Atmospheric boundary layer height inversion method and system based on laser radar

Abstract

The invention relates to an atmospheric boundary layer height inversion method and system based on a laser radar, and belongs to the field of atmospheric detection, and the method comprises the steps: employing the laser radar to carry out the vertical detection of a to-be-detected region, obtaining an echo signal, and carrying out the preprocessing of the echo signal, and obtaining a preprocessed echo signal; obtaining an optimal scale parameter of wavelet covariance transformation by using a particle swarm optimization algorithm, and carrying out wavelet covariance transformation on the preprocessed echo signal based on the optimal scale parameter to obtain a wavelet covariance coefficient curve; the local maximum value position of the wavelet covariance coefficient curve is used as an ABLH candidate value; calculating feature data of the preprocessed echo signals, and clustering the preprocessed echo signals into an aerosol layer, a cloud layer and a common atmosphere based on the feature data and an optimal covariance coefficient corresponding to the optimal scale parameter; setting the maximum aerosol concentration change position in the aerosol layer as an ABLH threshold value; the ABLH candidate value closest to the ABLH threshold value is determined to be the atmospheric boundary layer height.

Description

Technical Field

[0001] The present invention relates to the field of atmospheric detection technology, and in particular to an atmospheric boundary layer height inversion method system based on laser radar. Background Art

[0002] The atmospheric boundary layer (ABL) is the transition layer between the Earth's surface and the free atmosphere of the troposphere. Its thickness typically ranges from several hundred to several thousand meters. As a key parameter describing the structure and evolution of the ABL, the ABL height is of great significance in fields such as weather forecasting, atmospheric pollution research, climate change monitoring, and aerospace safety.

[0003] LiDAR is a highly precise and sensitive atmospheric detection instrument used to measure parameters such as the concentration, size, and distribution of atmospheric aerosols. By emitting pulsed laser light into the atmosphere and receiving backscattered signals, LiDAR can generate vertical profiles of aerosol concentration. These echoes reveal the vertical variations in aerosol concentration in the atmosphere, providing important information for identifying the boundary layer height.

[0004] Currently, a variety of methods have been developed to invert the atmospheric boundary layer height from lidar echo signals. These include the gradient method, the variance method, and the wavelet covariance transform method. The gradient method finds the ABLH by locating the location where the lidar echo signal attenuates most rapidly, i.e., the location where the gradient is minimum. The variance method determines the ABLH (Atmospheric Boundary Layer Height) by locating the entrainment layer between the boundary layer and the troposphere. The wavelet covariance transform method uses the Haar step function to detect step changes in the lidar signal to determine the boundary layer location.

[0005] Traditional methods, such as gradient and variance methods, are computationally simple but susceptible to noise and cloud interference. The performance of the wavelet covariance transform (WCT) depends heavily on the choice of scale parameter, which can lead to significant deviations in the location of boundary layer heights. Furthermore, the selection of scale parameters in the WCT often relies on empirical formulas or human judgment, lacking adaptability.

[0006] Therefore, traditional inversion methods are difficult to guarantee their robustness and accuracy, especially when atmospheric conditions are complex and changeable. LiDAR echo signals are often affected by clouds and noise background, making it difficult for traditional methods to accurately identify the boundary layer height. Summary of the Invention

[0007] In view of the above analysis, the embodiments of the present invention aim to provide a lidar-based atmospheric boundary layer height inversion method and system to solve the technical problems that traditional inversion methods are difficult to ensure robustness and accuracy under complex atmospheric conditions, and the selection of scale parameters relies on empirical formulas or human judgment and lacks adaptability.

[0008] The purpose of the present invention is mainly achieved through the following technical solutions:

[0009] The present invention provides a method for inverting the atmospheric boundary layer height based on laser radar, comprising the following steps:

[0010] Step S1: Use a laser radar to perform vertical detection on the area to be measured, obtain an echo signal, and pre-process it to obtain a pre-processed echo signal;

[0011] Step S2: using a particle swarm optimization algorithm to obtain an optimal scale parameter for a wavelet covariance transform, applying a wavelet covariance transform based on the optimal scale parameter to the preprocessed echo signal to obtain a wavelet covariance coefficient curve; and using the local maximum position of the wavelet covariance coefficient curve as an ABLH candidate value;

[0012] Step S3: Calculate the characteristic data of the preprocessed echo signal, and cluster the preprocessed echo signal into three categories: aerosol layer, cloud layer, and normal atmosphere layer based on the characteristic data and the optimal covariance coefficient corresponding to the optimal scale parameter; set the position with the largest aerosol concentration change in the aerosol layer as the ABLH threshold; and determine the ABLH candidate value closest to the ABLH threshold as the atmospheric boundary layer height.

[0013] Furthermore, the characteristic data includes relative rate of change, gradient and variance; and step S3 includes:

[0014] K-means clustering was performed based on the normalized relative rate of change, gradient, variance, and optimal covariance coefficient, and three clusters were obtained;

[0015] Calculate the weighted sum of the average relative change rate, average gradient and average variance corresponding to the p echo signal sampling data points closest to each cluster center

[0016] based on Determining categories of the three clusters, the categories including an aerosol layer, a cloud layer, and a general atmosphere layer;

[0017] Performing a wavelet covariance transform on the preprocessed echo signal corresponding to the aerosol layer to obtain an optimal scale parameter of the corresponding aerosol layer, and obtaining a wavelet covariance coefficient curve of the corresponding aerosol layer based on the optimal scale parameter;

[0018] Finding the local maximum position of the wavelet covariance coefficient curve of the aerosol layer, where the local maximum position is the position where the aerosol concentration changes the most, and serving as the ABLH threshold;

[0019] The ABLH candidate value closest to the ABLH threshold among the ABLH candidate values ​​is used as the atmospheric boundary layer height value.

