Method for generating simulation data of coherent wind lidar echo signal spectrum

Abstract

A coherent wind lidar echo signal spectrum simulation data generation method is provided, comprising: obtaining the bottom noise of the echo signal; removing the bottom noise from the echo signal spectrum data and counting the variance of the random noise; performing normalization processing on the random noise and performing Gaussian fitting to obtain the amplitude, Doppler shift and bandwidth of the echo signal; recording the Doppler shift difference between adjacent range gates to obtain the power spectrum peak energy of a single spectrum data and the correction amount of the distance; respectively counting the statistical parameters of the random variable, the power spectrum peak energy, the correction amount of the distance and the bandwidth; generating a group of wind speed profile graphs by using random walk and normal distribution characteristics of random variables, generating bandwidths under different range gates that satisfy the Lorentz distribution, power spectrum peak energies of a single simulated spectrum that satisfy the chi-square distribution, and distance correction amounts that satisfy the normal distribution; and generating a single simulated spectrum by a radar equation and repeatedly generating complete spectrum simulation data.

Description

Technical Field

[0001] This disclosure relates to the fields of laser remote sensing and optical technology, and in particular to a method for generating spectral simulation data of coherent wind-measuring lidar echo signals. Background Technology

[0002] In recent years, with the development of erbium-doped fiber amplifiers (EDFAs) and fiber technology, the application of large-mode-field fiber diameters has significantly improved the output power of fiber lasers. Coherent wind radar based on fiber lasers, with its compact structure, excellent performance, high wind measurement accuracy, and fast time response, is receiving increasing attention in applications such as boundary layer atmospheric wind profile measurement, wind shear early warning, aircraft wake detection, and wind energy resource utilization. Typically, we use pulse pairs, discrete spectrum peak maximum likelihood estimation, and Gaussian fitting to invert wind speed. However, as the signal-to-noise ratio decreases, the inversion accuracy drops significantly, and the inversion distance is limited. Using neural networks to build models for wind speed inversion can better distinguish wind speed locations at low signal-to-noise ratios, but training such models requires a large amount of simulation data. Traditional lidar equations only provide a general simulation process, neglecting several statistical characteristics in the echo signal. Summary of the Invention

[0003] To address the aforementioned issues, this disclosure provides a method for generating simulated spectrum data of coherent wind-measuring lidar echo signals, thereby alleviating the aforementioned technical problems in the prior art.

[0004] (I) Technical Solution

[0005] This disclosure provides a method for generating simulated spectrum data of coherent wind-measuring lidar echo signals, comprising: obtaining the noise floor of the echo signal through acquired real echo signal spectrum data; removing the noise floor from each spectrum data in the echo signal spectrum data and calculating the variance of random noise; normalizing the random noise; performing Gaussian fitting on the echo signal at each range gate in each normalized spectrum data to obtain the amplitude, Doppler frequency shift, and bandwidth of the echo signal; recording the Doppler frequency shift difference between adjacent range gates, and using inverse ratio fitting to obtain the number of single spectrum data. The peak energy of the power spectrum and the distance correction are obtained; statistical parameters are obtained by statistically analyzing the random variable, the peak energy of the power spectrum, the distance correction, and the bandwidth; a set of wind speed profile maps are generated using random walk and the normal distribution characteristics of random variables; bandwidths at different distances satisfying the Lorentz distribution are generated using a random number generator; the peak energy of the power spectrum of a single simulated spectrum satisfying the chi-square distribution and the distance correction satisfying the normal distribution are obtained; and a single simulated spectrum image is generated through the simplified radar equation, and the complete spectrum simulation data is generated by repeating this process.

[0006] According to embodiments of this disclosure, the noise floor S d (f) is represented as:

[0007]

[0008] in, Indicates the nth i The spectral distribution at each distance gate is given by t, where t is the cumulative number of distance gates, and i represents the corresponding index of a distance gate sequence.

[0009] According to embodiments of this disclosure, the variance of random noise is expressed as:

[0010]

[0011] Where M is the number of sampling points on a single range gate, t is the cumulative number of range gates, i is the corresponding index under a range gate sequence, and j is the index corresponding to the spectral frequency under a single range gate. For the power spectrum signal at the nth time... i The power spectral intensity value dispersed at the j-th frequency value on the distance gate, S d (f j ) represents the power spectral intensity value of the noise floor at the j-th frequency, which is dispersed across a single distance gate.

