A precipitation estimation method, system, device and storage medium applied in low-altitude economic services

Abstract

The application provides a precipitation estimation method, system, device and storage medium applied in low-altitude economic services, and the method comprises the following steps: acquiring dual-polarization radar echo data and environmental parameters in a low-altitude atmosphere layer and taking the data and the parameters as inputs of a least mean square algorithm, combining a fuzzy logic control algorithm to dynamically adjust weight coefficients of an adaptive filter, using the adaptive filter with the adjusted weight to suppress noise of the dual-polarization radar echo data, and obtaining filtered radar echo data; extracting key precipitation characteristic parameters from the filtered radar echo data and calculating radar original precipitation estimation values; acquiring observation data of a ground meteorological station and performing fusion processing to generate regional precipitation reference data; using the regional precipitation reference data, determining correction coefficients by using a regularized linear regression method, correcting the radar original precipitation estimation values by using the correction coefficients, and outputting final precipitation estimation values. The application effectively improves the precipitation estimation precision in a low-altitude scene.

Description

Technical Field

[0001] This invention relates to the field of meteorological monitoring and radar technology, and in particular to a precipitation estimation method, system, equipment and storage medium for use in low-altitude economic services. Background Technology

[0002] With the rapid development of the low-altitude economy, low-altitude operations such as low-altitude tourism, logistics distribution, and agricultural and forestry plant protection are becoming increasingly frequent. These activities place extremely high demands on the accurate understanding of low-altitude meteorological conditions, especially precipitation. The uncertainty of precipitation intensity, range, and distribution directly affects low-altitude flight safety, cargo transportation efficiency, and the effectiveness of agricultural and forestry operations.

[0003] Traditional radar systems have many limitations in precipitation estimation. On the one hand, raw radar estimates often deviate from actual precipitation due to systematic errors and atmospheric propagation interference, resulting in poor estimation accuracy. On the other hand, facing the complex and ever-changing low-altitude atmospheric environment, such as differences in temperature, humidity, and air pressure in different regions, as well as rapidly changing wind conditions, traditional radars struggle to adjust data processing strategies in real time and dynamically. Noise interference severely affects the quality of echo data, making it impossible to obtain timely and accurate information on precipitation intensity and spatial distribution. This makes it difficult to meet the stringent requirements of high resolution, high precision, and high timeliness for meteorological information in low-altitude economic applications. Summary of the Invention This invention provides a precipitation estimation method, system, device, and storage medium for use in low-altitude economic services, to address the problems existing in related technologies. The technical solution is as follows: In a first aspect, embodiments of the present invention provide a precipitation estimation method applied in low-altitude economic services, comprising: Acquire dual-polarization radar echo data and environmental parameters within the lower atmosphere; The environmental parameters and signal characteristics of dual-polarization radar echo data are used as inputs to the least mean square algorithm. The weight coefficients of the adaptive filter are dynamically adjusted by combining the fuzzy logic control algorithm. The adaptive filter with adjusted weights is used to suppress noise in the dual-polarization radar echo data to obtain filtered radar echo data. Key precipitation characteristic parameters are extracted from the filtered radar echo data, and the original radar precipitation estimate is calculated based on the key precipitation characteristic parameters. This process involves acquiring observational data from multiple ground meteorological stations within the radar's coverage area, fusing this data, and generating regional precipitation baseline data corresponding to the radar's coverage area. The fusing process includes: establishing a state-space model of the precipitation data, comprising a state equation and an observation equation; the state equation describing the temporal variation of the regional precipitation baseline data, and the observation equation describing the linear relationship between the observed values ​​of each ground meteorological station and the regional precipitation baseline data; predicting the prior estimate of the regional precipitation state at the current moment based on the previous moment's regional precipitation baseline data using the state equation, and calculating the covariance matrix of the prediction error; acquiring observational data from all ground meteorological stations within the radar's coverage area at the current moment, calculating the Kalman gain matrix, using the Kalman gain matrix to weight and correct the prior estimate of the regional precipitation state, fusing the observational information from multiple ground meteorological stations to obtain the posterior estimate of the regional precipitation state at the current moment, updating the error covariance matrix, and finally obtaining the regional precipitation baseline data for the current moment. Using regional precipitation baseline data, a regularized linear regression method is used to determine the correction coefficients, and the original radar precipitation estimates are corrected using the correction coefficients to output the final precipitation estimates.

[0004] In one embodiment, the method further includes preprocessing the dual-polarization radar echo data; the preprocessing includes: Wavelet transform is performed on the dual-polarization radar echo data to obtain wavelet coefficients; Wavelet coefficients with absolute values ​​greater than or equal to the threshold are retained, while wavelet coefficients with absolute values ​​less than the threshold are set to zero. Then, the denoised echo data is reconstructed through inverse wavelet transform. A bandpass filter is used to filter the denoised echo data. The passband frequency range of the bandpass filter matches the frequency range of the precipitation echo signal in the dual-polarization radar echo data. This allows the precipitation echo signal to pass through while suppressing interference signals outside the passband range, resulting in preprocessed dual-polarization radar echo data, which is then used for subsequent noise suppression and extraction of key precipitation characteristic parameters.

