Typhoon path inversion method and system based on ocean background noise spectrum characteristics

Abstract

The invention relates to the technical field of marine disaster early warning, and discloses a typhoon path inversion method and system based on marine background noise spectrum features, and the method comprises the steps: S1, constructing an LSTM-Attention feature fusion model; s2, noise data and meteorological data of at least three nodes are collected once every five minutes; the noise data comprises 8-20Hz infrasonic wave signals, and the meteorological data comprises air pressure data, wind speed data and temperature data; s3, screening the noise data to obtain an effective infrasonic wave signal; and S4, performing spectrum analysis and feature extraction on the effective infrasonic wave signals to obtain infrasonic wave features and the like. According to the method and system, a physical coupling mechanism of'typhoon infrasonic wave-barometric gradient-sea surface wind field 'is established through coupling modeling of ocean background noise infrasonic wave features (8-20 Hz) and meteorological parameters, and typhoon path deviation is pre-warned 6-12 hours in advance.

Description

Technical Field

[0001] This invention relates to the field of marine disaster early warning technology, specifically to a typhoon path inversion method and system based on the spectral characteristics of marine background noise. Background Technology

[0002] Offshore wind farms, marine observation stations, and other marine engineering scenarios urgently require real-time sensing of typhoon dynamics. Existing typhoon monitoring technologies have significant shortcomings: 1. The update cycle of satellite cloud images and meteorological buoy data generally reaches or exceeds 30 minutes. This time lag makes it difficult for the system to capture sudden changes in typhoon paths, including key dynamic features such as sharp changes in direction or sudden acceleration, severely affecting the real-time nature of early warnings; 2. There are significant blind spots in nearshore monitoring. When heavy rainfall occurs near the typhoon eyewall, traditional radar signals are easily interfered with, resulting in path prediction errors often exceeding 50 kilometers. This accuracy gap is particularly prominent when typhoons approach land, directly threatening the accuracy of disaster prevention decisions in coastal areas; 3. Existing monitoring methods have relatively limited data dimensions, mainly relying on changes in single parameters such as air pressure or wind speed, lacking a comprehensive characterization mechanism for the physical processes of typhoon-sea surface interaction. This fails to fully reflect the energy transfer and structural evolution of the typhoon system, leading to insufficient response of prediction models to complex environments. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the present invention aims to provide a typhoon path inversion method and system based on the spectral characteristics of marine background noise. This method and system can realize intelligent early warning of typhoon path inversion in real time, and is applicable to scenarios such as offshore wind farms and marine observation stations that need to sense typhoon dynamics in advance, solving the problems of low time resolution and large near-shore blind zone of traditional monitoring methods.

[0004] To achieve the above objectives, the present invention provides the following technical solution: A typhoon track retrieval method based on the spectral characteristics of ocean background noise includes: S1, constructing an LSTM-Attention feature fusion model; S2, collecting noise data and meteorological data from at least three nodes, every 5 minutes; the noise data includes 8-20Hz infrasound signals, and the meteorological data includes air pressure data, wind speed data, and temperature data; S3, filtering the noise data to obtain effective infrasound signals; S4, performing spectral analysis and feature extraction on the effective infrasound signals to obtain infrasound features, including center frequency offset and frequency band energy ratio; S5, filtering and verifying the meteorological data to obtain accurate air pressure data, wind speed data, and temperature data; S6, calculating the air pressure gradient and wind shear index based on the accurate air pressure and wind speed data, as meteorological parameters; S7, unifying the timestamps of the infrasound features and meteorological parameters; S8, inputting the timestamp-unified infrasound features and meteorological parameters as feature values ​​into the LSTM-Attention feature fusion model, and the LSTM-Attention feature fusion model outputs the typhoon track. The coordinates of the typhoon center and the typhoon's moving speed are calculated; S9, the average of the typhoon coordinates and typhoon moving speed obtained from each node is used as the final coordinates and final typhoon moving speed of the typhoon center; In S8, the process of the LSTM-Attention feature fusion model outputting the coordinates of the typhoon center and the typhoon's moving speed based on infrasound features and meteorological parameters includes: S81, converting the infrasound features and meteorological parameters into time-series data formats, and performing data normalization verification and adjustment to obtain a data tensor; S82, inputting the data tensor into the first layer LSTM for feature extraction, and outputting the first layer LSTM feature vector; S83, inputting the data tensor into the second layer LSTM for feature extraction, and outputting the deep feature vector; S84, calculating the Attention weights, normalizing the attention weights, and performing feature weighting, concatenating the Attention output with the deep feature vector to obtain the fused feature vector; S85, initializing the fully connected network, inputting the fused feature vector into the fully connected network, performing fully connected layer feature mapping, and outputting the coordinates of the typhoon center and the typhoon's moving speed.

[0005] In this invention, preferably, step S3 includes: S31, removing invalid data and correcting the amplitude of the infrasound signal based on the ambient temperature at the time of acquisition; S32, processing the infrasound signal using a wavelet threshold denoising algorithm; and S33, verifying the signal-to-noise ratio of the infrasound signal and correcting infrasound signal segments that are higher than the signal-to-noise ratio threshold to obtain a valid infrasound signal.

[0006] In this invention, preferably, step S4 includes: S41, performing wavelet packet decomposition on the effective infrasound signal to generate several sub-bands, and calculating the energy of each sub-band; S42, calculating the energy percentage of all sub-bands and sorting them from largest to smallest energy percentage; S43, accumulating the energy percentages, extracting sub-bands with a cumulative percentage > 80% or extracting sub-bands of 11-14Hz, and merging them into a main frequency band; S44, calculating the center frequency f_current of the main frequency band, comparing it with the reference center frequency f_benchmark when there is no typhoon, and obtaining the center frequency offset Δf = f_current - f_benchmark; S45, calculating the total energy E of the main frequency band. 主 Total energy E across the entire frequency band （8-20）Hz The proportion of the frequency band energy is used to obtain the frequency band energy ratio R=E. 主 / E （8-20）Hz .

[0007] In this invention, preferably, step S5 includes: S51, removing outliers from the air pressure data, wind speed data, and temperature data; S52, filling in missing values ​​from the air pressure data, wind speed data, and temperature data; and S53, performing consistency verification on the air pressure data, wind speed data, and temperature data, with the data that passes the verification being accurate data.