[0012] According to embodiments of this disclosure, the amplitude A1(R) of the echo signal is obtained by performing Gaussian fitting on the echo signal at each distance gate in each normalized spectral data, and is expressed as:

[0013]

[0014] Where R is the distance, A0 represents the energy of a single laser pulse, η(R) is the antenna efficiency, and T(R) is the one-way atmospheric transmission loss.

[0015] According to embodiments of this disclosure, the Doppler frequency shift f(R) of the echo signal is obtained by performing Gaussian fitting on the echo signal at each distance gate in each normalized spectral data, and is expressed as:

[0016] f(R)=f d (R);

[0017] Among them, f d (R) is a function of Doppler frequency shift as a function of distance.

[0018] According to embodiments of this disclosure, the bandwidth σ1(R) of the echo signal is obtained by performing Gaussian fitting on the echo signal at each distance gate in each normalized spectral data, and is expressed as:

[0019] σ(R)=σ(R);

[0020] Where σ(R) is a function of the spectral bandwidth as a function of distance.

[0021] According to an embodiment of this disclosure, the Doppler frequency shift difference p between adjacent distance gates is recorded and expressed as:

[0022] p = f((i+1)R0) - f(iR0);

[0023] Where R0 is the range resolution, f(iR0) is the Doppler frequency shift at the i-th range gate, and f((i+1)R0) is the Doppler frequency shift at the (i+1)-th range gate.

[0024] According to embodiments of this disclosure, histogram counting is used to statistically analyze random variables, peak power spectrum energy, distance correction, and bandwidth to obtain corresponding statistical parameters.

[0025] According to embodiments of this disclosure, the simplified lidar equation is expressed as:

[0026]

[0027] Wherein, the peak power energy of a single analog spectrum is A², R is the distance, σ(R) is the function of spectral bandwidth as a function of distance, b is the distance correction offset under inverse square fitting, and f is the spectral frequency. d (R) represents the Doppler frequency shift at distance R, N(0, σ) noise ) represents a normally distributed random noise, σ noise The mean squared error is the random noise.

[0028] According to embodiments of this disclosure, the random number generators used include Lorentz random number generators and chi-square random number generators.

[0029] (II) Beneficial Effects

[0030] As can be seen from the above technical solution, the method for generating coherent wind lidar echo signal spectrum simulation data disclosed herein has at least one or a part of the following beneficial effects:

[0031] (1) The distribution characteristics of each parameter in the lidar equation were verified, and the lidar equation was simplified, which simplified the complexity of spectrum generation.

[0032] (2) The simulated echo signal spectrum can approximate the real spectrum data to the greatest extent, providing a statistical reference for signal distribution that cannot be analyzed at high altitudes;

[0033] (3) It has a high degree of simplicity, and can start directly from the spectrum data without discussing the parameter calculations caused by the complex system parameters. It also provides a scientific reference for the learning of neural networks and the wind speed inversion process based on it, reducing the human error that the inversion algorithm itself may bring. Attached Figure Description

[0034] Figure 1a A schematic diagram of Gaussian fitting for the echo signal at the near-field gate;

[0035] Figure 1b A schematic diagram of Gaussian fitting for the echo signal at a long-distance gate;

[0036] Figure 2 This is a schematic diagram comparing the neural network method and the traditional centroid algorithm for wind speed inversion.

[0037] Figure 3 This is a schematic diagram illustrating the working principle of the method for generating simulated data of coherent wind-measuring lidar echo signal spectrum according to an embodiment of the present disclosure.

[0038] Figure 4 This is a flowchart illustrating the method for generating simulated spectrum data of coherent wind-measuring lidar echo signals according to an embodiment of this disclosure. Detailed Implementation

[0039] This disclosure provides a method for generating simulated spectral data of coherent wind-measuring lidar echo signals, which helps improve the reconstruction of the real wind field from the simulated signal, especially by statistically analyzing the high signal-to-noise ratio signal at low altitudes to infer the distribution pattern of the signal at higher altitudes. Combining this method with a neural network model can reduce noise interference with the signal itself, providing a foundation for reconstructing the wind field at high altitudes with low signal-to-noise ratios.