[0005] In one implementation, key precipitation characteristic parameters include correlation coefficient, differential propagation phase shift, and differential propagation phase shift rate; the radar raw precipitation estimate calculated based on the key precipitation characteristic parameters includes: A precipitation inversion model based on neural network training is adopted, and the correlation coefficient, differential propagation phase shift and differential propagation phase shift rate among the key precipitation characteristic parameters are used as input features and fed into the precipitation inversion model. The precipitation inversion model adopts a multilayer perceptron structure, which consists of an input layer, multiple hidden layers and an output layer. The neurons in the hidden layers perform nonlinear transformation and feature extraction on the input information through activation functions. The precipitation inversion model outputs precipitation intensity values, which are then used as the raw precipitation estimates from the radar.

[0006] In one implementation, environmental parameters include temperature, humidity, air pressure, and wind field; after collecting environmental parameters, the system further includes: A moving average filtering algorithm is used to remove high-frequency noise from environmental parameters, resulting in preprocessed environmental parameters that serve as input to the least mean square algorithm. The weight coefficients of the adaptive filter are then dynamically adjusted using a fuzzy logic control algorithm.

[0007] In one implementation, correcting the raw radar precipitation estimate using a correction factor includes: The ridge regression algorithm is used, and a regularization term is added to the objective function of the least squares method. The correction coefficient is obtained by adjusting the regularization parameter. The K-fold cross-validation method is used to calculate the prediction error under different regularization parameter values. The regularization parameter value that minimizes the prediction error of the validation set and its corresponding correction coefficient are selected. The selected correction coefficient is then linearly calculated with the original radar precipitation estimate to obtain the corrected precipitation estimate, which is used as the final precipitation estimate.

[0008] Secondly, embodiments of the present invention provide a precipitation estimation system applied in low-altitude economic services, which executes the precipitation estimation method applied in low-altitude economic services as described above, including: The data acquisition module is used to acquire dual-polarization radar echo data and environmental parameters within the lower atmosphere. The adaptive filtering module is used to take environmental parameters and signal characteristics of dual-polarization radar echo data as input to the least mean square algorithm, and dynamically adjust the weight coefficients of the adaptive filter in combination with fuzzy logic control algorithm. The adaptive filter with adjusted weights is used to suppress noise in the dual-polarization radar echo data to obtain filtered radar echo data. The precipitation estimation module is used to extract key precipitation characteristic parameters from the filtered radar echo data and calculate the original radar precipitation estimation value based on the key precipitation characteristic parameters. The ground data fusion module acquires observation data from multiple ground meteorological stations within the radar's coverage area, fuses this data, and generates regional precipitation baseline data corresponding to the radar's coverage area. The fusion process includes: establishing a state-space model of the precipitation data, comprising a state equation and an observation equation; the state equation describes the temporal variation of the regional precipitation baseline data, while the observation equation describes the linear relationship between the observed values ​​of each ground meteorological station and the regional precipitation baseline data; based on the regional precipitation baseline data from the previous moment, predicting the prior estimate of the regional precipitation state at the current moment using the state equation, and calculating the covariance matrix of the prediction error; acquiring observation data from all ground meteorological stations within the radar's coverage area at the current moment, calculating the Kalman gain matrix, using the Kalman gain matrix to weight and correct the prior estimate of the regional precipitation state, fusing the observation information from multiple ground meteorological stations to obtain the posterior estimate of the regional precipitation state at the current moment, updating the error covariance matrix, and finally obtaining the regional precipitation baseline data for the current moment. The correction output module is used to determine the correction coefficients using regional precipitation baseline data and a regularized linear regression method, and then uses the correction coefficients to correct the original radar precipitation estimates, outputting the final precipitation estimate.

[0009] Thirdly, embodiments of the present invention provide an electronic device comprising a memory and a processor. The memory and the processor communicate with each other via an internal connection path. The memory stores instructions, and the processor executes the instructions stored in the memory. When the processor executes the instructions stored in the memory, it causes the processor to perform the method described in any of the above embodiments.

[0010] Fourthly, embodiments of the present invention provide a computer-readable storage medium that stores a computer program, wherein when the computer program is run on a computer, the methods in any of the embodiments described above are executed.

[0011] The advantages or beneficial effects of the above technical solutions include at least the following: This invention inputs both low-altitude atmospheric environmental parameters and radar echo signal characteristics into an adaptive filter. By combining the least mean square algorithm with fuzzy logic control, it achieves adaptive suppression of complex noise in low-altitude radar echo data, significantly improving the signal-to-noise ratio of low-altitude precipitation signals. Based on this, it extracts key dual-polarization feature parameters and inverts the original precipitation estimates. Simultaneously, it fuses data from multiple ground meteorological stations within the radar coverage area to generate a regional precipitation benchmark. Then, it corrects the original estimates using a regularized linear regression method, ultimately outputting high-precision and high-stability precipitation estimation results. This invention forms a closed loop from data acquisition, noise suppression, feature inversion to multi-source fusion correction, giving full play to the dynamic guiding role of environmental parameters on filtering, the physical sensitivity of dual polarization parameters to liquid water, and the systematic calibration capability of ground station data for radar inversion. It collaboratively solves the technical problem that precipitation estimation in the low-altitude atmosphere is susceptible to interference from ground clutter, non-meteorological echoes, and atmospheric attenuation. It effectively improves the estimation accuracy and generalization ability under conditions of heavy precipitation, light precipitation, and complex wind fields, providing real-time and reliable meteorological data support for low-altitude aircraft route planning, hazardous weather avoidance, and take-off and landing decisions.

[0012] The above overview is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments, and features described above, further aspects, embodiments, and features of the invention will become readily apparent from the accompanying drawings and the following detailed description. Attached Figure Description

[0013] In the accompanying drawings, unless otherwise specified, the same reference numerals throughout the various drawings denote the same or similar parts or elements. These drawings are not necessarily drawn to scale. It should be understood that these drawings depict only some embodiments disclosed in the invention and should not be construed as limiting the scope of the invention.