[0020] Furthermore, the weighted sum of the average relative change rate, average gradient and average variance corresponding to the p echo signal sampling data points closest to each cluster center is calculated. as follows:

[0021]

[0022] Among them, n i is the cluster center c i The location of i is the cluster center c i The index of i=1, 2, 3 represents the aerosol layer, cloud layer and normal atmosphere layer respectively, k is the index of the cluster, α i (k),grd i (k) and var i (k) are the normalized relative rate of change, gradient and variance respectively; λ1, λ2, λ3 are the weight coefficients of the average relative rate of change, average gradient and average variance respectively.

[0023] Furthermore, The largest cluster is the cloud layer, The smallest cluster is the general atmosphere layer, and the remaining clusters are the aerosol layer.

[0024] Furthermore, the wavelet covariance transform is performed using the particle swarm optimization algorithm to optimize the wavelet scale parameters, and the optimal scale parameters are obtained, including:

[0025] Initialize the particle swarm algorithm parameters;

[0026] Set the particle swarm algorithm input to the wavelet scale parameter a;

[0027] Based on the current wavelet scale parameter, the preprocessed echo signal is transformed based on the Haar wavelet function to obtain the wavelet covariance coefficient, the fitness value of each particle is calculated based on the corresponding wavelet covariance coefficient, and the optimal position and optimal fitness value of the individual and the group are updated until the maximum number of iterations is reached, and the optimal position and optimal fitness value of the group are obtained;

[0028] The current wavelet scale parameter corresponding to the optimal position and optimal fitness value of the group is obtained as the optimal scale parameter a best , and a best The corresponding wavelet covariance coefficient curve W best .

[0029] Furthermore, the particle fitness value is calculated using the following fitness function:

[0030] f(a)=Max(λ(Max(W(a,b))-Min(W(a,b)))-(z1-z2))

[0031] Among them, λ is the weight factor, b is the position of the center of the Haar wavelet function, W(a,b) is the wavelet covariance coefficient, z1 and z2 are the heights of the upper and lower half-peak positions of the peak position of the wavelet covariance curve, respectively.

[0032] Furthermore, we use the built-in findperks function of Matlab to find the wavelet covariance coefficient curve W best The local maximum position of the ABLH is obtained as a plurality of local maximum positions as the candidate values ​​of the ABLH;

[0033] The multiple local maximum positions include atmospheric boundary layer height and cloud layer height positions.

[0034] Furthermore, the wavelet covariance coefficient W(a,b) is as follows:

[0035]

[0036] Among them, a is the wavelet scale parameter, and a = k × Δz, Δz is the vertical resolution of the lidar, b is the center position of the Haar wavelet function, k is the coefficient used to adjust the scale of the wavelet to adapt to different signal characteristics, r is the detection distance of the lidar, r b and r t are the lower and upper limits of the echo signal range, respectively, and N(r) is the echo signal after preprocessing; is the Haar wavelet function;

[0037] The Haar wavelet function is as follows:

[0038]

[0039] Furthermore, the echo signal is an optical signal; and the preprocessing of the echo signal includes:

[0040] Convert the received optical signal into an analog electrical signal;

[0041] amplifying the analog electrical signal to obtain an amplified electrical signal;

[0042] The amplified electrical signal is subjected to background denoising, smoothing filtering, geometric overlap factor correction processing and laser radar detection distance square correction in sequence to obtain a pre-processed echo signal.

[0043] The present invention also provides an atmospheric boundary layer height inversion system based on laser radar, which is characterized by comprising:

[0044] An echo signal acquisition module is used to use a laser radar to perform vertical detection on the area to be measured, obtain an echo signal, and perform preprocessing to obtain a preprocessed echo signal;

[0045] An ABLH candidate value generation module is configured to obtain an optimal scale parameter of a wavelet covariance transform using a particle swarm optimization algorithm, and to obtain a wavelet covariance coefficient curve by performing a wavelet covariance transform based on the optimal scale parameter on the preprocessed echo signal; and to use the local maximum position of the wavelet covariance coefficient curve as an ABLH candidate value;

[0046] An ABLH determination module is configured to calculate characteristic data of the preprocessed echo signal, cluster the preprocessed echo signal into three categories: an aerosol layer, a cloud layer, and a normal atmosphere layer based on the characteristic data and an optimal covariance coefficient corresponding to the optimal scale parameter; set the location in the aerosol layer where the aerosol concentration changes the most as the ABLH threshold; and determine the ABLH candidate value closest to the ABLH threshold as the atmospheric boundary layer height.

[0047] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0048] 1. This invention optimizes the scale parameters of the wavelet covariance transform by introducing a particle swarm optimization algorithm, which can automatically select the optimal scale parameters, thereby improving the accuracy of atmospheric boundary layer height inversion. In traditional methods, the selection of scale parameters for the wavelet covariance transform relies on empirical formulas or manual judgment, which lacks adaptability. However, this technical solution achieves adaptive optimization of the scale parameters through the particle swarm optimization algorithm;

[0049] 2. The present invention can effectively distinguish between aerosol layers, cloud layers, and ordinary atmosphere layers by combining multiple characteristic data (relative rate of change, gradient, variance) of the echo signal and the optimal covariance coefficient for K-means cluster analysis. This method can reduce the interference of clouds and noise on the inversion of atmospheric boundary layer height and improve the robustness of the algorithm under complex atmospheric conditions. Traditional methods such as the gradient method and the variance method are prone to errors when faced with cloud interference and noise, while the present invention significantly enhances the anti-interference ability of the algorithm through cluster analysis and feature extraction;

[0050] 3. This invention realizes the automation and intelligentization of atmospheric boundary layer height inversion. Through particle swarm optimization algorithm and K-means cluster analysis, the algorithm can automatically identify and screen candidate ABLH values ​​without human intervention. This not only improves the inversion efficiency, but also reduces the influence of human factors on the results, making the inversion of atmospheric boundary layer height more scientific and objective.