[0008] In this invention, preferably, in step S6, the pressure gradient The calculation method is as follows: calculate the straight-line distance D_ij between adjacent nodes, calculate the pressure difference ΔP_ij between adjacent nodes, and obtain the pressure gradient of each adjacent node. _ij = ΔP_ij / D_ij, and the maximum value is taken as the pressure gradient for this sampling period. The wind shear index α is calculated as follows: for each node, obtain the wind speed v at a height of 10m. (Note: The original text contains some formatting errors and inconsistencies. A more accurate translation would require the full context.) 10 Wind speed at 100m altitude v 100 α is calculated using the logarithmic wind profile model, where α = ln(v 100 / v 10 ) / ln (100 / 10) = ln (v 100 / v 10 ) / 2.303, if v 10 =0, then α takes the value 0.25, if v 100 ≤v 10 If so, wind speed data will be collected again.

[0009] In this invention, preferably, S7 includes: S71, extracting the collection timestamps of infrasound features and meteorological parameters, and uniformly converting them into the format of "year-month-day-hour:minute:second"; S72, if the difference between the timestamps of infrasound features and meteorological parameters in the same collection period is greater than 1 minute, it is determined as "time inaccuracy" and the data of that period is discarded.

[0010] In this invention, preferably, it further includes: S10, introducing a Kalman filter to fuse satellite cloud image data, and correcting the errors in the final coordinates of the typhoon center and the final typhoon movement speed through dynamic weight allocation.

[0011] In this invention, preferably, it further includes: S11, if the path deviation angle is >25° or the speed increment is >10km / h, a first-level warning is triggered, prompting the wind turbines of the offshore wind farm to shut down and the ships to evacuate urgently; S12, if 15° < path deviation angle ≤25° or 5km / h < speed increment ≤10km / h, a second-level warning is triggered, prompting the wind turbines to start operating at reduced power and notifying the ships to prepare for evacuation.

[0012] A typhoon path inversion system based on the spectral characteristics of ocean background noise includes: a data acquisition module for acquiring noise and meteorological data from at least three nodes, performed every 5 minutes; the noise data includes 8-20Hz infrasound signals, and the meteorological data includes air pressure, wind speed, and temperature data; a sound signal filtering module for filtering the noise data to obtain valid infrasound signals; a sound signal analysis module for performing spectral analysis and feature extraction on the valid infrasound signals to obtain infrasound features, including center frequency offset and frequency band energy ratio; and a meteorological filtering module for filtering and verifying meteorological data to obtain accurate air pressure, wind speed, and temperature data. Temperature data; a meteorological calculation module, used to calculate the pressure gradient and wind shear index based on accurate air pressure and wind speed data, as meteorological parameters; a timestamp unification module, used to unify the timestamps of infrasound features and meteorological parameters; a neural network module, used to input the timestamp-unified infrasound features and meteorological parameters as feature values ​​into the LSTM-Attention feature fusion model, the LSTM-Attention feature fusion model outputs the coordinates of the typhoon center and the typhoon's moving speed; an output module, used to calculate the average of the typhoon coordinates and typhoon moving speed obtained from each node, as the final coordinates of the typhoon center and the final typhoon moving speed; a correction module, used to introduce... The system integrates Kalman filtering and satellite cloud image data, and corrects errors in the final coordinates and speed of the typhoon center through dynamic weight allocation. An early warning module triggers a Level 1 warning if the path deviation angle is greater than 25° or the speed increment is greater than 10 km / h, prompting offshore wind turbines to shut down and vessels to evacuate immediately. A Level 2 warning is triggered if 15° < path deviation angle ≤ 25° or 5 km / h < speed increment ≤ 10 km / h, prompting wind turbines to operate at reduced power and vessels to prepare for evacuation. The neural network module includes an LSTM-Attention feature fusion model that outputs the typhoon center coordinates and speed based on infrasound features and meteorological parameters. The process involves: converting the time-series data format of infrasound features and meteorological parameters, performing data normalization verification and adjustment to obtain a data tensor; inputting the data tensor into the first-layer LSTM for feature extraction, outputting the first-layer LSTM feature vector; inputting the data tensor into the second-layer LSTM for feature extraction, outputting the deep feature vector; calculating attention weights, normalizing the attention weights, and weighting the features; concatenating the attention output with the deep feature vector to obtain the fused feature vector; initializing a fully connected network, inputting the fused feature vector into the fully connected network, performing fully connected layer feature mapping, and outputting the coordinates of the typhoon center and the typhoon's movement speed.

[0013] In this invention, preferably, the acoustic signal filtering module includes: an invalid rejection unit for rejecting invalid data and correcting the amplitude of the infrasound signal based on the ambient temperature at the time of acquisition; a denoising unit for processing the infrasound signal using a wavelet threshold denoising algorithm; and a verification unit for verifying the signal-to-noise ratio of the infrasound signal and correcting infrasound signal segments with a signal-to-noise ratio higher than the threshold to obtain valid infrasound signals. The acoustic signal analysis module includes: a decomposition unit for performing wavelet packet decomposition on the valid infrasound signal to generate several sub-bands and calculating the energy of each sub-band; a statistics unit for calculating the energy percentage of all sub-bands and sorting them from largest to smallest energy percentage; a merging unit for accumulating the energy percentage, extracting sub-bands with a cumulative percentage > 80% or extracting sub-bands of 11-14Hz, and merging them into a main frequency band; and a center frequency offset calculation unit for calculating the center frequency f_current of the main frequency band and comparing it with the reference center frequency f_benchmark when there is no typhoon to obtain the center frequency offset Δf = f_current - f_benchmark; Bandwidth energy ratio calculation unit, used to calculate the total energy E of the main frequency band. 主 Total energy E across the entire frequency band （8-20）Hz The proportion of the frequency band energy is used to obtain the frequency band energy ratio R=E. 主 / E （8-20）Hz The meteorological screening module includes: an anomaly removal unit for removing outliers from air pressure, wind speed, and temperature data; a missing value filling unit for filling in missing values ​​from air pressure, wind speed, and temperature data; and a verification unit for performing consistency verification on the air pressure, wind speed, and temperature data, with data passing the verification being considered accurate. In the meteorological calculation module, the air pressure gradient... The calculation method is as follows: calculate the straight-line distance D_ij between adjacent nodes, calculate the pressure difference ΔP_ij between adjacent nodes, and obtain the pressure gradient of each adjacent node. _ij = ΔP_ij / D_ij, and the maximum value is taken as the pressure gradient for this sampling period. The wind shear index α is calculated as follows: for each node, obtain the wind speed v at a height of 10m. (Note: The original text contains some formatting errors and inconsistencies. A more accurate translation would require the full context.) 10 Wind speed at 100m altitude v 100 α is calculated using the logarithmic wind profile model, where α = ln(v 100 / v 10 ) / ln (100 / 10) = ln (v 100 / v 10 ) / 2.303, if v 10 =0, then α takes the value 0.25, if v 100 ≤v 10If the wind speed data is not collected, the time stamp unification module includes: a unification unit, used to extract the collection timestamps of infrasound features and meteorological parameters, and uniformly convert them into the format of "year-month-day-hour:minute:second"; and an inaccuracy elimination unit, used to determine that if the difference between the timestamps of infrasound features and meteorological parameters in the same collection period is greater than 1 minute, it is considered to be time inaccuracy and the data of that period is eliminated.