[0014] Figure 1 This is a flowchart illustrating the precipitation estimation method of the present invention applied to low-altitude economic services. Figure 2 This is a schematic diagram of the data correction workflow of the present invention; Figure 3 This is a schematic diagram of the precipitation estimation system of the present invention applied to low-altitude economic services. Figure 4 This is a structural block diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0015] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of the invention. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0016] Example 1 This embodiment provides a precipitation estimation method applied in low-altitude economic services, with reference to... Figure 1 As shown, Figure 1 As shown, the method specifically includes the following steps: Step S1: Acquire dual-polarization radar echo data and environmental parameters within the lower atmosphere; wherein, the lower atmosphere refers to the boundary layer from the Earth's surface to a height of 3 kilometers.

[0017] Radar stations are strategically deployed within the low-altitude economic service area in advance, and raw radar echo data is acquired using traditional radar systems. This embodiment utilizes dual-polarization radar data, which boasts a spatial resolution down to the hundred-meter level. This allows for the differentiation of precipitation from non-precipitation at a smaller spatial scale, enabling more precise, stable, and timely high-resolution precipitation monitoring. It is particularly suitable for localized severe convection, complex terrain, and low-altitude flight support in low-altitude economic scenarios. Its working principle is based on the transmission and reception of electromagnetic waves. By emitting electromagnetic pulses of specific frequency and power into the low-altitude atmosphere, some energy is scattered back when these electromagnetic waves encounter precipitation particles. The radar receives the scattered echo signals. The size, shape, phase, and spatial distribution of different precipitation particles cause variations in the intensity, frequency, and phase of the echo signals. These echo signals contain rich information related to low-altitude atmospheric precipitation, providing a raw data source for subsequent analysis and processing.

[0018] During the collection of dual-polarization radar echo data, it is necessary to record the data acquisition time and radar operating parameters (such as transmission frequency and pulse width) for subsequent analysis and processing. Simultaneously, preliminary quality control should be performed on the collected radar echo data to check for missing data, outliers, etc. If problems are found, data repair or re-acquisition should be carried out promptly.

[0019] Meanwhile, various sensors are used to collect environmental parameters within the lower atmosphere, including temperature, humidity, air pressure, and wind field. These sensors must possess high accuracy and high reliability. Specifically, platinum resistance temperature sensors can be used for atmospheric temperature, with a measurement accuracy of ±0.1℃; capacitive humidity sensors are used, with an accuracy of ±2%RH; and silicon piezoresistive pressure sensors are used, with a measurement accuracy of ±0.1hPa. Atmospheric wind field information can be collected using ultrasonic anemometers, which offer high measurement accuracy and fast response.

[0020] After collecting environmental parameters in the lower atmosphere and signal characteristics (including echo intensity, signal-to-noise ratio, and spectral width) of dual-polarization radar echo data, both are preprocessed to improve data quality and provide stable and standardized input for adjusting the parameters of the adaptive filter.

[0021] For temperature, humidity, and air pressure parameters, a moving average filtering algorithm is used to remove high-frequency noise. Specifically, let the time series of a certain environmental parameter be x(n), where n is the sampling time number. Selecting a sliding window size M, for each current time n, the arithmetic mean of the M most recent sampling points is taken as the filtered output y(n), that is: ; The window size M can be set according to the sensor's sampling frequency and noise characteristics, with a typical value range of 3 to 11. By using a moving average filter, random high-frequency fluctuations in temperature, humidity, and air pressure data can be effectively suppressed, preserving the true trend of environmental parameters.

[0022] For dual-polarization radar echo data, this embodiment preprocesses it to remove various noise and interference signals, improve data quality, and provide reliable input for subsequent adaptive filtering and precipitation feature extraction. Specifically: Median filtering is used to remove isolated impulse noise and point clutter from radar echo data. Specifically, for a one-dimensional radar echo data sequence s(i), where i = 1, 2, ..., L; i is the range library index, and L is the number of data points on a single radial line, a window size M is selected. The window size is adjusted based on the noise characteristics and signal detail requirements of the data: a larger window is selected if there is a lot of noise and the signal detail requirement is low; a smaller window is selected if more signal detail needs to be retained. Taking a window size of M=5 as an example, for the i-th range library, five data points s[i] are taken, including the two points before and after the i-th range library and the current point. 2], s[i Given s[i], s[i+1], and s[i+2], sort these five data points by their numerical values ​​and take the median value as the filtered output y. med [i]. The median filtering process is completed by sliding the window point by point along the radial direction of the radar. This step effectively removes isolated noise points while preserving the main characteristics of the echo signal.

[0023] Based on median filtering, spectral analysis and band-stop filtering techniques are used to suppress interference signals at specific frequencies. First, the median-filtered data sequence y... med [i] Perform a Fast Fourier Transform (FFT) to convert the time-domain signal into a frequency-domain signal. The formula for its Discrete Fourier Transform is: ; Where k = 0, 1, ..., L 1; L represents the number of data points. Through spectrum analysis, the energy distribution of the signal at different frequencies is obtained, and the frequency range of the interfering signal is identified.

[0024] Design a band-stop filter based on the known noise frequency range. For example, using an elliptic band-stop filter, determine the filter's passband boundary frequency f. p1 f p2 Stopband boundary frequency f s1 f s2 Maximum passband attenuation A P and stopband minimum attenuation A S Parameters such as these are used to design the filter's transfer function H. bs (z). Using the designed band-stop filter for y med [i] Perform filtering to remove irrelevant frequency signals (such as electromagnetic interference of a specific frequency, frequency components corresponding to ground clutter, etc.) to obtain the output y. bs [i].