[0051] 4. By optimizing the scale parameters of the wavelet covariance transform and combining it with K-means cluster analysis, the present invention can quickly and accurately determine the atmospheric boundary layer height. Compared with traditional methods, this method reduces the amount of computation, improves inversion efficiency, and can provide reliable ABLH results in a shorter time, making it suitable for real-time monitoring and rapid response applications.

[0052] 5. By adjusting the wavelet scale parameters and weight coefficients, the present invention can adapt to the vertical resolution and sampling interval of different lidars, showing strong adaptability. The method works effectively with both high-resolution and low-resolution lidar data, providing a universal inversion method for different types of lidars.

[0053] In the present invention, the above-mentioned technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of the present invention will be described in the following description, and some advantages will become apparent from the description or be learned through practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the contents particularly pointed out in the description and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The accompanying drawings are only for the purpose of illustrating particular embodiments and are not to be considered limiting of the present invention. Like reference symbols denote like parts throughout the drawings.

[0055] Figure 1 Flowchart of the atmospheric boundary layer height inversion method based on lidar in an embodiment of the present invention;

[0056] Figure 2 This is a working diagram of a radar system in an embodiment of the present invention;

[0057] Figure 3 Schematic diagram of the laser radar geometric overlap factor in an embodiment of the present invention;

[0058] Figure 4 This is a flow chart of improving wavelet covariance transform using particle swarm optimization in an embodiment of the present invention;

[0059] Figure 5 Schematic diagram of the atmospheric boundary layer height inversion system module based on lidar in an embodiment of the present invention. DETAILED DESCRIPTION

[0060] The preferred embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, and are not used to limit the scope of the present invention.

[0061] In order to further improve the accuracy and reliability of atmospheric boundary layer height inversion, the present invention proposes a new atmospheric boundary layer height inversion method and system based on lidar, which solves the above technical problems and is expected to be widely used in environmental monitoring, atmospheric science and other fields.

[0062] Example 1:

[0063] A specific embodiment of the present invention, as Figure 1 As shown, a method for inverting the atmospheric boundary layer height based on lidar is disclosed, comprising the following steps:

[0064] Step S1: Use a laser radar to perform vertical detection on the area to be measured, obtain an echo signal, and pre-process it to obtain a pre-processed echo signal;

[0065] Step S2: using a particle swarm optimization algorithm to obtain an optimal scale parameter for a wavelet covariance transform, applying a wavelet covariance transform based on the optimal scale parameter to the preprocessed echo signal to obtain a wavelet covariance coefficient curve; and using the local maximum position of the wavelet covariance coefficient curve as an ABLH candidate value;

[0066] Step S3: Calculate the characteristic data of the preprocessed echo signal, and cluster the preprocessed echo signal into three categories: aerosol layer, cloud layer, and normal atmosphere layer based on the characteristic data and the optimal covariance coefficient corresponding to the optimal scale parameter; set the position with the largest aerosol concentration change in the aerosol layer as the ABLH threshold; and determine the ABLH candidate value closest to the ABLH threshold as the atmospheric boundary layer height.

[0067] Step S1 includes steps S11-S12.

[0068] Step S11: Use the laser radar to perform vertical detection to obtain an echo signal.

[0069] The laser radar system used in this embodiment includes: a laser radar, a gimbal (for adjusting the laser emission angle), a tripod (for fixing and stabilizing the laser radar equipment), and a computing and processing device (for data collection and preliminary processing).

[0070] Exemplarily, the laser radar in the present invention is a non-coaxial laser radar.

[0071] The workflow of the LiDAR system is as follows: Figure 2 As shown, the control system controls the pulse laser emitter to generate laser, which is amplified by the amplifier and then emitted into the atmosphere; the pulse laser is emitted in a vertical direction along the area to be measured.

[0072] The gimbal adjusts the laser emission angle and adjusts the pulse laser transmitter to be 90 degrees perpendicular to the ground to emit the laser.

[0073] For example, the areas to be tested can be of the following types:

[0074] (1) Urban areas: used to study urban air pollution, boundary layer structure, and urban heat island effect;

[0075] (2) Rural or natural areas: used to study background atmospheric conditions, distribution and transport of natural aerosols (such as dust and sea salt);

[0076] (3) Industrial areas: used to monitor the impact of industrial emissions on the atmospheric environment;

[0077] (4) Coastal areas: used to study the interaction between the ocean and the atmosphere, and the distribution of sea salt aerosols.

[0078] The laser pulses emitted by the pulsed laser transmitter interact with aerosol particles in the atmosphere to generate echo signals. The echo signals are captured by the lidar receiver, collected by the information acquisition unit, processed by the signal processing unit, and sent to the control system.

[0079] Step S12: pre-process the echo signal to obtain a pre-processed echo signal.

[0080] The echo signal is an optical signal; and the preprocessing of the echo signal includes:

[0081] Convert the received optical signal into an analog electrical signal;

[0082] amplifying the analog electrical signal to obtain an amplified electrical signal;

[0083] The amplified electrical signal is subjected to background denoising, smoothing filtering, geometric overlap factor correction processing and laser radar detection distance square correction in sequence to obtain a pre-processed echo signal.

[0084] Step S12 is divided into steps S121-S122.

[0085] Step S121: convert the echo signal.

[0086] The received echo signal is converted into an analog electrical signal through a photoelectric detector;

[0087] The analog electrical signal is then amplified through a signal amplifier to improve the intensity and quality of the echo signal. Due to the use of a vertical fixed observation method, after processing by computing and processing equipment, data on the change of backscatter signal intensity at different heights over time can be generated.

[0088] The backscattered light is the echo signal.

[0089] Step S122: The amplified electrical signal is subjected to background denoising, smoothing filtering, geometric overlap factor correction, and laser radar color greed distance square correction in sequence.