[0014] Compared with the prior art, the beneficial effects of the present invention are: The method and system of this invention establish a physical coupling mechanism of "typhoon infrasound-pressure gradient-sea surface wind field" by coupling the infrasound characteristics of ocean background noise (8–20Hz) with meteorological parameters, enabling early warning of typhoon path deviation 6–12 hours in advance; it utilizes 8–20Hz ocean background noise infrasound (low attenuation, long-distance propagation characteristics) to supplement nearshore monitoring blind spots; it couples infrasound spectral characteristics with meteorological parameters to construct multi-dimensional data input, improving the model's comprehensive representation ability of typhoon dynamics; and it adopts a hybrid architecture of LSTM-Attention + ensemble Kalman filtering to balance time-series dependency capture and prediction bias correction, achieving real-time inversion and providing emergency response decision support for wind farms in "instantaneous strong wind events" (such as wind turbine shutdown and ship evacuation). Attached Figure Description

[0015] Figure 1 This is a flowchart of a typhoon path inversion method based on the spectral characteristics of ocean background noise, according to an embodiment of the present invention.

[0016] Figure 2 This is a flowchart of S3 in a typhoon path inversion method based on the spectral characteristics of ocean background noise according to an embodiment of the present invention.

[0017] Figure 3 This is a flowchart of S4 in a typhoon path inversion method based on the spectral characteristics of ocean background noise according to an embodiment of the present invention.

[0018] Figure 4 This is a flowchart of step S5 in a typhoon path inversion method based on the spectral characteristics of ocean background noise according to an embodiment of the present invention.

[0019] Figure 5 This is a flowchart of S7 in a typhoon path inversion method based on the spectral characteristics of ocean background noise according to an embodiment of the present invention.

[0020] Figure 6 This is a flowchart of step S8 in a typhoon path inversion method based on the spectral characteristics of ocean background noise according to an embodiment of the present invention.

[0021] Figure 7 This is a flowchart of another embodiment of the present invention, which is a typhoon path inversion method based on the spectral characteristics of ocean background noise.

[0022] Figure 8 This is a schematic diagram of the structure of a typhoon path inversion system based on the spectral characteristics of ocean background noise, according to another embodiment of the present invention.

[0023] In the attached diagram: 1. Acquisition module; 2. Sound signal filtering module; 3. Sound signal analysis module; 4. Meteorological filtering module; 5. Meteorological calculation module; 6. Timestamp unification module; 7. Neural network module; 8. Output module; 9. Correction module; 10. Early warning module. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] It should be noted that when a component is described as "fixed to" another component, it can be directly on the other component or may have a component in between. When a component is considered "connected to" another component, it can be directly connected to the other component or may have a component in between. When a component is considered "set on" another component, it can be directly set on the other component or may have a component in between. The terms "vertical," "horizontal," "left," "right," and similar expressions used in this document are for illustrative purposes only.

[0026] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0027] Please see Figure 1 A preferred embodiment of the present invention provides a typhoon path inversion method based on the spectral characteristics of ocean background noise, comprising: Typhoon track inversion methods based on ocean background noise spectral characteristics include: S1, Construct an LSTM-Attention feature fusion model.

[0028] The LSTM-Attention feature fusion model takes four-dimensional temporal features as input and the coordinates (longitude and latitude) of the typhoon center and the typhoon's moving speed as output, and adopts a two-layer LSTM+Attention architecture.

[0029] Input layer structure: Based on a time step of timesteps=12 (corresponding to 1 hour of historical data, 5 minutes / time * 12 times = 60 minutes), the standardized four-dimensional features (Δf, R, ...) are processed. The input matrix (α) is reconstructed into a temporal input matrix with dimensions [batch_size, 12, 4] (batch_size=32, optimized based on GPU memory). Here, Δf is the center frequency offset, and R is the band energy ratio. α represents the pressure gradient, and α represents the wind shear index.

[0030] The first layer LSTM (64 units) structure is used to extract the short-term temporal dependence of a single feature (such as the continuous shift trend of Δf in 1 hour); the activation function is tanh (hidden layer) + sigmoid (gating), dropout=0.2 (to prevent overfitting), and returns the hidden state of each time step (dimension [32,12,64]).

[0031] The second-layer LSTM (32-unit) structure is used to capture deep dependencies between multiple features (such as Δf and Δf). (Co-changes); the activation function is in the same first layer as dropout, the input is the output of the first layer LSTM, and the return is the hidden state of the final time step (dimension [32,32]).

[0032] Enhanced Attention Mechanism: Its function is to focus on the moment of Δf abrupt change (a precursor to typhoon turning, where Δf abrupt change is defined as the absolute value of the difference between adjacent time steps ≥ 2Hz); its calculation process is as follows: (1) Define the Query vector (32-dimensional, trainable), the Key vector (output of the first LSTM layer, [32, 12, 64]), and the Value vector (consistent with the Key); (2) Calculate the attention score: score(Q,K_i)=W_v・tanh(W_q・Q + W_k・K_i) (W_q / W_k / W_v is the trainable weight matrix); (3) Weight enhancement at mutation time: the fraction of the Δf mutation time step is multiplied by a factor of 1.5; (4) Normalization and weighting: softmax normalizes the weights, and weighted summation is used to obtain the Attention output (dimension [32,32]).

[0033] Output layer: Fully connected network (16-unit ReLU activation → 3-unit linear output), outputting the longitude (λ, °), latitude (φ, °), and typhoon movement speed (v, km / h) of the typhoon center for the next 1-6 hours.

[0034] Training data: Typhoon monitoring data of a wind farm for one year (365 days) (including more than 10,000 time series samples).

[0035] Training parameters: MSE loss function (suitable for continuous value prediction), Adam optimizer (initial learning rate 0.001, decaying by 10% every 5 epochs), early stopping strategy (stop if the MSE on the validation set does not decrease for 5 consecutive epochs, and save the optimal model). Model deployment: Exported in ONNX format and deployed to edge computing nodes (such as the server in the control room of a wind farm), with a single sample inference time of ≤0.1 seconds.