[0025] Furthermore, to further suppress background noise, a wavelet transform denoising algorithm is employed. For y bs [i] is considered as a sample of a continuous signal f(t), and wavelet transform is performed to obtain the wavelet coefficients W. f (a, b), where a is the scale parameter and b is the translation parameter. In the wavelet domain, the wavelet coefficients of a signal typically have large amplitudes and exhibit regularity at certain scales and locations, while the wavelet coefficients of noise are relatively small and uniformly distributed. A threshold T is set to perform thresholding on the wavelet coefficients: when |W f When (a, b)|>T, the coefficient is considered to correspond to the signal component and is retained; when |W f When (a, b) | ≤ T, the corresponding noise component is considered and set to zero. The threshold function can be either a hard threshold function or a soft threshold function. Hard thresholding function: Retains coefficients whose absolute value is greater than the threshold, and sets the rest to zero; its expression is: ; Soft thresholding function: shrinks coefficients whose absolute value is greater than the threshold to zero (subtracts the threshold), and sets the rest to zero; its expression is: ; It should be noted that the superscript T above refers to the coefficient after thresholding.

[0026] After thresholding, inverse wavelet transform is performed to reconstruct the denoised signal.

[0027] Subsequently, bandpass filtering was used to extract the precipitation signal. Precipitation signals typically have a specific frequency range, while interference signals are distributed in other frequency bands. For example, using a Butterworth bandpass filter, its squared amplitude response is: ; Where f0 is the center frequency, B is the bandwidth, and N is the filter order. By appropriately selecting parameters such as f0, B, and N, the filter achieves high gain within the precipitation signal frequency range and low gain outside this range. This bandpass filter is used to filter the echo data after wavelet transform denoising, allowing the precipitation echo signal to pass through while effectively filtering out residual interference signals outside the passband.

[0028] Finally, the min-max normalization method is used to process the data. After normalization, all data are normalized to the [0, 1] interval, which eliminates the influence of different magnitudes of data on subsequent processing and improves the stability and accuracy of data processing.

[0029] After undergoing the aforementioned median filtering, spectrum analysis and band-stop filtering, wavelet transform denoising, bandpass filtering and normalization, preprocessed dual-polarization radar echo data is obtained. This data has a high signal-to-noise ratio and clear precipitation signal characteristics, providing a high-quality data foundation for subsequent adaptive filtering adjustments and extraction of key precipitation characteristic parameters.

[0030] Step S2: Using environmental parameters and signal characteristics of dual-polarization radar echo data as input to the least mean square algorithm, and combining fuzzy logic control algorithm to dynamically adjust the weight coefficients of the adaptive filter, the adaptive filter with adjusted weights is used to suppress noise in the dual-polarization radar echo data to obtain filtered radar echo data.

[0031] In this embodiment, the core function of the adaptive filter module is to dynamically adjust the filter's weighting coefficients based on real-time collected environmental parameters (temperature, humidity, air pressure, wind field) and signal characteristics (echo intensity, signal-to-noise ratio, spectral width) of the low-altitude atmosphere and dual-polarization radar echo data. This effectively suppresses noise in the radar echo data and outputs filtered data with a high signal-to-noise ratio. Specifically: Let the input signal vector of the adaptive filter be X(t) = [x1(t), x2(t), ..., x...]. Lf (t)] T L fLet t be the filter order, and t be the discrete-time index. Here, the superscript T indicates that the row vector is transposed into a column vector. The elements in the input vector include: normalized environmental parameters such as temperature, humidity, air pressure, and wind field, as well as signal characteristics such as echo intensity, signal-to-noise ratio, and spectral width. The filter weight coefficient vector is W(t) = [w1(t), w2(t), ..., w...]. Lf (t)] T Here, the superscript T indicates the transpose operation. The actual output is y(t) = W. T The expression is defined as y(t)X(t), where the superscript T denotes the transpose operation. The desired output d(t) is an estimate of the clean radar echo signal (obtainable through statistics from periods without precipitation or adjacent beams). The error signal is defined as e(t) = d(t) - y(t).

[0032] The LMS algorithm dynamically updates the weight coefficients based on real-time input environmental parameters and signal characteristics. The update formula is as follows: W(t+1)=W(t)+μ(t)·e(t)·X(t); where μ(t) is a step size parameter that changes with time, and its value is dynamically adjusted by the fuzzy logic controller.

[0033] To achieve optimal convergence performance under different meteorological conditions, this embodiment introduces a fuzzy logic controller to adjust the step size parameter μ(t) in real time. The controller's inputs are the current error e(t) and the error change rate Δe(t) = e(t) - e(t-1). Based on empirical rules, when the error is large or the error change rate is large, the step size is increased to accelerate convergence; when the error is small and tends to stabilize, the step size is decreased to reduce steady-state error. Through fuzzy inference and defuzzification, the current step size adjustment Δμ(t) is obtained, thereby updating the step size parameter. μ(t+1) = μ(t) + Δμ(t); And restrict μ(t+1) to a preset minimum value μ. min With the maximum value μ max Between (e.g., μ) min =0.001, μ max =0.1). In this way, the filter can adapt to changes in meteorological environments such as temperature and humidity, and maintain fast convergence and low steady-state error under different conditions.