[0090] (1) Background noise removal: When the laser radar receiving telescope receives backscattered light, it not only receives the laser echo signal, but also receives the part of the light from light sources such as sunlight and lamplight that has the same spectrum as the laser beam wavelength. This is especially serious during the day. This part of the signal is called background noise. Background noise should be removed when processing the echo signal.

[0091] Background denoising selects the relatively flat signal at the end of the detection distance (7-10km) in the echo signal, calculates the average value of the flat signal, and uses it as the background noise. The background noise is subtracted from the echo signal to obtain the echo signal without background noise.

[0092] (2) Smoothing Filtering: In order to remove high-frequency noise, enhance boundary features, improve the robustness of subsequent algorithms and the credibility of results, and improve the intuitiveness and processing efficiency of data, a sliding average filter is used to smooth the echo signal after background denoising. Smoothing filtering is performed on the middle, starting, and ending parts of the echo signal.

[0093] For the middle portion of the echo signal after background denoising, a sliding average is performed using a fixed window length. To prevent phase shift, the window size needs to be an odd number. To avoid loss of valid information due to an overly large window, a window size of 5 is chosen as an example based on comprehensive considerations.

[0094] For each position n in the middle part, the smoothing filter formula is as follows:

[0095]

[0096] Where x(k) is the input echo signal after background denoising, k is a positive integer; y avg (n) is the sliding average result at position n. k is the index in the signal, indicating the specific sampling data point position.

[0097] For the beginning and end parts of the denoised echo signal, a shorter window length is used for sliding average to reduce the influence of the boundary effect, that is:

[0098]

[0099] Where 1 and m are the starting and ending positions of the echo signal. The first three formulas smooth the starting portion of the echo signal, while the last three formulas smooth the ending portion of the echo signal.

[0100] The smoothed and filtered echo signal is obtained, which effectively removes high-frequency noise and enhances boundary features.

[0101] (3) Geometric overlap factor correction processing: Figure 3 As shown in the figure, when a non-coaxial laser radar is measuring, the emission field of view of the pulse laser transmitter and the receiving field of view of the receiving telescope are not on the same axis, and the angles between the two are different. Therefore, the detection distance r from the laser radar satisfies:

[0102] a) When r≤R1, the geometric overlap factor η=0. Within this distance range, the laser radar's transmitting field of view and receiving field of view do not overlap, so the laser radar cannot receive any echo signal. This area is the radar blind spot;

[0103] b) When R1<r<R2, only some photons will be received, 0<η<1; the geometric overlap factor η is between 0-1. Within this range, only some photons will be received, and the lidar can receive some echo signals. This area is the transition zone;

[0104] c) When r ≥ R2, the geometric overlap factor η = 1. Within this range, the laser radar's transmitting field of view and receiving field of view completely overlap, and the laser radar can receive all echo signals. This area is the bright area.

[0105] The geometric overlap factor was corrected experimentally. Experiments were conducted in a clean environment. For example, lidar echo signals were collected on a night after a recent rain. During this time, aerosol concentrations within a certain range near the ground are low, and the atmospheric distribution is relatively uniform, making the environment considered clean.

[0106] In this environment, lidar is used for detection, and signals are continuously collected for linear fitting calculation to obtain the geometric overlap factor within the corresponding distance.

[0107] R1 and R2 are two key distance parameters that describe the change in the geometric overlap factor η. R1 is the end of the radar blind zone; R2 is the end of the radar transition zone.

[0108] The lidar equation including the geometric overlap factor is:

[0109]

[0110] Where P is the laser radar signal intensity power; r is the laser radar detection distance; E is the laser radar fixed parameter, which is a constant; β is the atmospheric backscattering coefficient; σ is the atmospheric extinction coefficient.

[0111] Transform the above formula into:

[0112] ln(P(r)·r 2)=ln(E(β(r)))+ln(η(r))-2σ(r)·r Formula (4)

[0113] Arranging the above formula can obtain the geometric overlap factor calculation formula:

[0114]

[0115] During the experiment, data outside the transition region of the echo signal (r ≥ R2) are intercepted. At this time, η(r) = 1. The measured data are fitted with the function of formula (4) to obtain the constant terms ln(E(β(r))) and 2σ(r) in formula (5). Then, the geometric overlap factor in the transition region is calculated by substituting them into formula (5).

[0116] According to the calculated geometric overlap factor η(r), the geometric overlap factor correction processing is performed on the echo signal after smoothing and filtering, as follows:

[0117]

[0118] Among them, Q(r) is the echo signal after geometric overlap factor correction processing.

[0119] (4) LiDAR detection distance square correction: As the light pulse emitted by the LiDAR propagates, its energy rapidly decreases with the square of the distance. According to the radar equation (Formula (3)), the received echo signal strength (power) is inversely proportional to the square of the distance to the target. Therefore, performing distance square correction on the radar echo signal can improve the significance and effectiveness of the long-range echo signal.

[0120] Among them, the laser radar detection distance square calibration formula is:

[0121] N(r)=Q(r)·r 2 Formula (7)

[0122] Where N(r) is the signal after square correction of the laser radar detection distance.

[0123] The radar echo signal is preprocessed to obtain a preprocessed echo signal N(r).

[0124] The function of step S1 is to use the laser radar to detect the vertical direction of the area to be measured, obtain the echo signal, and preprocess it, including signal conversion, amplification, background denoising, smoothing filtering, geometric overlap factor correction and laser radar detection distance square correction, and finally obtain the preprocessed echo signal suitable for subsequent analysis.

[0125] Step S2 includes steps S21-S22.

[0126] Step S21: Use PSO (Particle Swarm Optimization) to obtain the optimal scale parameter of wavelet covariance transform. Use particle swarm optimization to improve the process of wavelet covariance transform, such as Figure 4 shown.