[0036] S2 collects noise and meteorological data from at least 3 nodes every 5 minutes; the noise data includes 8-20Hz infrasound signals, and the meteorological data includes air pressure, wind speed and temperature data.

[0037] In this embodiment, the application scenario is an offshore wind farm, and data is collected using four nodes. Four noise measuring instruments and four weather stations are placed in a corner of the wind farm area to form a noise measurement and meteorological parameter acquisition matrix. The noise measuring instruments (sampling rate 100Hz, center frequency 10Hz) capture 8-20Hz infrasound waves; the weather stations monitor air pressure, temperature, humidity and wind speed data.

[0038] Start the noise measuring instruments at all four nodes, ensuring the equipment is in an unobstructed and electromagnetic interference-free state, and collect 8-20Hz infrasound signals in real time. Automatically save the time-domain data every 5 minutes, and record the ambient temperature at the time of acquisition (for subsequent signal temperature compensation).

[0039] Four multi-parameter weather stations were simultaneously activated, with 10m / 100m height anemometers, barometric pressure sensors, and temperature and humidity sensors operating concurrently. After data collection, the system automatically marked data integrity (e.g., "complete," "missing wind speed," "missing barometric pressure") to facilitate subsequent filtering.

[0040] S3 filters and processes the noise data to obtain the effective infrasound signal.

[0041] Specifically, such as Figure 2 As shown, S3 includes: S31, Remove invalid data and correct the amplitude of the infrasound signal based on the ambient temperature at the time of acquisition.

[0042] Remove invalid data, i.e. delete data that was collected at the time of equipment failure (such as sensor offline, signal amplitude of 0) or data that exceeded the environmental interference limit (such as strong vibration signal of 1-5Hz generated by sudden failure of fan). The amplitude of the infrasound signal is corrected based on the ambient temperature at the time of acquisition. For every 1°C change in temperature, the signal amplitude correction factor is 0.98. The formula is: A_compensated = A_original × 0.98^(T-25), where T is the ambient temperature, A_compensated is the corrected value of the infrasound signal amplitude, and A_original is the original value of the infrasound signal amplitude. S32 uses a wavelet threshold denoising algorithm to process the infrasound signal.

[0043] The wavelet thresholding denoising algorithm is used, and the specific steps are as follows: (1) Select the db4 wavelet basis and perform 4-level wavelet decomposition on the 5-minute time domain signal (30,000 data points) to obtain the approximation coefficients and detail coefficients; (2) Apply a “soft threshold” to the detail coefficients (threshold = 0.5 × max (absolute value of detail coefficients)) to suppress high-frequency noise (such as 50Hz power frequency interference); (3) Reconstructing the signal: Perform inverse wavelet transform on the processed approximation coefficients and detail coefficients to obtain the denoised 8-20Hz pure infrasound signal.

[0044] S33, verify the signal-to-noise ratio of the infrasound signal, correct the infrasound signal segments that are higher than the signal-to-noise ratio threshold, and obtain the effective infrasound signal.

[0045] Specifically, the denoised 5-minute signal is divided into 1-minute segments to obtain 5 sub-signals (6000 data points per segment). The signal-to-noise ratio (SNR) of each sub-signal is calculated. Sub-signals with an SNR < 10dB are considered invalid and replaced with the average of the two valid sub-signals before and after them to ensure that each signal segment meets the requirements of subsequent spectrum analysis.

[0046] S4 performs spectral analysis and feature extraction on the effective infrasound signal to obtain infrasound features, which include center frequency offset and bandwidth energy ratio.

[0047] Specifically, such as Figure 3 As shown, S4 includes: S41, perform wavelet packet decomposition on the effective infrasound signal to generate several sub-bands, and calculate the energy of each sub-band.

[0048] In wavelet packet decomposition, each effective sub-signal segment undergoes 6 levels of wavelet packet decomposition, generating 2^6=64 sub-bands covering a frequency range of 8-20Hz. The energy of each sub-band is calculated (formula: E=Σ|x_i|). 2 (where x_i is the signal sampling point within the sub-band).

[0049] S42, calculate the energy percentage of all sub-bands and sort them from largest to smallest energy percentage.

[0050] S43, cumulative energy percentage, extract sub-bands with a cumulative percentage >80% or extract sub-bands of 11-14Hz, and merge them into the main frequency band.

[0051] In practical applications, the 11-14Hz frequency band can be used as the default main frequency band. If the cumulative proportion is not met, the frequency band range can be dynamically adjusted.

[0052] S44. Calculate the center frequency f_current of the main frequency band and compare it with the reference center frequency f_benchmark when there is no typhoon to obtain the center frequency offset Δf = f_current - f_benchmark.

[0053] S45, Calculate the total energy E of the main frequency band. 主 Total energy E across the entire frequency band （8-20）Hz The proportion of the frequency band energy is used to obtain the frequency band energy ratio R=E. 主 / E （8-20）Hz .

[0054] S5 filters and verifies meteorological data to obtain accurate air pressure, wind speed, and temperature data.

[0055] Specifically, such as Figure 4 As shown, S5 includes: S51 removes outliers from air pressure, wind speed, and temperature data.

[0056] Using the three-standard-deviation principle, data with air pressure greater than 1050 hPa or less than 950 hPa, wind speed greater than 50 m / s, temperature less than -10℃ or greater than 40℃, and humidity less than 0% or greater than 100% were deleted.

[0057] S52 fills in missing values ​​in air pressure, wind speed, and temperature data.

[0058] The rules for filling in the gaps are: (1) If a parameter is missing in a single node (such as air pressure), calculate the “distance-weighted average” of that parameter for the other 3 nodes (weight = 1 / straight-line distance between nodes, sum to 1). (2) If multiple nodes are missing the same parameter (less than or equal to 2 nodes), use the "time series interpolation method" (fitting based on the effective data of the 5 collection cycles before and after the node); (3) If the same parameter is missing in ≥3 nodes, mark the collection period as “invalid meteorological data”, suspend the calculation of subsequent meteorological parameters, and wait for the next collection period.

[0059] S53 performs consistency checks on air pressure, wind speed, and temperature data; data that passes the check is considered accurate.

[0060] Specifically, compare the wind speed at 10m and 100m altitudes at the same node. Under normal circumstances, the wind speed at 100m should be greater than that at 10m (difference > 0.5m / s). If this is not the case, re-collect the wind speed data for that node. Compare the air pressure at adjacent nodes. The normal range for the difference is 0.1~2hPa. If it exceeds this range, check the weather station for leaks, obstructions, or other problems. After repairing these issues, re-collect the data.

[0061] S6 calculates the pressure gradient and wind shear index based on accurate air pressure and wind speed data, which are used as meteorological parameters.