[0040] Coherent lidar targets atmospheric aerosols. It receives the backscattered signals and the local master beat frequency, generating a radio frequency heterodyne signal. This signal is then subjected to FFT and incoherent frequency-domain accumulation. A suitable algorithm is used to estimate the Doppler frequency shift and derive the radial wind speed. Commonly used Doppler frequency shift estimation algorithms include pulse pair techniques, maximum likelihood discrete peak estimation, and Gaussian fitting. For coherent wind lidar, because it receives scattered signals from soft atmospheric targets, the echo signal generated by a single pulse is very weak. Wind speed inversion is essentially a frequency estimation problem for weak signals. Therefore, it is usually necessary to perform frequency-domain accumulation of multiple pulses to obtain a high signal-to-noise ratio and improve inversion accuracy. Assuming the average value of aerosol particles is f... d M data samples are taken within the distance gate, with a sampling interval of T. r Assuming that the amplitude of each pulse echo signal remains constant and is 'a' within the same range gate, then when sampling the echo signal of a single pulse within the range gate of the target location, it can be considered as sampling the echo signal of a pulse with frequency f.d The sine wave is sampled. The sampled value of the target at the i-th sampling point can then be expressed as:

[0041] i = 0, 1, ..., M-1;

[0042] M is the number of sampling points within a single range gate. If coherent accumulation is used, and the number of FFT sampling points is N, then the output of the Doppler channel in the frequency domain is:

[0043]

[0044] Generally, Doppler lidar emits pulse signals with a Gaussian time-domain distribution, and the sampling sequence contains uncorrelated noise:

[0045]

[0046] Its autocovariance is not zero, which can be expressed as:

[0047] R kl = <x k x l * >=R SN exp[[(-2(πwT r ) 2 ]exp(2πjf d kT r )+δ k ;

[0048] R sv The ratio of average signal power to noise power, w is the power spectral width, and δ is the signal-to-noise ratio. k Let Kronecker function be the periodogram. In this case, using the periodogram maximum value method under incoherent accumulation can greatly simplify the calculation, and the calculated periodogram can be regarded as the power spectrum of the obtained echo signal. The periodogram is defined as:

[0049] k = 0, 1, ..., M-1;

[0050] Using incoherent accumulation, i.e., ignoring the phase information of each pulse echo, the amplitude is accumulated to obtain Z, where Z represents the number of accumulated pulses.

[0051]

[0052] This yields discrete spectrum data of size L×M, where L is the number of range gates, M is the number of sampling points on a single range gate, the sampling zero-frequency position is the AOM modulation frequency shift, and the spectral resolution is:

[0053]

[0054] The signal strength at each range gate of the collected single-spectrum signal all follows a Gaussian distribution. Based on the lidar equations, the relationship between signal power and range and frequency should be:

[0055]

[0056] Where R is the distance, η(R) is the antenna efficiency, T(R) is the one-way atmospheric transmission loss, and σ noise Let $\mathbf{a}$ be the mean square error of the Gaussian white noise. It's easy to see from the above equation that the signal power decreases with the square of the distance. At the near-range gate, the signal is strong, and Gaussian fitting, maximum likelihood estimation, or the spectral centroid algorithm can effectively invert the position of the Doppler frequency shift, thus accurately obtaining wind speed information. However, at the far-range gate, the signal power decreases sharply, and traditional algorithms can no longer accurately invert the position of the Gaussian peak from the periodogram. In this case, a better option is to use a neural network algorithm to invert the radial wind speed.

[0057] like Figure 1a and Figure 1b As shown, at the near-range gate (distance R=5), the signal is very strong, and the curve can be well fitted with Gaussian. However, as the distance increases, at the range gate (distance R=40), the signal-to-noise ratio decreases, and the error caused by using Gaussian fitting and the spectral centroid algorithm to obtain the center position also increases. Figure 2 A comparison of wind speed inversion using a neural network method and the traditional centroid algorithm is presented. The neural network model was trained using data generated by this simulation method. The light-colored areas near the curves represent the error range of the wind speed inversion. It can be seen that the centroid algorithm has an extremely limited range for wind speed inversion, and its error range decreases sharply with increasing distance, at which point the signal is essentially submerged in noise. At close range, the neural network inversion accuracy is lower than the centroid algorithm, but its error range is less affected by noise and signal strength, and the true wind speed value is basically contained within the error range. This demonstrates that using a neural network model to establish wind speed inversion at low signal-to-noise ratios is unmatched by traditional algorithms.