[0127] The wavelet covariance transform is performed using the particle swarm optimization algorithm to optimize the wavelet scale parameters and obtain the optimal scale parameters, including:

[0128] Initialize some parameters of the particle swarm algorithm, including: the population size of the particle swarm N = 10, the dimension of the particle d = 1, which only requires the optimization of the scale parameter; the minimum position value of the particle swarm X min =1, position maximum value X max =50, that is, the range of the scale parameter is 1Δz~50Δz; the maximum number of iterations T=50, the inertia weight w=0.8, the learning factors c1=1.5, c2=1.5, and the maximum particle speed V max =5, minimum particle speed V min =-5;

[0129] Set the particle swarm algorithm input to the wavelet scale parameter a;

[0130] Based on the current wavelet scale parameter, the preprocessed echo signal is transformed based on the Haar wavelet function to obtain the wavelet covariance coefficient, the fitness value of each particle is calculated based on the corresponding wavelet covariance coefficient, and the optimal position and optimal fitness value of the individual and the group are updated until the maximum number of iterations is reached, and the optimal position and optimal fitness value of the group are obtained;

[0131] The current wavelet scale parameter corresponding to the optimal position and optimal fitness value of the group is obtained as the optimal scale parameter a best , and a best The corresponding wavelet covariance coefficient curve W best .

[0132] The particle swarm optimization algorithm is introduced into the wavelet covariance transform. By maximizing the peak significance and concentration of the wavelet covariance coefficient curve, the optimal scale parameter of the Haar wavelet is obtained. Finally, the wavelet covariance transform of the preprocessed echo signal is performed using the optimal scale parameter to obtain the wavelet covariance coefficient curve.

[0133] The wavelet covariance coefficient W(a,b) is as follows:

[0134]

[0135] Among them, a is the wavelet scale parameter, and a=k×Δz, Δz is the vertical resolution of the lidar, b is the center position of the Haar wavelet function, k is the coefficient used to adjust the scale of the wavelet to adapt to different signal characteristics, r is the detection distance of the lidar, r b and r t are the lower and upper limits of the echo signal range, respectively, and N(r) is the echo signal after preprocessing; is the Haar wavelet function;

[0136] The Haar wavelet function is as follows:

[0137]

[0138] k is a coefficient used to adjust the scale of the wavelet to accommodate different signal characteristics. By adjusting the value of k, the width of the wavelet changes, thereby adapting to signal characteristics of different scales. For example, a larger k value allows the wavelet to cover a wider area, suitable for detecting larger-scale signal changes; a smaller k value is suitable for detecting smaller-scale signal changes. Based on the characteristic scale of the echo signal, the noise level, and the specific vertical resolution requirements of the lidar, selecting an appropriate k value can effectively improve the accuracy and reliability of signal processing. For example, in lidar signal processing, the coefficient k ranges from 1 to 10.

[0139] The particle fitness value is calculated using the following fitness function:

[0140] f(a)=Max(λ(Max(W(a,b))-Min(W(a,b)))-(z1-z2))

[0141] Formula (10)

[0142] Among them, λ is the weight factor, b is the position of the center of the Haar wavelet function, W(a,b) is the wavelet covariance coefficient, z1 and z2 are the heights of the upper and lower half-peak positions of the peak position of the wavelet covariance curve, respectively.

[0143] This fitness function incorporates the criteria of peak significance and concentration. The difference between the maximum and minimum values ​​of the covariance curve helps measure the prominence of the peak in the covariance curve. Increasing the difference improves peak significance, making boundary layer signals easier to identify. Furthermore, a smaller half-peak width and a more concentrated peak indicate more accurate positioning of the boundary layer height signal.

[0144] The initial position and velocity of each particle are randomly initialized according to the set value range.

[0145] According to the current scale parameter, the wavelet covariance coefficient is calculated by formula (8). The fitness value of each particle is calculated and the optimal position and optimal fitness value of the individual and the group are updated.

[0146] Recalculate the fitness value of each particle and update the optimal position and optimal fitness value of the individual and the group until the maximum number of iterations is reached.

[0147] Output the optimal position and optimal fitness value of the group to obtain the optimal scale parameter a best , and get a best Wavelet covariance coefficient curve W corresponding to the optimal scale parameter best .

[0148] Step S22: Determine the candidate value of ABLH.

[0149] Use the findperks function built into Matlab to find the wavelet covariance coefficient curve W best The local maximum position of the ABLH is obtained as a plurality of local maximum positions as the candidate values ​​of the ABLH;

[0150] The multiple local maximum positions include atmospheric boundary layer height and cloud layer height positions.

[0151] In practical applications, the number of ABLH candidate values ​​is usually 2 to 3. The specific number depends on the echo signal characteristics, environmental conditions, and the vertical resolution of the lidar.

[0152] Single boundary layer: Under simple atmospheric conditions, there is usually only one significant maximum, corresponding to the top of the atmospheric boundary layer.

[0153] Cloudy or complex boundary layer: Under complex meteorological conditions, multiple maxima may be detected.

[0154] Among the candidate ABLH values, there are locations with the greatest changes in aerosol concentration (true ABLH) and "false aerosol layers," that is, locations with the greatest changes in cloud concentration (false ABLH). Subsequent steps are required to distinguish true and false values ​​and determine the true ABLH value.

[0155] The function of step S2 is to optimize the scale parameter of the wavelet covariance transform by using the particle swarm optimization algorithm to obtain the optimal scale parameter, and use the optimal scale parameter to perform a wavelet covariance transform on the preprocessed echo signal to obtain a wavelet covariance coefficient curve, and then determine the candidate value of the atmospheric boundary layer height.

[0156] Step S3 is divided into steps S31-S33.