[0062] Specifically, pressure gradient The calculation method is as follows: Calculate the straight-line distance D_ij between adjacent nodes, calculate the pressure difference ΔP_ij between adjacent nodes, and obtain the pressure gradient of each adjacent node. _ij = ΔP_ij / D_ij, and the maximum value is taken as the pressure gradient for this sampling period. Keep two decimal places.

[0063] The wind shear index α is calculated as follows: For each node, obtain the wind speed v at a height of 10m. 10 Wind speed at 100m altitude v 100 α is calculated using the logarithmic wind profile model, where α = ln(v) 100 / v 10 ) / ln (100 / 10) = ln (v 100 / v 10 ) / 2.303, if v 10 =0, then α takes the value 0.25, if v 100 ≤v 10 If so, wind speed data will be collected again.

[0064] S7 unifies the timestamps for infrasound characteristics and meteorological parameters.

[0065] Specifically, such as Figure 5 As shown, S7 includes: S71, extract the collection timestamps of infrasound features and meteorological parameters, and convert them into a unified format of "year-month-day-hour:minute:second".

[0066] S72. If the difference between the timestamp of the infrasound characteristics and the meteorological parameters is greater than 1 minute within the same acquisition cycle, it is judged as "time inaccuracy" and the data of that cycle is removed.

[0067] S8 inputs the infrasound features and meteorological parameters after unifying the timestamps as feature values ​​into the LSTM-Attention feature fusion model. The LSTM-Attention feature fusion model outputs the coordinates of the typhoon center and the typhoon's movement speed.

[0068] Specifically, such as Figure 6 As shown in Figure S8, the process by which the LSTM-Attention feature fusion model outputs the coordinates of the typhoon center and the typhoon's movement speed based on infrasound features and meteorological parameters includes: S81 performs time-series data format conversion on infrasound characteristics and meteorological parameters, and performs data normalization verification and adjustment to obtain data tensors.

[0069] Receive preprocessed four-dimensional feature data (Δf, R, Each feature contains continuous time-series data. The feature data is divided into fixed-length time-series segments according to a timestep of timesteps=12 (corresponding to 1 hour of historical data, with 1 data point every 5 minutes). An input data tensor is constructed with the shape [number of samples, timesteps=12, number of features=4], where each sample represents a feature sequence of 12 consecutive time steps.

[0070] Check the normalization of the input data to ensure that all feature values ​​are within the range [-3, 3] (compliant with Z-Score normalization). Truncate outliers that are outside the range (values ​​less than -3 are set to -3, and values ​​greater than 3 are set to 3). Convert the input data to a tensor format that the model can process (such as a Tensor in PyTorch or a TensorFlow Tensor), and set the data type to float32.

[0071] S82 inputs the data tensor into the first-layer LSTM for feature extraction and outputs the first-layer LSTM feature vector.

[0072] Initialize the first layer of the LSTM network (64 neurons) and set the following parameters: Input dimension: 4 (corresponding to 4 features); Hidden layer dimension: 64; Activation function: tanh function is used for hidden states, sigmoid is used for gated units, and Dropout rate is 0.2 (to prevent overfitting). Return sequence: True (Returns the output of each time step, used for subsequent Attention mechanisms).

[0073] Input the input tensor [number of samples, 12, 4] into the first layer of the LSTM and perform forward propagation computation: calculate the activation values ​​of the forget gate, input gate and output gate to dynamically control the retention and updating of information; update the cell state and hidden state to capture the short-term temporal dependencies of the feature sequence.

[0074] The output of the first LSTM layer is in the shape of [number of samples, 12, 64], which means that each time step corresponds to a 64-dimensional feature vector, i.e., the feature vector of the first LSTM layer.

[0075] S83 inputs the data tensor into the second LSTM layer for feature extraction and outputs a deep feature vector.

[0076] Initialize the second LSTM network (32 neurons) with the following parameters: Input dimension: 64 (receives the output of the first LSTM layer); Hidden layer dimension: 32; Activation function: Same as the first layer (tanh + sigmoid); Dropout rate: 0.2; Return sequence: False (only returns the output of the last time step).

[0077] The output of the first LSTM layer [number of samples, 12, 64] is input into the second LSTM layer for forward propagation computation: further learning the long-term dependencies in the feature sequence and integrating the correlation between multiple features; compressing the feature dimension, reducing the 64-dimensional features to 32-dimensional features.

[0078] The output of the second LSTM layer is in the shape of [number of samples, 32], that is, each sample corresponds to a 32-dimensional deep feature vector.

[0079] S84 performs Attention weight calculation, normalizes the attention weights, performs feature weighting, and concatenates the Attention output with the deep feature vector to obtain the fused feature vector.

[0080] Define the key vector for the Attention mechanism: Query vector: A randomly initialized 32-dimensional trainable vector representing the type of features the model is interested in; Key vector: The output of the first LSTM layer [number of samples, 12, 64] represents the feature information at each time step; Value vector: Same as the key vector, i.e., the output of the first LSTM layer [number of samples, 12, 64].

[0081] Calculate the similarity (attention score) between the Query and the Key at each time step: map the Query and Key to the same dimension through a linear transformation; calculate the score using additive attention.

[0082] Enhance the attention score at the abrupt change in Δf: detect the abrupt change in Δf in the input sequence (absolute value of the difference between adjacent time steps Δf ≥ 2Hz); multiply the attention score at the abrupt change by a weight of 1.5 to enhance the influence of the key signal.

[0083] The attention scores at 12 time steps are normalized using the softmax function to obtain the weight value for each time step, with the shape being [number of samples, 12] and the sum of the weights being 1. The Value vector is weighted and summed based on the attention weights to obtain the output of the Attention mechanism, with the shape being [number of samples, 64]. The Attention output is concatenated with the output of the second LSTM layer to obtain the fused feature vector, with the shape being [number of samples, 64+32=96].

[0084] S85 initializes the fully connected network, inputs the fused feature vector into the fully connected network, performs fully connected layer feature mapping, and outputs the coordinates of the typhoon center and the typhoon's movement speed.

[0085] Specifically, initialize the fully connected network: First layer (hidden layer): 16 neurons, using the ReLU activation function; Second layer (output layer): 3 neurons, no activation function (output continuous value).

[0086] Input the fused feature vector [number of samples, 96] into the fully connected network: Hidden layer computation: Mapping 96-dimensional features to 16-dimensional features through linear transformation and ReLU activation; Output layer computation: Maps the 16-dimensional features to 3 dimensions, corresponding to 3 prediction targets; Output the prediction results, in shape [number of samples, 3], representing: Longitude of the typhoon center (°) in the next 1-6 hours; Latitude (°) of the typhoon center in the next 1-6 hours; Typhoon's movement speed (km / h) in the next 1-6 hours.