[0058] Neural network algorithms construct the relationships between different nodes or layers through a pre-defined theoretical or data model to achieve the purpose of data or information processing. In order to make the neural network model more applicable to the coherent lidar spectrum, it needs to be trained extensively. The usual practice is to use equation (1) to generate training data. However, equation (1) only gives the Gaussian distribution characteristics of the echo signal spectrum in a general sense, but does not give the distribution characteristics of its spectral intensity, width, etc. with time or distance. It has been proven that the spectral data has good statistical distribution characteristics under high signal-to-noise ratio conditions. These characteristics include the trend of wind speed change, the distribution characteristics of spectral bandwidth, and the correction that the Gaussian peak value at high altitudes decreases inversely with the square of the distance. Due to the incoherent accumulation of pulses, the received pulse signals have different phases, which means that the signal amplitude obtained directly from the frequency domain is lower than that of the time domain signal. There is a spectral leakage phenomenon. Therefore, the laser single pulse energy value A0 after leakage is no longer a specific constant in the power spectrum. This also shows that directly establishing the simulation of the spectral signal has higher determinism for wind speed inversion.

[0059] Based on the above considerations, this invention aims to establish a method for generating spectrum simulation data. The spectrum of coherent wind-measuring radar echo signals generated by this method has the following advantages: First, the spectrum simulation data generated by this method can better represent the real wind field than the simulation data generated by simply using lidar equations, thereby improving the efficiency and accuracy of neural network training; second, by generating a large amount of near-real spectrum data for neural network training, it can provide the possibility for wind speed inversion at long distances with low signal-to-noise ratios and high speeds; third, it is universally applicable to coherent lidars of different models and operating wavelengths.

[0060] To make the objectives, technical solutions, and advantages of this disclosure clearer, the following detailed description is provided in conjunction with specific embodiments and the accompanying drawings.

[0061] In this embodiment of the disclosure, a method for generating simulated spectrum data of coherent wind-measuring lidar echo signals is provided, such as... Figure 3 and Figure 4 As shown, the method for generating simulated spectrum data of coherent wind-measuring lidar echo signals includes operations S1-S8:

[0062] Operation S1: Obtain the noise floor of the echo signal by acquiring the actual echo signal spectrum data;

[0063] Operation S2: Remove background noise from each spectrum data in the echo signal spectrum data and calculate the variance of random noise;

[0064] Operation S3: Normalize the random noise;

[0065] Operation S4: Perform Gaussian fitting on the echo signal at each distance gate in each normalized spectral data to obtain the amplitude, Doppler shift, and bandwidth of the echo signal;

[0066] Operation S5: Record the Doppler frequency shift difference between adjacent distance gates, and use inverse ratio fitting to obtain the power spectrum peak energy and distance correction amount of a single spectrum data;

[0067] Operation S6: Perform statistical analysis on the random variable, peak power spectrum energy, distance correction, and bandwidth to obtain the corresponding statistical parameters;

[0068] Operation S7: Generate a set of wind speed profiles using random walk and the normal distribution characteristics of random variables; generate bandwidths at different distance gates satisfying the Lorentz distribution using a random number generator; obtain the peak power spectrum of a single simulated spectrum satisfying the chi-square distribution; and obtain the distance correction amount satisfying the normal distribution.

[0069] Operation S8: Generate a single simulated spectrum image using the simplified radar equation; and further repeat operations S1-S8 to generate more simulated spectrum images to meet the required number, thus completing the generation of complete spectrum simulation data.

[0070] Suppose the obtained spectrum data has a size of N×L×M, where N is the number of spectrum samples, L is the number of range gates, and M is the number of sampling points on a single range gate. Let R be the spectral distribution at the nth range gate, taken from a single spectrum sample. n (f), the noise floor distribution is S d (f), the signal distribution is S n (f), then it is obvious that:

[0071] R n (f)=S d (f)+S n (f)+N(0,σ) noise );

[0072] When n is large, to reduce the impact of randomness caused by noise, a series of n1, ..., n are taken. t By averaging the spectral sums at these distance gates, the noise floor can be expressed as:

[0073]

[0074] in, Indicates the nth i The spectral distribution at each distance gate is given by t, where t is the cumulative number of distance gates, and i represents the corresponding index of a distance gate sequence.