[0157] The characteristic data includes relative rate of change, gradient and variance; the step S3 includes:

[0158] K-means clustering was performed based on the normalized relative rate of change, gradient, variance, and optimal covariance coefficient, and three clusters were obtained;

[0159] Calculate the weighted sum of the average relative change rate, average gradient and average variance corresponding to the p echo signal sampling data points closest to each cluster center

[0160] based on Determining categories of the three clusters, the categories including an aerosol layer, a cloud layer, and a general atmosphere layer;

[0161] Performing a wavelet covariance transform on the preprocessed echo signal corresponding to the aerosol layer to obtain an optimal scale parameter of the corresponding aerosol layer, and obtaining a wavelet covariance coefficient curve of the corresponding aerosol layer based on the optimal scale parameter;

[0162] Finding the local maximum position of the wavelet covariance coefficient curve of the aerosol layer, where the local maximum position is the position where the aerosol concentration changes the most, and serving as the ABLH threshold;

[0163] The ABLH candidate value closest to the ABLH threshold among the ABLH candidate values ​​is used as the atmospheric boundary layer height value.

[0164] Step S2 calculates a globally optimal scaling parameter, which provides the basis for the wavelet covariance transform across the entire signal range. Step S3 then builds on this by making more refined adjustments to the local characteristics of the aerosol layer to improve the accuracy of identifying the location with the maximum aerosol concentration change.

[0165] The optimal scale parameter in step S2 is calculated for the entire preprocessed echo signal, while the optimal scale parameter in step S3 is optimized specifically for the signal characteristics of the aerosol layer. Although both methods optimize the scale parameter through wavelet covariance transform, their scope of application is different.

[0166] Step S31: Calculate characteristic data of the preprocessed echo signal.

[0167] For the preprocessed echo signal, the relative rate of change, gradient, and variance are calculated to provide features for K-means clustering.

[0168] The relative change rate is the relative change rate of the backscatter signal (i.e., echo signal) intensity at two points at adjacent heights;

[0169]

[0170] Wherein, α(r) represents the relative rate of change at height z, and N(r) represents the echo signal at height r after preprocessing.

[0171] The calculation formula of the gradient is as follows:

[0172]

[0173] Where grd(r) represents the gradient at height r.

[0174] The calculation formula for variance is as follows:

[0175]

[0176] Where var(r) represents the variance at height r, and n represents the number of points at which the variance is calculated.

[0177] Step S32: Perform K-means clustering based on the normalized relative change rate, gradient, variance, and optimal covariance coefficient to obtain three clusters.

[0178] Using the optimal scale parameter a best The corresponding optimal covariance coefficient and characteristic data (relative rate of change, gradient, variance parameter) are used to cluster the preprocessed echo signals using the K-means method to obtain three clusters.

[0179] K-means is a classic unsupervised machine learning algorithm that aims to partition a dataset into k clusters, such that samples within the same cluster are as similar as possible, while samples across clusters are as different as possible. The core idea is to iteratively optimize the location of cluster centers to minimize the sum of the squared distances from samples within a cluster to the centers.

[0180] The specific steps of using K-means clustering method to segment the cloud layer and the atmospheric boundary layer are as follows:

[0181] Using the optimal scale parameter a best The corresponding optimal covariance coefficient, as well as characteristic data (relative rate of change, gradient, variance parameter), constructs the data set Where L is the length of the data sequence and M is the data dimension. For example, the length of the laser radar echo signal is 1000 sampling points, and the value of M is 4.

[0182] Normalize the data to prevent the order of magnitude difference of features in different dimensions from being too large, which will affect the clustering results.

[0183] Setting the number of clusters k to 3 clusters the echo signals into three clusters: the aerosol layer, the cloud layer, and the normal atmosphere. However, it is not clear which cluster is the aerosol layer.

[0184] The cluster centers are set to random positions: c i =(W i ,α i ,grd i,var i ), (i=1,2,3), where W is the normalized optimal covariance coefficient, α is the normalized relative rate of change, grd is the normalized gradient, and var is the normalized variance.

[0185] Calculate the Euclidean distance between each sampling point and the cluster center and perform clustering. The calculation formula is as follows:

[0186]

[0187] Among them, x is the sampling point in the data set, c i is the i-th cluster center, dist(x,c i ) is the sampling point x and the cluster center c i The Euclidean distance between j and c ij x and c respectively i The value in the jth dimension.

[0188] Based on the results of the previous round of clustering, update the cluster center of each cluster. The calculation formula is as follows:

[0189]

[0190] in, is the updated i-th cluster center, S i is the i-th type of data set after clustering, ||S i |The number of sample data points in this set.

[0191] Clustering is completed until the maximum number of iterations is reached or the cluster center no longer changes. Three clusters are obtained through K-means clustering.

[0192] Step S33: The three clusters obtained by K-means clustering are determined to be the aerosol layer, the cloud layer, and the normal atmosphere layer; the position where the aerosol concentration changes the most in the aerosol layer is set as the ABLH threshold; and the ABLH candidate value closest to the ABLH threshold is determined as the atmospheric boundary layer height.

[0193] Calculate the weighted sum of the average relative change rate, average gradient and average variance corresponding to the p echo signal sampling data points closest to each cluster center as follows:

[0194]

[0195] Among them, n i is the cluster center c i The location of i is the cluster center c iThe index of i=1, 2, 3 represents the aerosol layer, cloud layer and normal atmosphere layer respectively, k is the index of the cluster, α i (k),grd i (k) and var i (k) are the normalized relative rate of change, gradient and variance respectively; λ1, λ2, λ3 are the weight coefficients of the average relative rate of change, average gradient and average variance respectively.

[0196] For example, the value of p is 40.

[0197] Among them, the variance var i (k) is a positive number, so there is no need to find its absolute value.

[0198] For example, λ1=0.3, λ2=0.5, and λ3=0.2.

[0199] The largest cluster is the cloud layer, The smallest cluster is the general atmosphere layer, and the remaining clusters are the aerosol layer.