[0087] S9 calculates the average of the typhoon coordinates and typhoon movement speed obtained from each node, and uses it as the final coordinates and final typhoon movement speed of the typhoon center.

[0088] In a preferred embodiment of the present invention, such as Figure 7 As shown, the typhoon track inversion method based on the spectral characteristics of ocean background noise also includes: S10 introduces ensemble Kalman filtering to fuse satellite cloud image data, and corrects the errors in the final coordinates of the typhoon center and the final typhoon movement speed through dynamic weight allocation.

[0089] During periods of heavy rainfall within the eyewall of a typhoon, model predictions are prone to bias and require correction by incorporating satellite cloud imagery data (updated every 30 minutes). (1) The LSTM-Attention prediction results (i.e., the final coordinates of the typhoon center and the final typhoon movement speed) are used as the initial values, and the typhoon position extracted from the satellite cloud image is used as the observation value. (2) Calculate the covariance of prediction error and the covariance of observation error, and dynamically allocate weights (satellite data has a weight of 0.3 when there is no rainfall and a weight of 0.5 when there is rainfall). (3) Output the corrected typhoon path to ensure that the prediction error is ≤20km.

[0090] In a preferred embodiment of the present invention, such as Figure 7 As shown, the typhoon track inversion method based on the spectral characteristics of ocean background noise also includes: S11 If the path deviation angle is greater than 25° or the speed increment is greater than 10km / h, a Level 1 warning will be triggered, prompting the wind turbines of the offshore wind farm to shut down and the ships to evacuate urgently.

[0091] S12, if 15° < path offset angle ≤ 25° or 5km / h < speed increment ≤ 10km / h, a level 2 warning is triggered, prompting the wind turbines to start operating at reduced power and notifying ships to prepare for evacuation.

[0092] Another embodiment of the present invention also provides a typhoon path inversion system based on the spectral characteristics of ocean background noise, such as... Figure 8 As shown, it includes: Acquisition module 1 is used to collect noise data and meteorological data from at least 3 nodes, once every 5 minutes; the noise data includes 8-20Hz infrasound signals, and the meteorological data includes air pressure data, wind speed data, and temperature data.

[0093] The acoustic signal filtering module 2 is used to filter noise data to obtain effective infrasound signals.

[0094] The sound signal filtering module 2 includes: The invalid rejection unit is used to reject invalid data and correct the amplitude of the infrasound signal based on the ambient temperature at the time of acquisition. The denoising unit is used to process the infrasound signal using a wavelet threshold denoising algorithm; The verification unit is used to verify the signal-to-noise ratio of the infrasound signal, and to correct the infrasound signal segments that are higher than the signal-to-noise ratio threshold to obtain a valid infrasound signal.

[0095] The acoustic signal analysis module 3 is used to perform spectral analysis and feature extraction on the effective infrasound signal to obtain infrasound features, including center frequency offset and frequency band energy ratio.

[0096] The acoustic signal analysis module 3 includes: The decomposition unit is used to perform wavelet packet decomposition on the effective infrasound signal, generate several sub-bands, and calculate the energy of each sub-band. The statistical unit is used to calculate the energy percentage of all sub-bands and sort them from largest to smallest energy percentage. The merging unit is used to accumulate the energy percentage, extract sub-bands with a cumulative percentage >80% or extract sub-bands of 11-14Hz, and merge them into the main frequency band; The center frequency offset calculation unit is used to calculate the center frequency f_current of the main frequency band and compare it with the reference center frequency f_benchmark when there is no typhoon to obtain the center frequency offset Δf = f_current - f_benchmark. The frequency band energy ratio calculation unit is used to calculate the total energy E of the main frequency band. 主 Total energy E across the entire frequency band （8-20）Hz The proportion of the frequency band energy is used to obtain the frequency band energy ratio R=E. 主 / E （8-20）Hz .

[0097] Meteorological filtering module 4 is used to filter and verify meteorological data to obtain accurate air pressure data, wind speed data, and temperature data.

[0098] Weather screening module 4 includes: The anomaly removal unit is used to remove outliers from air pressure, wind speed, and temperature data. Missing data filling unit, used to fill in missing values ​​in air pressure data, wind speed data, and temperature data; The verification unit is used to verify the consistency of air pressure data, wind speed data, and temperature data. Data that passes the verification is considered accurate.

[0099] Meteorological calculation module 5 is used to calculate the pressure gradient and wind shear index based on accurate air pressure and wind speed data, as meteorological parameters.

[0100] In meteorological calculation module 5, pressure gradient The calculation method is as follows: Calculate the straight-line distance D_ij between adjacent nodes, calculate the pressure difference ΔP_ij between adjacent nodes, and obtain the pressure gradient of each adjacent node. _ij = ΔP_ij / D_ij, and the maximum value is taken as the pressure gradient for this sampling period. Round to two decimal places; The wind shear index α is calculated as follows: For each node, obtain the wind speed v at a height of 10m. 10 Wind speed at 100m altitude v 100 α is calculated using the logarithmic wind profile model, where α = ln(v) 100 / v 10 ) / ln (100 / 10) = ln (v 100 / v 10 ) / 2.303, if v 10 =0, then α takes the value 0.25, if v 100 ≤v 10 If so, wind speed data will be collected again.

[0101] The timestamp unification module 6 is used to unify the timestamps of infrasound characteristics and meteorological parameters.

[0102] Timestamp unification module 6 includes: A unified unit is used to extract the acquisition timestamps of infrasound features and meteorological parameters, and convert them into a unified format of "year-month-day-hour:minute:second". The inaccuracy elimination unit is used to determine that if the difference between the timestamp of the infrasound characteristics and the meteorological parameters is greater than 1 minute within the same acquisition period, it is considered to be time inaccuracy and the data of that period is eliminated.

[0103] Neural network module 7 is used to input the infrasound features and meteorological parameters after unifying the timestamps as feature values ​​into the LSTM-Attention feature fusion model. The LSTM-Attention feature fusion model outputs the coordinates of the typhoon center and the typhoon's movement speed.

[0104] Output module 8 is used to calculate the average value of the typhoon coordinates and typhoon movement speed obtained from each node, and use it as the final coordinates and final typhoon movement speed of the typhoon center.

[0105] Correction module 9 is used to introduce ensemble Kalman filter fusion satellite cloud image data and correct the errors in the final coordinates of the typhoon center and the final typhoon movement speed through dynamic weight allocation.

