract candidate data of internal waves.

[0195] Next, the team constructed a set of mathematical equations describing the motion and propagation characteristics of ocean internal waves, including the basic equations of fluid mechanics, the buoyancy frequency equation, the dispersion relation equation, the density perturbation equation, the energy propagation equation, and the phase velocity equation. Substituting the candidate internal wave data into these equations for verification, the results showed:

[0196] (1) Internal wave dispersion relation Angular frequency ω and wave number k in x、k z meet theoretical predictions;

[0197] (2) Density perturbation The amplitude a in the background density gradient ±30% of the range;

[0198] (3) Energy propagation direction Compared with the measured group velocity Basically consistent.

[0199] Through this verification process, the team screened out valid data that meets the characteristics of internal wave theory.

[0200] Next, the team fed this valid internal wave data into a pre-trained convolutional neural network model for classification. This model, consisting of four convolutional layers, two pooling layers, and three fully connected layers, was trained using labeled historical internal wave data for supervised learning. After classification, the internal wave data was divided into three types: free internal waves, accounting for 45%; forced internal waves, accounting for 35%; and vortex internal waves, accounting for 20%.

[0201] Finally, the team used Fourier transform, wavelet transform, Hilbert transform and empirical mode decomposition to extract waveform features such as the main frequency band, energy concentration segment, envelope mutation segment and characteristic mode segment from the effective internal wave data. Among them, the main frequency band waveform extracted by Fourier transform is as follows Figure 2 As shown in the figure, the main frequency is about 0.005 rad / s. The upper part shows the time domain waveform, showing the amplitude change of the internal wave over time; the lower part shows the frequency domain analysis results, which clearly shows that the main frequency is around 0.005 rad / s. Figure 3 This figure shows a comparison of typical waveform characteristics of three different types of internal waves. From top to bottom, they are the time domain waveforms of free internal waves, forced internal waves, and vortex internal waves. This figure shows the significant differences in the waveform structures of the three types of internal waves.

[0202] Figure 4 This is a performance comparison chart of the proposed method and traditional methods. The bar chart shows the performance of different methods in three dimensions: recognition accuracy, computation time, and memory usage. This chart clearly demonstrates that the proposed method outperforms traditional methods in all aspects, especially in computational efficiency and accuracy.

[0203] The above description is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed by the present invention, which should be covered by the scope of protection of the present invention.

Claims

1. A method for collecting ocean internal wave data, characterized in that: The following steps are involved: S10, continuously and in real time using an acoustic Doppler current profiler to collect ocean water movement data, the ocean water movement data including horizontal current velocity data and vertical current velocity data; simultaneously using a temperature-salinity-depth profiler to collect seawater temperature, salinity, and depth data, and calculating density gradient data and background density data based on the seawater state equation; inverting satellite altimeter data to obtain sea surface wave number direction data; and measuring group velocity data using a deployed drifting buoy; S20, performing time-frequency analysis on the ocean water movement data to obtain ocean internal wave frequency range data; S30, filtering the ocean water motion data using a bandpass filter according to the ocean internal wave frequency range data to obtain ocean internal wave candidate data; S40, constructing a mathematical equation group for ocean internal waves, wherein the mathematical equation group for ocean internal waves includes a basic equation of fluid mechanics, a buoyancy frequency equation, a dispersion relation equation, a density perturbation equation, an energy propagation equation, and a phase velocity equation; S50, substituting the candidate ocean internal wave data into the ocean internal wave mathematical equation group for verification, and screening out valid ocean internal wave data that satisfies the ocean internal wave mathematical equation group; S60, inputting the ocean internal wave valid data into a pre-trained ocean internal wave classification model for classification to obtain ocean internal wave type data; S70, using a preset extraction equation group to extract key waveform segments from the ocean internal wave valid data according to the ocean internal wave type data, generate ocean internal wave characteristic waveform data, and output the data; The ocean internal wave classification model adopts a convolutional neural network structure, specifically comprising four convolutional layers, two pooling layers, and three fully connected layers; the preset extraction equation group includes a Fourier transform equation, a wavelet transform equation, a Hilbert transform equation, and an empirical mode decomposition equation; the Fourier transform equation is used to extract the main frequency waveform segment, the input is the effective ocean internal wave data, and the output is the main frequency band waveform data; The wavelet transform equation is used to extract energy concentrated waveform segments, the input is effective ocean internal wave data, and the output is energy concentrated segment waveform data; The Hilbert transform equation is used to extract the envelope mutation waveform segment, the input is the effective data of the ocean internal wave, and the output is the envelope mutation waveform data; The empirical mode decomposition equation is used to extract characteristic mode waveform segments, the input is effective ocean internal wave data, and the output is characteristic mode segment waveform data.

2. The method for collecting ocean internal wave data according to claim 1, wherein: The basic fluid mechanics equation is used to calculate the motion state of seawater particles. The input includes horizontal velocity data and vertical velocity data, and the output is the motion trajectory data of seawater particles. The buoyancy frequency equation is used to calculate the seawater stratification intensity. The input includes density gradient data and gravity acceleration data, and the output is the buoyancy frequency value. The dispersion relation equation is used to verify the internal wave propagation characteristics. The input includes angular frequency data and wave number data, and the output is the internal wave dispersion relation verification result. The density perturbation equation is used to calculate density field changes. The input includes background density data and perturbation amplitude data, and the output is density perturbation distribution data. The energy propagation equation is used to calculate the energy propagation direction. The input includes group velocity data and wave number direction data, and the output is energy propagation vector data. The phase velocity equation is used to calculate the wave propagation velocity. The input includes angular frequency data and wave number data, and the output is the phase velocity value.

3. The method for collecting ocean internal wave data according to claim 2, wherein: In the step of screening out valid ocean internal wave data that meets the ocean internal wave mathematical equation group, the screening criteria are that the internal wave dispersion relationship verification result meets the theoretical prediction value, the density disturbance distribution data is within the preset threshold range and the energy propagation vector data points reasonably.

4. The method for collecting ocean internal wave data according to claim 3, wherein: The ocean internal wave classification model uses labeled historical ocean internal wave data to train the convolutional neural network, specifically using historical ocean internal wave data with labeled information for supervised learning training.

5. The method for collecting ocean internal wave data according to claim 4, characterized in that: The preset threshold range of the density perturbation distribution data is ±30% of the background density gradient.

6. A computer-readable storage medium, characterized in that The computer-readable storage medium stores program instructions, and when the program instructions are run in a computer, they are used to execute the ocean internal wave data acquisition method according to any one of claims 1 to 5.

7. An ocean internal wave data acquisition system, characterized in that: Contains the computer-readable storage medium of claim 6.

Images