[0034] The weighted adaptive filter is applied to suppress noise in dual-polarization radar echo data. For each elevation layer, in each azimuth direction, the process is performed radially from near to far, sliding through range libraries: in the current range library, environmental parameters and signal characteristics at that moment are used to construct the input vector; the filtered output is calculated based on the current weight coefficients; the weight coefficients are updated using the LMS algorithm formula; the process is repeated in the next range library until the entire radial line is completed. By traversing all azimuth and elevation layers, the filtered radar echo data is finally obtained.

[0035] In the filtered data, ground clutter, electromagnetic interference, and random noise are significantly suppressed, while the structural information of precipitation echoes is preserved. This data is then output to the subsequent key precipitation feature parameter extraction module.

[0036] Step S3: Extract key precipitation characteristic parameters from the filtered radar echo data, and calculate the original radar precipitation estimate based on the key precipitation characteristic parameters.

[0037] Key precipitation characteristic parameters in this embodiment include correlation coefficient, differential propagation phase shift, and differential propagation phase shift rate.

[0038] This embodiment extracts correlation coefficients from filtered radar echo data. For horizontally polarized and vertically polarized echo sequences along the same radial line, the zero-hysteresis cross-correlation value of the two sequences is calculated, and divided by the geometric mean of their respective autocorrelation values ​​to obtain the correlation coefficient. A correlation coefficient close to 1 indicates that the echo mainly originates from precipitation particles, while a coefficient significantly less than 1 indicates the presence of ground clutter or a mixed phase. The calculated correlation coefficient is output as one of the key precipitation characteristic parameters.

[0039] The differential propagation phase shift is extracted by acquiring the phase values ​​of the horizontally polarized wave and the vertically polarized wave at various range stations along the radar radial direction. Subtracting the phase of the vertically polarized wave from the phase of the horizontally polarized wave at the same range station yields the differential propagation phase shift at that range station. This parameter reflects the cumulative phase difference between the horizontally and vertically polarized waves as the electromagnetic wave passes through precipitation particles. It is sensitive to liquid water content but less affected by hail, bright bands, and ground clutter. The calculated differential propagation phase shift is output as one of the key precipitation characteristic parameters.

[0040] The differential propagation phase shift rate is defined as the rate of change of the differential propagation phase shift per unit distance along the radar radial direction, and its magnitude directly reflects the precipitation intensity. Since the differential propagation phase shift is discrete range database data and cannot be directly differentiated, a differential approximation method is used for calculation. Specifically, for the current range database, several range databases symmetrically arranged before and after it are selected (usually one or two range databases before and after, corresponding to a sliding window of three to five range databases). The differential propagation phase shift of the forward range database is subtracted from the differential propagation phase shift of the backward range database, and then divided by the total distance of the sliding window (i.e., the product of twice the number of half-points and the range database interval). The result is the differential propagation phase shift rate of the current range database. This calculation is performed sliding along the radar radial direction, one range database at a time, to obtain the differential propagation phase shift rate distribution along the entire radial line. This parameter is sensitive to liquid water content and less affected by hail, bright bands, and ground clutter, making it one of the key parameters for quantitatively estimating heavy precipitation. The calculated differential propagation phase shift rate is output as a key precipitation characteristic parameter.

[0041] In this embodiment, based on key precipitation characteristic parameters (including correlation coefficient, differential propagation phase shift, and differential propagation phase shift rate) extracted from filtered low-altitude radar echo data, a neural network precipitation inversion model trained with historical data is used to calculate the estimated value of the original radar precipitation. Specifically: The neural network model employs a multilayer perceptron structure, consisting of an input layer, multiple hidden layers, and an output layer. The feature vector received by the input layer includes: correlation coefficient ρ, differential propagation phase shift Φ. DP and differential propagation phase shift rate K DP The hidden layer contains several neurons. Each neuron performs a weighted sum of the input information, followed by a non-linear transformation using an activation function. The activation function can be the sigmoid function. σ(x) is the output value of the sigmoid function, which maps any real number input x to the interval (0, 1); Or the ReLU function: f(x) is the output value of the ReLU function, which maps the input x to max(0, x), that is, if x>0, then output x, otherwise output 0.

[0042] The output layer outputs precipitation intensity values ​​(unit: mm / h) and their spatial distribution information.

[0043] The neural network model is trained based on a large amount of historical radar echo data and its corresponding actual surface precipitation data. During training, the network's predicted output is set as follows: The actual precipitation data is y i The loss function uses mean squared error: ; Where 'a' represents the number of training samples. The network parameters (including weights and biases between neurons in each layer) are continuously adjusted using the backpropagation algorithm to minimize the loss function, thereby improving the model's prediction accuracy.

[0044] The correlation coefficient, differential propagation phase shift, and differential propagation phase shift rate extracted in real time are input into the trained neural network model. After forward propagation calculation, the model outputs the precipitation intensity value, which is used as the radar's raw precipitation estimate for subsequent correction steps.

[0045] Step S4: Acquire observation data from multiple ground meteorological stations within the radar detection coverage area, fuse the observation data from multiple ground meteorological stations, and generate regional precipitation benchmark data corresponding to the radar detection coverage area.