[0200] Aerosol concentrations vary at higher locations within the atmospheric boundary layer, reaching their maximum rate of change near the top of the atmospheric boundary layer. This is because turbulent mixing within the atmospheric boundary layer makes the aerosol distribution more uniform in the vertical direction, while the gradient of aerosol concentration is most significant at the top of the atmospheric boundary layer.

[0201] The optimal scale parameter a′ of the aerosol layer is obtained by using the particle swarm algorithm for the preprocessed echo signal corresponding to the aerosol layer. best Then, based on the optimal scale parameter of the aerosol layer, the preprocessed echo signal corresponding to the aerosol layer is subjected to wavelet covariance transformation to obtain the wavelet covariance coefficient curve of the aerosol layer. The position with the largest aerosol concentration change in the aerosol layer is set as the ABLH threshold using the built-in findperks function in MATLAB; the ABLH candidate value closest to the ABLH threshold is determined as the atmospheric boundary layer height.

[0202] The true ABLH value among the candidate ABLH values ​​is determined, and the correct ABLH value is selected from the candidate ABLH values.

[0203] Step S3 calculates the characteristic data (relative rate of change, gradient, and variance) of the preprocessed echo signal and performs K-means clustering based on the optimal covariance coefficient to classify the preprocessed echo signal into three categories: aerosol layer, cloud layer, and normal atmosphere layer. The ABLH threshold is set by analyzing the locations within the aerosol layer where aerosol concentration changes the most. The candidate ABLH values ​​are then selected to determine the value closest to this threshold, which is used as the final atmospheric boundary layer height.

[0204] Example 2:

[0205] Another embodiment of the present invention discloses a laser radar-based atmospheric boundary layer height inversion system, thereby implementing the laser radar-based atmospheric boundary layer height inversion method of embodiment 1. The specific implementation of each module refers to the corresponding description of embodiment 1.

[0206] like Figure 5 As shown, a laser radar-based atmospheric boundary layer height inversion system in this embodiment includes the following modules:

[0207] The echo signal acquisition module M1 is used to use the laser radar to perform vertical detection on the area to be measured, obtain the echo signal and perform preprocessing to obtain the preprocessed echo signal;

[0208] An ABLH candidate value generation module M2 is configured to obtain an optimal scale parameter for wavelet covariance transform using a particle swarm optimization algorithm, and to obtain a wavelet covariance coefficient curve by performing a wavelet covariance transform based on the optimal scale parameter on the preprocessed echo signal; and to use the local maximum position of the wavelet covariance coefficient curve as an ABLH candidate value;

[0209] The ABLH determination module M3 is configured to calculate characteristic data of the preprocessed echo signal, cluster the preprocessed echo signal into three categories: an aerosol layer, a cloud layer, and a normal atmosphere layer based on the characteristic data and the optimal covariance coefficient corresponding to the optimal scale parameter; set the location in the aerosol layer where the aerosol concentration changes the most as the ABLH threshold; and determine the ABLH candidate value closest to the ABLH threshold as the atmospheric boundary layer height.

[0210] In summary, the lidar-based atmospheric boundary layer height inversion method and system according to the embodiments of the present invention have the following beneficial effects:

[0211] 1. This invention optimizes the scale parameters of the wavelet covariance transform by introducing a particle swarm optimization algorithm, which can automatically select the optimal scale parameters, thereby improving the accuracy of atmospheric boundary layer height inversion. In traditional methods, the selection of scale parameters for the wavelet covariance transform relies on empirical formulas or manual judgment, which lacks adaptability. However, this technical solution achieves adaptive optimization of the scale parameters through the particle swarm optimization algorithm;

[0212] 2. The present invention can effectively distinguish between aerosol layers, cloud layers, and ordinary atmosphere layers by combining multiple characteristic data (relative rate of change, gradient, variance) of the echo signal and the optimal covariance coefficient for K-means cluster analysis. This method can reduce the interference of clouds and noise on the inversion of atmospheric boundary layer height and improve the robustness of the algorithm under complex atmospheric conditions. Traditional methods such as the gradient method and the variance method are prone to errors when faced with cloud interference and noise, while the present invention significantly enhances the anti-interference ability of the algorithm through cluster analysis and feature extraction;

[0213] 3. This invention realizes the automation and intelligentization of atmospheric boundary layer height inversion. Through particle swarm optimization algorithm and K-means cluster analysis, the algorithm can automatically identify and screen candidate ABLH values ​​without human intervention. This not only improves the inversion efficiency, but also reduces the influence of human factors on the results, making the inversion of atmospheric boundary layer height more scientific and objective.

[0214] 4. By optimizing the scale parameters of the wavelet covariance transform and combining it with K-means cluster analysis, the present invention can quickly and accurately determine the atmospheric boundary layer height. Compared with traditional methods, this method reduces the amount of computation, improves inversion efficiency, and can provide reliable ABLH results in a shorter time, making it suitable for real-time monitoring and rapid response applications.

[0215] 5. By adjusting the wavelet scale parameters and weight coefficients, the present invention can adapt to the vertical resolution and sampling interval of different lidars, showing strong adaptability. The method works effectively with both high-resolution and low-resolution lidar data, providing a universal inversion method for different types of lidars.

[0216] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by any technician familiar with this technical field within the technical scope disclosed by the present invention should be covered by the scope of protection of the present invention.

Claims

1. A method for inverting the atmospheric boundary layer height based on lidar, characterized in that: The steps include: Step S1: Use a laser radar to perform vertical detection on the area to be measured, obtain an echo signal, and pre-process it to obtain a pre-processed echo signal; Step S2: using a particle swarm optimization algorithm to obtain an optimal scale parameter for a wavelet covariance transform, applying a wavelet covariance transform based on the optimal scale parameter to the preprocessed echo signal to obtain a wavelet covariance coefficient curve; and using the local maximum position of the wavelet covariance coefficient curve as an ABLH candidate value; Step S3: Calculate the characteristic data of the preprocessed echo signal, and cluster the preprocessed echo signal into three categories: aerosol layer, cloud layer, and normal atmosphere layer based on the characteristic data and the optimal covariance coefficient corresponding to the optimal scale parameter; set the position with the largest aerosol concentration change in the aerosol layer as the ABLH threshold; and determine the ABLH candidate value closest to the ABLH threshold as the atmospheric boundary layer height.