[0075] Noise removal and noise normalization are performed, where the root mean square error σ of Gaussian white noise (random noise) is... noiseIt can be represented as:

[0076]

[0077] Where M is the number of sampling points on a single range gate, t is the cumulative number of range gates, i is the corresponding index under a range gate sequence, and j is the index corresponding to the spectral frequency under a single range gate. For the power spectrum signal at the nth time... i The power spectral intensity value dispersed at the j-th frequency value on the distance gate, S d (f j ) represents the power spectral intensity value of the noise floor at the j-th frequency, which is dispersed across a single distance gate.

[0078] According to the lidar equation:

[0079]

[0080] The spectral Gaussian fitting parameters for a single distance gate are as follows:

[0081]

[0082] f(R)=f d (R);

[0083] σ1(R)=σ(R);

[0084] Where R is the distance, A0 represents the energy of a single laser pulse, η(R) is the antenna efficiency, and T(R) is the one-way atmospheric transmission loss. d σ(R) is a function of Doppler frequency shift as a function of distance. σ(R) is a function of spectral bandwidth as a function of distance.

[0085] Additionally: η(R)=1 / (1+z) R / R 2 Analysis of a large number of echo signals shows that when R is large, the lidar equation can be rewritten as:

[0086]

[0087] Let σ(R, t), A2(t), and b(t) be random variables that satisfy a certain statistical distribution, namely the Lorentz distribution, the chi-square distribution, and the normal distribution, respectively, and these distributions can be considered to be independent in terms of distance and time.

[0088]

[0089]

[0090]

[0091] Let the nth i The wind speed detected by the distance gate is f i , nth i+1 The detected wind speed is f i+1 Let the random variable p = f i+1 -f i If so, it satisfies the characteristics of a normal distribution. In fact, the variance of this normal distribution gradually increases with distance, but the increase is slow.

[0092]

[0093] Based on the existing data, analyze the statistical values ​​of each parameter, and use σ p A set of simulated wind speed locations is generated using random walks. The initial wind speed is denoted as f0, and a recursive formula (generally μ) is used. p =0):

[0094] f i+1 =f i +A p N(0, σ) p )

[0095] This allows us to simulate a set of wind speed values, which can then be substituted into the following equation for f. d In (R):

[0096]

[0097] The distance correction b is determined by a normal distribution N(μ) b , σ b The random number generator is used to generate σ(R), and similarly, σ(R) is generated by a random number generator with parameter (A). σ x c A Lorentz random number generator (w) generates L sets of data, where L is the number of distance gates. The generation of A2 follows a chi-square distribution with n degrees of freedom. This yields a single spectral image. Repeating this process until the statistical regularity is evident results in a good set of simulated spectral data.

[0098] In the embodiments disclosed herein, such as Figure 3 and Figure 4As shown, a set of echo signal spectrum data with a specification of 14742×100×100 was obtained from the coherent wind lidar. Assuming that no effective signal can be detected starting from the 70th range gate, the signals from the last 30 range gates of each spectrum data are accumulated and averaged to obtain the echo signal noise floor. The corresponding noise floor is subtracted from each spectrum, and the variance of the random noise is calculated. The random noise is normalized by dividing the spectrum by the variance of the random noise. Next, each processed spectrum is analyzed, and Gaussian fitting is performed on the signal at each range gate to obtain the corresponding amplitude A1(R), Doppler frequency shift f(R), and bandwidth σ1(R), for a total of N×L sets. Record the Doppler frequency shift difference between adjacent range gates as p = f((i+1)R0) - f(iR0), where R0 is the range resolution, f(iR0) is the Doppler frequency shift at the i-th range gate, and f((i+1)R0) is the Doppler frequency shift at the (i+1)-th range gate. Using the inverse proportionality:

[0099]

[0100] The values ​​of single spectrum A2 and b are obtained. Considering the specifications of the initial data, if the influence of weak signals at high altitudes and the inability to perform Gaussian fitting is not considered, then a total of 14742×100 sets of f(R) and σ1(R), and 14742 sets of A2 and b values ​​are obtained. The random variables p, A2, b and σ1(R) are statistically analyzed respectively, and the corresponding statistical parameters (6) are obtained by using the statistical methods mentioned in the technical content section. The statistics here are performed by histogram counting.