[0046] When fusing data from multiple ground meteorological stations to generate regional precipitation baseline data, in addition to using a weighted averaging method, the Kalman filter algorithm can also be introduced. The Kalman filter algorithm can utilize the system's state equation and observation equation to perform optimal estimation of noisy observation data. Specifically: Establish a state-space model of precipitation data, including state equations and observation equations; the state equations are used to describe the variation of regional precipitation baseline data over time, and the observation equations are used to describe the linear relationship between the observed values ​​of each surface meteorological station and the regional precipitation baseline data. Based on the regional precipitation baseline data of the previous moment, the prior estimate of the regional precipitation state at the current moment is predicted by the state equation, and the covariance matrix of the prediction error is calculated. The system acquires observational data from all ground meteorological stations within the radar's coverage area at the current moment, calculates the Kalman gain matrix, uses the Kalman gain matrix to weight and correct the prior estimate of regional precipitation status, fuses observational information from multiple ground meteorological stations to obtain the posterior estimate of regional precipitation status at the current moment, and updates the error covariance matrix simultaneously. Finally, it obtains the regional precipitation baseline data for the current moment. Let... for The actual state of precipitation in the region at any given time. for The state equation for the weather station's observations at a given time can be expressed as: , The observation equation is: , in Here is the state transition matrix. For the observation matrix, and These are process noise and observation noise, respectively, both of which follow a Gaussian distribution.

[0047] Kalman filtering is calculated iteratively through two steps: prediction and update. Specifically, the prediction step uses the state estimate from the previous time step to predict the prior state estimate for the current time step through the state equation and calculates the prediction error covariance matrix.

[0048] The update step involves acquiring observation data from all ground meteorological stations at the current moment, calculating the Kalman gain matrix, using this gain to weight and correct the prior estimate, fusing observation information from multiple stations to obtain the posterior estimate of the state at the current moment, and updating the error covariance matrix at the same time.

[0049] The posterior estimate is output as the regional precipitation baseline data for the current time step and used as the input for the prediction at the next time step. The above steps are performed time-by-time to generate dynamically optimized regional precipitation baseline data.

[0050] Furthermore, spatial interpolation algorithms can be used to fuse data from ground weather stations. For example, Kriging interpolation can be employed, which considers the spatial autocorrelation of the data. First, a spatial coordinate matrix is ​​constructed based on the geographic coordinates of each ground weather station, and the distances between the stations are calculated. Then, an experimental variogram model (e.g., a spherical model or an exponential model) is selected to fit the spatial variability characteristics of the precipitation data, obtaining the variogram parameters. Based on the fitted variogram, a system of Kriging equations is established, and the optimal weighting coefficients for the observations at each weather station are solved. Finally, for any spatial location to be estimated, the observations from each weather station and their optimal weighting coefficients are weighted and summed to obtain the estimated precipitation value for that location. The above process is repeated for all points to be estimated within the entire radar detection coverage area to generate continuous regional precipitation baseline data.

[0051] In this embodiment, regional precipitation baseline data obtained through Kalman filtering or Kriging interpolation is used to correct the original radar precipitation estimates, thereby improving the accuracy and reliability of precipitation estimates.

[0052] Step S5: Using regional precipitation baseline data, a regularized linear regression method is used to determine the correction coefficients, and the original radar precipitation estimates are corrected using the correction coefficients to output the final precipitation estimates.

[0053] Combination Figure 2 As shown in the figure, in this embodiment, the regional precipitation baseline data obtained by fusing ground meteorological station data is y. i (The actual precipitation value of the i-th sample), corresponding to the radar's raw precipitation estimate, is x. i The relationship between the two is described using a linear regression model: y i =β0+β1x i +ε i ; Where β0 and β1 are the correction coefficients to be determined, ε i This is the error term.

[0054] To avoid overfitting, ridge regression is used, incorporating a regularization term into the objective function of least squares. For the multivariate case (considering multiple features), the objective function is: ; Where λ is the regularization parameter (λ≥0), and p is the number of independent variables (in this embodiment, p=1 can be used, i.e., univariate correction). By adjusting λ, a balance is achieved between fitting accuracy and model complexity, thereby improving the generalization ability of the correction coefficients.

[0055] The optimal λ value is selected using a K-fold cross-validation method. The historical dataset (containing paired regional precipitation baseline data y and radar raw precipitation estimates x) is randomly divided into K parts (e.g., K=5 or K=10). K-1 parts are used alternately as the training set, and the remaining part as the validation set. For each candidate regularization parameter λ, the correction coefficients β0 and β1 are solved using ridge regression on the training set. Then, the error (e.g., mean squared error) between the corrected predicted values ​​and the true values ​​is calculated on the validation set. This process is repeated for all K partitions, calculating the average validation error for each λ. The λ value that minimizes the average validation error is selected, and the dataset is retrained using this λ value to obtain the final correction coefficients β0 and β1.

[0056] The determined correction coefficients β0 and β1 are applied to the raw precipitation estimates from the real-time radar, and linear calculations are performed: y corrected =β0+β1·x raw ;x raw These are the raw precipitation estimates from the radar. The final corrected precipitation estimate y is obtained. corrected This will be output as the final precipitation estimate.

[0057] Meanwhile, an update cycle (e.g., weekly or monthly) is set to periodically re-collect the latest regional precipitation baseline data and corresponding radar raw precipitation estimates, repeat the above cross-validation and ridge regression solution process, and update the correction coefficients to adapt to changes in the meteorological environment and radar system characteristics.

[0058] Example 2 This embodiment provides a precipitation estimation system applied in low-altitude economic services, executing the precipitation estimation method for low-altitude economic services as described above. Figure 3 As shown, the system includes at least a data acquisition module, an adaptive filtering module, a precipitation estimation module, a ground data fusion module, and a correction output module.