2. The method according to claim 1, characterized in that The characteristic data includes relative rate of change, gradient and variance; the step S3 includes: K-means clustering was performed based on the normalized relative rate of change, gradient, variance, and optimal covariance coefficient, and three clusters were obtained; Calculate the weighted sum of the average relative change rate, average gradient and average variance corresponding to the p echo signal sampling data points closest to each cluster center based on Determining categories of the three clusters, the categories including an aerosol layer, a cloud layer, and a general atmosphere layer; Performing a wavelet covariance transform on the preprocessed echo signal corresponding to the aerosol layer to obtain an optimal scale parameter of the corresponding aerosol layer, and obtaining a wavelet covariance coefficient curve of the corresponding aerosol layer based on the optimal scale parameter; Finding the local maximum position of the wavelet covariance coefficient curve of the aerosol layer, where the local maximum position is the position where the aerosol concentration changes the most, and serving as the ABLH threshold; The ABLH candidate value closest to the ABLH threshold among the ABLH candidate values ​​is used as the atmospheric boundary layer height value.

3. The method according to claim 2, characterized in that Calculate the weighted sum of the average relative change rate, average gradient and average variance corresponding to the p echo signal sampling data points closest to each cluster center as follows: Among them, n i is the cluster center c i The location of i is the cluster center c i The index of i=1, 2, 3 represents the aerosol layer, cloud layer and normal atmosphere layer respectively, k is the index of the cluster, α i (k),grd i (k) and var i (k) are the normalized relative rate of change, gradient and variance respectively; λ1, λ2, λ3 are the weight coefficients of the average relative rate of change, average gradient and average variance respectively.

4. The method according to claim 2, characterized in that The largest cluster is the cloud layer, The smallest cluster is the general atmosphere layer, and the remaining clusters are the aerosol layer.

5. The method according to claim 1, characterized in that: The wavelet covariance transform is performed using the particle swarm optimization algorithm to optimize the wavelet scale parameters and obtain the optimal scale parameters, including: Initialize the particle swarm algorithm parameters; Set the particle swarm algorithm input to the wavelet scale parameter a; Based on the current wavelet scale parameter, the preprocessed echo signal is transformed based on the Haar wavelet function to obtain the wavelet covariance coefficient, the fitness value of each particle is calculated based on the corresponding wavelet covariance coefficient, and the optimal position and optimal fitness value of the individual and the group are updated until the maximum number of iterations is reached, and the optimal position and optimal fitness value of the group are obtained; The current wavelet scale parameter corresponding to the optimal position and optimal fitness value of the group is obtained as the optimal scale parameter a best , and a best The corresponding wavelet covariance coefficient curve W best .

6. The method according to claim 5, characterized in that The particle fitness value is calculated using the following fitness function: f(a)=Max(λ(Max(W(a,b))-Min(W(a,b)))-(z1-z2)) Among them, λ is the weight factor, b is the position of the center of the Haar wavelet function, W(a,b) is the wavelet covariance coefficient, z1 and z2 are the heights of the upper and lower half-peak positions of the peak position of the wavelet covariance curve, respectively.

7. The method according to claim 5, characterized in that Use the findperks function built into Matlab to find the wavelet covariance coefficient curve W best The local maximum position of the ABLH is obtained as a plurality of local maximum positions as the candidate values ​​of the ABLH; The multiple local maximum positions include atmospheric boundary layer height and cloud layer height positions.

8. The method according to claim 6, characterized in that The wavelet covariance coefficient W(a,b) is as follows: Among them, a is the wavelet scale parameter, and a = k × Δz, Δz is the vertical resolution of the lidar, b is the center position of the Haar wavelet function, k is the coefficient used to adjust the scale of the wavelet to adapt to different signal characteristics, r is the detection distance of the lidar, r b and r t are the lower and upper limits of the echo signal range, respectively, and N(r) is the echo signal after preprocessing; is the Haar wavelet function; The Haar wavelet function is as follows:

9. The method according to any one of claims 1 to 8, characterized in that The echo signal is an optical signal; and the preprocessing of the echo signal includes: Convert the received optical signal into an analog electrical signal; amplifying the analog electrical signal to obtain an amplified electrical signal; The amplified electrical signal is subjected to background denoising, smoothing filtering, geometric overlap factor correction processing and laser radar detection distance square correction in sequence to obtain a pre-processed echo signal.

10. An atmospheric boundary layer height inversion system based on laser radar, characterized in that: include: An echo signal acquisition module is used to use a laser radar to perform vertical detection on the area to be measured, obtain an echo signal, and perform preprocessing to obtain a preprocessed echo signal; An ABLH candidate value generation module is configured to obtain an optimal scale parameter of a wavelet covariance transform using a particle swarm optimization algorithm, and to obtain a wavelet covariance coefficient curve by performing a wavelet covariance transform based on the optimal scale parameter on the preprocessed echo signal; and to use the local maximum position of the wavelet covariance coefficient curve as an ABLH candidate value; An ABLH determination module is configured to calculate characteristic data of the preprocessed echo signal, cluster the preprocessed echo signal into three categories: an aerosol layer, a cloud layer, and a normal atmosphere layer based on the characteristic data and an optimal covariance coefficient corresponding to the optimal scale parameter; set the location in the aerosol layer where the aerosol concentration changes the most as the ABLH threshold; and determine the ABLH candidate value closest to the ABLH threshold as the atmospheric boundary layer height.

Images