[0106] The early warning module 10 is used to trigger a Level 1 early warning if the path deviation angle is greater than 25° or the speed increment is greater than 10km / h, prompting the wind turbines of the offshore wind farm to shut down and the ships to evacuate urgently; if 15° < path deviation angle ≤ 25° or 5km / h < speed increment ≤ 10km / h, it triggers a Level 2 early warning, prompting the wind turbines to start operating at reduced power and notifying the ships to prepare for evacuation.

[0107] In neural network module 7, the process by which the LSTM-Attention feature fusion model outputs the coordinates of the typhoon center and the typhoon's movement speed based on infrasound features and meteorological parameters includes: The infrasound characteristics and meteorological parameters are converted into time-series data formats, and the data is normalized, verified, and adjusted to obtain data tensors; Input the data tensor into the first LSTM layer for feature extraction, and output the first LSTM layer feature vector. The data tensor is input into the second LSTM layer for feature extraction, and the deep feature vector is output. The attention weights are calculated, normalized, and feature weights are applied. The attention output is then concatenated with the deep feature vector to obtain the fused feature vector. Initialize the fully connected network, input the fused feature vector into the fully connected network, perform feature mapping of the fully connected layer, and output the coordinates of the typhoon center and the typhoon's moving speed.

[0108] The above description is a detailed description of the preferred embodiments of the present invention. However, the embodiments are not intended to limit the scope of the patent application of the present invention. All equivalent changes or modifications made under the technical spirit of the present invention should fall within the patent scope covered by the present invention.

Claims

1. A method for retrieving typhoon track based on spectral characteristics of ocean ambient noise, characterized in that, The application relates to a method for tracking a typhoon center, which comprises the following steps: S1, constructing an LSTM-Attention feature fusion model; S2, collecting noise data and meteorological data of at least three nodes, and performing the collection once every 5 minutes; the noise data comprises 8-20Hz infrasound signals, and the meteorological data comprises air pressure data, wind speed data and temperature data; S3, performing screening processing on the noise data to obtain effective infrasound signals; S4, performing frequency spectrum analysis and feature extraction on the effective infrasound signals to obtain infrasound features, wherein the infrasound features comprise a center frequency offset and a frequency band energy ratio; S5, performing screening and verification on the meteorological data to obtain accurate air pressure data, wind speed data and temperature data; S6, calculating an air pressure gradient and a wind shear index according to the accurate air pressure data and the wind speed data, and taking the air pressure gradient and the wind shear index as meteorological parameters; S7, performing time stamp unification on the infrasound features and the meteorological parameters; S8, inputting the infrasound features and the meteorological parameters after time stamp unification as feature values into the LSTM-Attention feature fusion model, and outputting coordinates of a typhoon center and a typhoon moving speed by the LSTM-Attention feature fusion model; S9, calculating average values of the typhoon coordinates and the typhoon moving speed obtained by each node, and taking the average values as final coordinates of the typhoon center and a final typhoon moving speed; In the S8, the process that the LSTM-Attention feature fusion model outputs the coordinates of the typhoon center and the typhoon moving speed according to the infrasound features and the meteorological parameters comprises the following steps: S81, performing time sequence data format conversion on the infrasound features and the meteorological parameters, and performing data normalization verification and adjustment to obtain a data tensor; S82, inputting the data tensor into a first-layer LSTM to perform feature extraction, and outputting a first-layer LSTM feature vector; S83, inputting the data tensor into a second-layer LSTM to perform feature extraction, and outputting a deep-layer feature vector; S84, performing attention weight calculation, normalizing the attention weight, performing feature weighting, splicing an attention output result and the deep-layer feature vector, and obtaining a fused feature vector; S85, initializing a full connection network, inputting the fused feature vector into the full connection network, performing full connection layer feature mapping, and outputting the coordinates of the typhoon center and the typhoon moving speed.

2. The typhoon track inversion method according to claim 1, characterized in that, The S3 comprises the following steps: S31, removing invalid data, and correcting an infrasound signal amplitude according to an ambient temperature at a collection time; S32, processing the infrasound signal by adopting a wavelet threshold denoising algorithm; S33, verifying a signal-to-noise ratio of the infrasound signal, correcting an infrasound signal paragraph higher than a signal-to-noise ratio threshold, and obtaining effective infrasound signals.

3. The typhoon track inversion method of claim 1, wherein, The S4 comprises the following steps: S41, performing wavelet packet decomposition on the effective infrasound signals to generate a plurality of sub-bands, and calculating energy of each sub-band; S42, counting energy proportions of all the sub-bands, and sorting the sub-bands according to the energy proportions from large to small; S43, accumulating the energy proportions, extracting a sub-band with a cumulative proportion greater than 80% or extracting a 11-14Hz sub-band, and merging the sub-bands into a main frequency band. S44, calculate the center frequency f_current of the main frequency band, compare it with the benchmark center frequency f_benchmark in the absence of typhoon, and obtain the center frequency offset Δf = f_current - f_benchmark; S45, calculate the total energy E of the main frequency band 主 In the proportion of the total energy E of the full frequency band （8-20）Hz , get the frequency band energy ratio R = E 主 / E （8-20）Hz .

4. The typhoon track inversion method according to claim 1, wherein, The S5 comprises: S51, remove outliers in the air pressure data, wind speed data and temperature data; S52, fill in the missing values in the air pressure data, wind speed data and temperature data; S53, perform consistency check on the air pressure data, wind speed data and temperature data, and the data passing the check is accurate data.

5. The typhoon track inversion method according to claim 1, wherein, In the S6, barometric gradient is calculated as Calculate the straight-line distance D_ij between adjacent nodes, calculate the air pressure difference ΔP_ij between adjacent nodes, and obtain the air pressure gradient of each adjacent node _ij = ΔP_ij / D_ij, and take the maximum value as the air pressure gradient of the collection period ; The calculation method of the wind shear index a is: For each node, get the 10m height wind speed v 10 , 100m height wind speed v 100 , calculate a according to the logarithmic wind profile model, a = ln(v 100 / v 10 ) / ln (100 / 10) = ln(v 100 / v 10 ) / 2.303, if v 10 =0, then a takes 0.25, if v 100 ≤v 10 , then re-collect wind speed data.

6. The typhoon track inversion method according to claim 1, wherein, The S7 comprises: S71, extract the collection timestamps of the infrasound wave features and meteorological parameters, and convert them into the "year-month-day-hour-minute-second" format; S72, if the timestamp difference of the infrasound wave features and meteorological parameters in the same collection period is greater than 1 minute, it is determined that the time is inaccurate, and the data in this period is removed.