[0101] By utilizing random walks and the normal distribution characteristics of p, a set of wind speed profiles can be generated, while the bandwidth σ at different distances can be obtained using a Lorentz random number generator. i (R), a total of L groups. Similarly, a value for A2 and b can be obtained using a chi-square random number generator (7). Consider the simplified lidar equation:

[0102]

[0103] Wherein, the peak power energy of a single analog spectrum is A², R is the distance, σ(R) is the function of spectral bandwidth as a function of distance, b is the distance correction offset under inverse square fitting, and f is the spectral frequency. d (R) represents the Doppler frequency shift at distance R, N(0, σ) noise ) represents a normally distributed random noise, σ noise This represents the mean square error of the random noise. Based on this, a spectral image can be generated.

[0104] When 10,000 or more spectral data points are generated, the characteristics of the simulated data become increasingly closer to those of real spectral images. Using these simulated spectra generates highly accurate and reliable training of subsequent neural network models, thus achieving effective wind speed retrieval.

[0105] The embodiments of this disclosure have been described in detail above with reference to the accompanying drawings. It should be noted that implementations not illustrated or described in the drawings or the main text of the specification are forms known to those skilled in the art and are not described in detail. Furthermore, the definitions of the various elements and methods described above are not limited to the specific structures, shapes, or methods mentioned in the embodiments, and those skilled in the art can easily modify or substitute them.

[0106] Based on the above description, those skilled in the art should have a clear understanding of the method for generating coherent wind-measuring lidar echo signal spectrum simulation data disclosed herein.

[0107] In summary, this disclosure provides a method for generating simulated spectral data of coherent wind-measuring lidar echo signals. It presents the parametric characteristics of the echo signals, enabling the trained neural network model to achieve high wind speed inversion accuracy. This is beneficial for improving the reconstruction of the real wind field using simulated signals, especially by statistically analyzing high signal-to-noise ratio signals at low altitudes to infer the distribution patterns of signals at higher altitudes. Using this method in conjunction with the neural network model reduces noise interference with the signal itself, providing a foundation for reconstructing wind fields at high altitudes with low signal-to-noise ratios. The generated simulated spectral data better represents the real wind field than simulated data generated solely using lidar equations, thus improving the efficiency and accuracy of neural network training. Secondly, generating a large amount of near-realistic spectral data for neural network training makes it possible to invert wind speed at high altitudes with low signal-to-noise ratios over long distances. Thirdly, it is universally applicable to different models and operating wavelengths of coherent lidar.

[0108] It should also be noted that the above are different embodiments provided by this disclosure. These embodiments are used to illustrate the technical content of this disclosure and are not intended to limit the scope of protection of this disclosure. A feature of one embodiment can be applied to other embodiments through suitable modifications, substitutions, combinations, or separations.

[0109] It should be noted that, unless otherwise specified herein, having "a" element is not limited to having a single element, but may include one or more of the element.

[0110] Furthermore, unless otherwise specified, the ordinal numbers such as "first," "second," etc., used herein are merely for distinguishing multiple elements with the same name and do not indicate any hierarchy, order of execution, or process sequence among them. A "first" element and a "second" element may appear together in the same component or separately in different components. The presence of an element with a higher ordinal number does not necessarily indicate the presence of another element with a lower ordinal number.

[0111] In this document, unless otherwise specified, the term "characteristic A" or "and / or" and "characteristic B" means that A exists alone, B exists alone, or A and B exist simultaneously; the term "characteristic A" and "and" or "and" and "and" and "characteristic B" means that A and B exist simultaneously; the terms "including", "containing", "having", and "containing" refer to, but are not limited to, these.

[0112] Furthermore, in this document, terms such as "up," "down," "left," "right," "front," "back," or "between" are used only to describe the relative positions of multiple elements and can be extended to include translation, rotation, or mirroring. Additionally, unless otherwise specified, the statement "one element is on another element" or similar statements do not necessarily indicate that the element is in contact with the other element.

[0113] Furthermore, unless specifically described or required to occur in a specific order, the order of the above steps is not limited to those listed above and can be varied or rearranged according to the desired design. Moreover, the above embodiments can be used in combination with each other or with other embodiments based on design and reliability considerations; that is, technical features from different embodiments can be freely combined to form more embodiments.