[0059] The data acquisition module is composed of a traditional radar system and various sensors. It acquires dual-polarization radar echo data in the lower atmosphere through traditional radar sensors and collects environmental parameters in the lower atmosphere through various sensors.

[0060] The adaptive filtering module, precipitation estimation module, ground data fusion module, and correction output module in the precipitation estimation system can be embedded in the traditional radar system, or they can be processed by an additional processor connected to the traditional radar system to achieve the final output.

[0061] The adaptive filtering module takes environmental parameters and signal characteristics of dual-polarization radar echo data as input to the least mean square algorithm, and dynamically adjusts the weight coefficients of the adaptive filter using a fuzzy logic control algorithm. The adjusted adaptive filter is then used to suppress noise in the dual-polarization radar echo data to obtain filtered radar echo data.

[0062] The precipitation estimation module is used to extract key precipitation characteristic parameters from the filtered radar echo data and calculate the original radar precipitation estimation value based on the key precipitation characteristic parameters.

[0063] The ground data fusion module is used to acquire observation data from multiple ground meteorological stations within the radar detection coverage area, fuse the observation data from multiple ground meteorological stations, and generate regional precipitation benchmark data corresponding to the radar detection coverage area.

[0064] The correction output module uses regional precipitation baseline data, employs a regularized linear regression method to determine correction coefficients, and uses these correction coefficients to correct the original radar precipitation estimates, outputting the final precipitation estimate.

[0065] The final precipitation estimate can be output to users of low-altitude economic meteorological services through a visual interface or in a standard data format, to guide low-altitude aircraft to avoid areas of heavy precipitation, optimize flight routes, or assist in takeoff and landing decisions.

[0066] In addition, the system in this embodiment also includes a data preprocessing module for preprocessing the acquired dual-polarization radar echo data and environmental parameters.

[0067] It should be noted that the functions of each module in this embodiment have been disclosed in the method of Embodiment 1, and will not be described again here.

[0068] Example 3 This embodiment provides an electronic device. Figure 4 A structural block diagram of an electronic device according to an embodiment of the present invention is shown. Figure 4As shown, the electronic device includes a memory 100 and a processor 200. The memory 100 stores a computer program that can run on the processor 200. When the processor 200 executes the computer program, it implements the precipitation estimation method applied in the low-altitude economic service described in the above embodiments. The number of memories 100 and processors 200 can be one or more.

[0069] The electronic device also includes: The communication interface 300 is used to communicate with external devices and perform data exchange and transmission.

[0070] If the memory 100, processor 200, and communication interface 300 are implemented independently, they can be interconnected via a bus to communicate with each other. This bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. This bus can be divided into address bus, data bus, control bus, etc.

[0071] Optionally, in a specific implementation, if the memory 100, processor 200, and communication interface 300 are integrated on a single chip, then the memory 100, processor 200, and communication interface 300 can communicate with each other through an internal interface.

[0072] This invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method provided in this invention.

[0073] This invention also provides a chip, which includes a processor for calling and executing instructions stored in a memory, causing a communication device on which the chip is installed to perform the method provided in this invention.

[0074] This invention also provides a chip, including: an input interface, an output interface, a processor, and a memory. The input interface, output interface, processor, and memory are connected through an internal connection path. The processor is used to execute code in the memory. When the code is executed, the processor is used to execute the method provided in this invention.

[0075] It should be understood that the aforementioned processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. General-purpose processors can be microprocessors or any conventional processor. It is worth noting that the processor can be a processor supporting the Advanced Reduced Instruction Set Computing (RISC) machine (ARM) architecture.

[0076] Further, optionally, the aforementioned memory may include read-only memory and random access memory, and may also include non-volatile random access memory. The memory may be volatile or non-volatile, or may include both. Non-volatile memory may include read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory may include random access memory (RAM), which serves as an external cache. Many forms of RAM are available by way of example, but not limitation. Examples include static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (SLDRAM), and direct rambus RAM (DR RAM).

[0077] In the above embodiments, implementation can be achieved, in whole or in part, by software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the present invention is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another.

[0078] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0079] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0080] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in the present invention, and these should all be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A precipitation estimation method applied in low-altitude economic services, characterized in that, include: Acquire dual-polarization radar echo data and environmental parameters within the lower atmosphere; The environmental parameters and the signal characteristics of the dual-polarization radar echo data are used as inputs to the least mean square algorithm. The weight coefficients of the adaptive filter are dynamically adjusted by combining the fuzzy logic control algorithm. The adaptive filter with adjusted weights is used to suppress noise in the dual-polarization radar echo data to obtain filtered radar echo data. Key precipitation characteristic parameters are extracted from the filtered radar echo data, and the original radar precipitation estimate is calculated based on the key precipitation characteristic parameters. The method involves acquiring observation data from multiple ground meteorological stations within the radar detection coverage area, fusing the observation data from these stations to generate regional precipitation baseline data corresponding to the radar detection coverage area. The fusing process includes: establishing a state-space model of the precipitation data, comprising a state equation and an observation equation; the state equation describing the temporal variation of the regional precipitation baseline data, and the observation equation describing the linear relationship between the observed values ​​of each ground meteorological station and the regional precipitation baseline data; based on the regional precipitation baseline data from the previous moment, predicting the prior estimate of the regional precipitation state at the current moment using the state equation, and calculating the covariance matrix of the prediction error; acquiring observation data from all ground meteorological stations within the radar detection coverage area at the current moment, calculating the Kalman gain matrix, using the Kalman gain matrix to weight and correct the prior estimate of the regional precipitation state, fusing the observation information from multiple ground meteorological stations to obtain the posterior estimate of the regional precipitation state at the current moment, updating the error covariance matrix, and finally obtaining the regional precipitation baseline data for the current moment. Using the regional precipitation baseline data, a regularized linear regression method is used to determine the correction coefficients, and the correction coefficients are used to correct the original radar precipitation estimates, outputting the final precipitation estimates.