7. The typhoon track inversion method according to claim 1, wherein, Further comprising: S10, introduce ensemble Kalman filtering to fuse satellite cloud image data, and correct the errors of the final coordinates of the typhoon center and the final typhoon moving speed through dynamic weight distribution.

8. The typhoon track inversion method according to claim 7, characterized in that, Further comprising: S11, if the path offset angle is greater than 25° or the moving speed increment is greater than 10 km / h, a first-level warning is triggered, prompting the wind turbine of the offshore wind farm to shut down and the ship to evacuate urgently; S12, if 15° < path offset angle ≤ 25° or 5 km / h < moving speed increment ≤ 10 km / h, a second-level warning is triggered, prompting the wind turbine to start power reduction operation and notifying the ship to prepare for evacuation.

9. A typhoon track inversion system based on the spectral characteristics of ocean ambient noise, characterized in that, Comprise: A collection module for collecting noise data and meteorological data of at least three nodes, once every 5 minutes; the noise data includes 8-20Hz infrasound wave signal, and the meteorological data includes air pressure data, wind speed data and temperature data; An acoustic signal screening module for screening and processing the noise data to obtain effective infrasound wave signal; An acoustic signal analysis module for performing spectrum analysis and feature extraction on the effective infrasound wave signal to obtain infrasound wave features, the infrasound wave features including center frequency offset and frequency band energy ratio; A meteorological screening module for screening and checking the meteorological data to obtain accurate air pressure data, wind speed data and temperature data; A meteorological calculation module for calculating the air pressure gradient and the wind shear index as meteorological parameters according to the accurate air pressure data and wind speed data; A timestamp unification module for unifying the timestamps of the infrasound wave features and the meteorological parameters; A neural network module for inputting the timestamp-unified infrasound wave features and meteorological parameters as characteristic values into an LSTM-Attention feature fusion model, and outputting the coordinates of the typhoon center and the typhoon moving speed from the LSTM-Attention feature fusion model; An output module for averaging the typhoon coordinates and the typhoon moving speed obtained by each node to obtain the final coordinates of the typhoon center and the final typhoon moving speed; A correction module for introducing ensemble Kalman filtering to fuse satellite cloud image data, and correcting the errors of the final coordinates of the typhoon center and the final typhoon moving speed through dynamic weight distribution; The early warning module is configured to trigger a first-level early warning and prompt the wind turbine of the offshore wind farm to stop and the ship to evacuate in an emergency if the path offset angle is greater than 25° or the speed increment is greater than 10 km / h; and trigger a second-level early warning and prompt the wind turbine to start to operate at a reduced power and notify the ship to prepare to evacuate if 15° is less than the path offset angle and less than or equal to 25° or 5 km / h is less than the speed increment and less than or equal to 10 km / h. In the neural network module, the process in which the LSTM-Attention feature fusion model outputs the coordinates of the typhoon center and the moving speed of the typhoon according to the infrasound wave features and the meteorological parameters comprises: The infrasound wave features and the meteorological parameters are converted into a time series data format, and data normalization verification and adjustment are performed to obtain a data tensor; The data tensor is input into a first-layer LSTM for feature extraction, and a first-layer LSTM feature vector is output; The data tensor is input into a second-layer LSTM for feature extraction, and a deep-layer feature vector is output; Attention weight calculation is performed, the attention weights are normalized, feature weighting is performed, the Attention output result is spliced with the deep-layer feature vector, and a fused feature vector is obtained; A full-connection network is initialized, the fused feature vector is input into the full-connection network, full-connection layer feature mapping is performed, and the coordinates of the typhoon center and the moving speed of the typhoon are output.

10. The typhoon path inversion system according to claim 9, characterized in that, The acoustic signal screening module comprises: An invalid data elimination unit configured to eliminate invalid data and correct the amplitude of the infrasound wave signal according to the ambient temperature at the collection time; A denoising unit configured to process the infrasound wave signal by using a wavelet threshold denoising algorithm; A verification unit configured to verify the signal-to-noise ratio of the infrasound wave signal, correct the infrasound wave signal segment with a signal-to-noise ratio higher than a signal-to-noise ratio threshold, and obtain an effective infrasound wave signal; The acoustic signal analysis module comprises: A decomposition unit configured to decompose the effective infrasound wave signal by using a wavelet packet decomposition algorithm, generate a plurality of sub-bands, and calculate the energy of each sub-band; A statistical unit configured to sort all the sub-bands according to the energy proportion from large to small; A merging unit configured to accumulate the energy proportion, extract the sub-band with a cumulative proportion greater than 80% or extract the sub-band of 11-14 Hz, and merge the sub-bands into a main frequency band; A center frequency offset calculation unit configured to calculate the center frequency f_current of the main frequency band, compare the center frequency f_current with a benchmark center frequency f_benchmark in the absence of a typhoon, and obtain a center frequency offset Δf = f_current - f_benchmark; a band energy ratio calculating unit for calculating a total energy E of the main band 主 In the total energy E of the full band （8-20）Hz The proportion of the main band, the band energy ratio R = E 主 / E （8-20）Hz ; The meteorological screening module comprises: An abnormal value elimination unit configured to eliminate abnormal values in the pressure data, the wind speed data and the temperature data; A missing value filling unit configured to fill in the missing values in the pressure data, the wind speed data and the temperature data; A verification unit configured to perform consistency verification on the pressure data, the wind speed data and the temperature data, and the data passing the verification is accurate data; The meteorological calculation module, the pressure gradient is calculated in the following way: Calculate the straight-line distance D_ij between adjacent nodes, calculate the air pressure difference ΔP_ij between adjacent nodes, and obtain the air pressure gradient of each adjacent node _ij = ΔP_ij / D_ij, take the maximum value as the air pressure gradient of this collection period , keep 2 decimal places; The wind shear index α is calculated in the following manner: For each node, get the 10m height wind speed v 10 , 100m height wind speed v 100 , calculate a according to the logarithmic wind profile model, a = ln(v 100 / v 10 ) / ln (100 / 10) = ln(v 100 / v 10 ) / 2.303, if v 10 =0, then a takes 0.25, if v 100 ≤v 10 , then re-collect wind speed data; The timestamp unification module comprises: A unification unit is configured to extract the time stamp of the infrasound wave feature and the weather parameter, and unify the conversion into the "year-month-day-hour: minute: second" format; An error elimination unit is configured to determine that the time is inaccurate if the time stamp difference between the infrasound wave feature and the weather parameter in the same collection cycle is greater than 1 minute, and eliminate the data in the cycle.

Images