[0114] The specific embodiments described above further illustrate the purpose, technical solutions, and beneficial effects of this disclosure. It should be understood that the above descriptions are merely specific embodiments of this disclosure and are not intended to limit this disclosure. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this disclosure should be included within the protection scope of this disclosure.

Claims

1. A method for generating simulated data of a coherent wind lidar return signal spectrum, comprising: obtaining a noise floor of a return signal from real return signal spectrum data; removing the noise floor and calculating a variance of random noise for each spectrum data in the real return signal spectrum data; normalizing the random noise; performing Gaussian fitting on return signals at each range gate in each spectrum data after normalization to obtain an amplitude, a Doppler shift, and a bandwidth of the return signals; recording a Doppler shift difference between adjacent range gates, and using inverse fitting to obtain a power spectrum peak energy and a correction of range for a single spectrum data; respectively calculating statistical parameters of random variables, power spectrum peak energies, corrections of range, and bandwidths; generating a set of wind speed profiles using random walk and normal distribution of random variables, and using a random number generator to generate bandwidths under different range gates satisfying a Lorentz distribution, power spectrum peak energies of a single simulated spectrum satisfying a chi-square distribution, and corrections of range satisfying a normal distribution; and generating a single simulated spectrum picture using a simplified radar equation, and repeatedly generating complete simulated spectrum data. 3.The method of claim 1, wherein the variance of the random noise is represented as: σ 2 ( R ) = σ 2 ( R ) + σ 2 ( R ).

2. The method of claim 1, wherein the noise floor S d (f) is represented as: wherein represents the nth i spectral distribution at the i-th distance gate, t is the cumulative number of distance gates, and i represents the corresponding subscript under a sequence of distance gates. 4.The method of claim 1, wherein the Gaussian fitting on return signals at each range gate in each spectrum data after normalization obtains an amplitude A1(R) of the return signals, which is represented as: A 1 ( R ) = A 0 ( R ) η ( R ) T ( R ) e - ( R 2 ) . wherein M is the number of sampling points on a single range gate, t is the cumulative number of range gates, i is the corresponding index under a range gate sequence, j is the corresponding index of the spectral frequency under a single range gate, S(n, j) is the power spectrum intensity value of the power spectrum signal at the nth i range gate at the jth frequency value, S d (f j ) is the power spectrum intensity value of the bottom noise at the jth frequency value under a single range gate. R is a distance, A0represents a single pulse energy of a laser, η(R) is an antenna efficiency, and T(R) is a single-path atmospheric transmission loss. wherein 5.The method of claim 4, wherein the Gaussian fitting on return signals at each range gate in each spectrum data after normalization obtains a Doppler shift f(R) of the return signals, which is represented as: f ( R ) = f ( R ) + f ( R ). 6.The method of claim 5, wherein the Gaussian fitting on return signals at each range gate in each spectrum data after normalization obtains a bandwidth σ1(R) of the return signals, which is represented as: σ 1 ( R ) = σ ( R ). f(R) = f d (R); wherein, f d (R) is a function of the Doppler shift as a function of distance. σ(R) is a function of a spectrum bandwidth changing with distance. 7.The method of claim 6, wherein a Doppler shift difference p between adjacent range gates is recorded, which is represented as: p = f ( ( i + 1 ) R 0 ) - f ( i R 0 ) . wherein R0is a distance resolution, f(iR0) is a Doppler shift at an i-th range gate, and f((i+1)R0) is a Doppler shift at an (i+1)-th range gate. 8.The method of claim 1, wherein the random variables, the power spectrum peak energies, the corrections of range, and the bandwidths are calculated using histogram counting to obtain corresponding statistical parameters. ​ wherein, ​ ​ 9. The method of claim 1, wherein the simplified lidar equation is represented as: wherein The power spectrum peak energy of the single analog spectrum is A2, R is the distance, σ(R) is a function of the spectrum bandwidth changing with the distance, b is the distance correction offset under the square inverse ratio fitting, f is the spectrum frequency, f d (R) is the Doppler shift under the distance R, N(0, σ noise ) is the noise normal distribution random, σ noise is the mean square error of the random noise.

10. The method of claim 1, wherein the random number generator utilized comprises a Lorentz random number generator, a chi-squared random number generator.

Images