2. The precipitation estimation method applied in low-altitude economic services according to claim 1, characterized in that, It also includes preprocessing the dual-polarization radar echo data; the preprocessing includes: Wavelet transform is performed on the dual-polarization radar echo data to obtain wavelet coefficients; Wavelet coefficients with absolute values ​​greater than or equal to the threshold are retained, while wavelet coefficients with absolute values ​​less than the threshold are set to zero. Then, the denoised echo data is reconstructed through inverse wavelet transform. A bandpass filter is used to filter the denoised echo data. The passband frequency range of the bandpass filter matches the frequency range of the precipitation echo signal in the dual-polarization radar echo data. This allows the precipitation echo signal to pass through while suppressing interference signals outside the passband range, resulting in preprocessed dual-polarization radar echo data, which is then used for subsequent noise suppression and extraction of key precipitation characteristic parameters.

3. The precipitation estimation method applied in low-altitude economic services according to claim 1, characterized in that, The key precipitation characteristic parameters include correlation coefficient, differential propagation phase shift, and differential propagation phase shift rate; the radar raw precipitation estimate calculated based on the key precipitation characteristic parameters includes: A precipitation inversion model based on neural network training is adopted, and the correlation coefficient, differential propagation phase shift, and differential propagation phase shift rate among the key precipitation characteristic parameters are used as input features and input into the precipitation inversion model; wherein, the precipitation inversion model adopts a multilayer perceptron structure, which consists of an input layer, multiple hidden layers and an output layer, wherein the neurons in the hidden layers perform nonlinear transformation and feature extraction on the input information through activation functions; The precipitation inversion model outputs a precipitation intensity value, which is then used as the radar's original precipitation estimate.

4. The precipitation estimation method applied in low-altitude economic services according to claim 1, characterized in that, The environmental parameters include temperature, humidity, air pressure, and wind field; after collecting the environmental parameters, the following are also included: The environmental parameters are processed by a moving average filtering algorithm to remove high-frequency noise, resulting in preprocessed environmental parameters that serve as input to the least mean square algorithm. The weight coefficients of the adaptive filter are then dynamically adjusted using a fuzzy logic control algorithm.

5. The precipitation estimation method applied in low-altitude economic services according to claim 1, characterized in that, The step of correcting the original radar precipitation estimate using the correction coefficient includes: The ridge regression algorithm is used, and a regularization term is added to the objective function of the least squares method. The correction coefficient is obtained by adjusting the regularization parameter. The prediction error under different regularization parameter values ​​is calculated using the K-fold cross-validation method. The regularization parameter value that minimizes the prediction error of the validation set and its corresponding correction coefficient are selected. The selected correction coefficient is then linearly calculated with the original radar precipitation estimate to obtain the corrected precipitation estimate, which is used as the final precipitation estimate.

6. A precipitation estimation system applied in low-altitude economic services, characterized in that, Performing the precipitation estimation method for low-altitude economic services as described in any one of claims 1 to 5 includes: The data acquisition module is used to acquire dual-polarization radar echo data and environmental parameters within the lower atmosphere. An adaptive filtering module is used to take the environmental parameters and the signal characteristics of the dual-polarization radar echo data as inputs to the least mean square algorithm, and dynamically adjust the weight coefficients of the adaptive filter in combination with a fuzzy logic control algorithm. The adaptive filter with adjusted weights is used to suppress noise in the dual-polarization radar echo data to obtain filtered radar echo data. The precipitation estimation module is used to extract key precipitation feature parameters from the filtered radar echo data and calculate the original radar precipitation estimation value based on the key precipitation feature parameters. A ground data fusion module is used to acquire observation data from multiple ground meteorological stations within the radar detection coverage area, and to fuse the observation data from these multiple ground meteorological stations to generate regional precipitation benchmark data corresponding to the radar detection coverage area. The fusion processing of the observation data from multiple ground meteorological stations includes: establishing a state-space model of the precipitation data, comprising a state equation and an observation equation; the state equation describes the change law of the regional precipitation benchmark data over time, and the observation equation describes the linear relationship between the observed values ​​of each ground meteorological station and the regional precipitation benchmark data; based on the regional precipitation benchmark data of the previous moment, predicting the prior estimate of the regional precipitation state at the current moment through the state equation, and calculating the covariance matrix of the prediction error; acquiring the observation data of all ground meteorological stations within the radar detection coverage area at the current moment, calculating the Kalman gain matrix, using the Kalman gain matrix to weight and correct the prior estimate of the regional precipitation state, fusing the observation information from multiple ground meteorological stations to obtain the posterior estimate of the regional precipitation state at the current moment, updating the error covariance matrix, and finally obtaining the regional precipitation benchmark data at the current moment. The correction output module is used to determine the correction coefficients using the regional precipitation baseline data and a regularized linear regression method, and then uses the correction coefficients to correct the original radar precipitation estimate and output the final precipitation estimate.

7. An electronic device, characterized in that, include: A processor and a memory, wherein the memory stores instructions that are loaded and executed by the processor to implement the precipitation estimation method for use in low-altitude economic services as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the precipitation estimation method for low-altitude economic services as described in any one of claims 1 to 5.

Images