阅读说明：本技术 一种非平稳天气系统高频强风的分级检测方法 (Grading detection method for high-frequency strong wind of non-stationary weather system ) 是由 刘辉 张炜 吴迪 田宏强 徐晶晶 陈思威 费晓虎 于 2021-08-06 设计创作，主要内容包括：本发明公开了一种非平稳天气系统高频强风的分级检测方法,包括以下步骤：1)测得瞬时风速；2)利用由瞬时风速表征的风脉动序列,提取高频风脉动；3)检测高频风脉动偏离高斯分布的阈值；4)利用分级的阈值判断高频风特性。针对的是秒到分钟时间尺度高频强风检测技术的研究空白,可以对高频破坏性强风进行更精准的预警。(The invention discloses a grading detection method for high-frequency strong wind of a nonstationary weather system, which comprises the following steps: 1) measuring the instantaneous wind speed; 2) extracting high-frequency wind pulsation by using a wind pulsation sequence represented by instantaneous wind speed; 3) detecting a threshold value of deviation of the high-frequency wind pulsation from Gaussian distribution; 4) and judging the high-frequency wind characteristics by utilizing the graded threshold value. The method aims at the research blank of the second-minute time scale high-frequency strong wind detection technology, and can perform more accurate early warning on high-frequency destructive strong wind.)

1. A grading detection method for high-frequency strong wind of a non-stationary weather system is characterized by comprising the following steps:

1) measuring the instantaneous wind speed;

2) extracting high-frequency wind pulsation by using a wind pulsation sequence represented by instantaneous wind speed;

3) detecting a threshold value of deviation of the high-frequency wind pulsation from Gaussian distribution;

4) and judging the high-frequency wind characteristics by utilizing the graded threshold value.

2. The hierarchical detection method according to claim 1, characterized in that: in the step 1), the instantaneous wind speed is measured by a cup anemometer.

3. The hierarchical detection method according to claim 1, wherein in step 2), the process of extracting the high-frequency wind pulsation includes:

21) the original wind pulsation sequence is decomposed into a series of internal model functions, which are recorded as:

where U (t) is the original sequence of wind pulsations, x_i(t) is the ith internal model function, N is the number of the internal model functions decomposed by the time series, and epsilon (t) is the residual low-frequency trend;

22) calculating each internal model function x_i(t) average frequency, adding the internal model functions in the order of seconds to minutes to obtain high-frequency wind pulsation U' (t), which is recorded as:

wherein M is the number of internal model functions in the high frequency interval.

4. The hierarchical detection method according to claim 1, wherein in step 3), the process of extracting the high-frequency wind pulsation includes:

by analyzing the probability density function of the high-frequency wind pulsation U' (t) and comparing the probability density function with the Gaussian distribution, the position where the probability density function begins to deviate from the Gaussian distribution is found and is set as a threshold value H₁Absolute value greater than threshold H₁Is considered to be a type of extreme wind U'_extreme1(t); the threshold for starting to deviate from the stable distribution is set to H₂Absolute value greater than threshold H₂Is considered to be a secondary extreme value wind U'_extreme2(t)。

5. The classification detection method according to claim 1, wherein in step 4), the method for determining the high-frequency wind characteristics is:

a) if the high-frequency wind pulsation completely falls within the range of Gaussian distribution, the wind is judged to be type A wind, and the wind is only provided with the base flow of the wind pulsation, is not high-frequency strong wind, and cannot harm outdoor equipment;

b) if the high-frequency wind pulsation falls between the Gaussian distribution and the stable distribution, judging the high-frequency wind pulsation to be B-type wind; at this time, according to the threshold value H₁Extracting extreme value wind U's'_extreme1(t) calculating the Hilbert transient spectrum of the wind, wherein if the amplitude of the Hilbert transient spectrum of extreme wind pulsation exceeds a threshold value, the wind is quite strong, destructive is generated, and outdoor equipment is possibly damaged;

c) if the high-frequency wind pulsation falls outside the stable distribution, judging the high-frequency wind pulsation to be C-type wind; at this time, according to the threshold value H₂Extracting two types of extreme value wind U'_extreme2(t) and harmful high frequency strong winds whose Hilbert spectra exceed a threshold can be a serious hazard to outdoor equipment.

Technical Field

The invention belongs to the technical field of meteorological monitoring, and particularly relates to a grading detection method for high-frequency strong wind of a nonstationary weather system.

Background

The destructive strong wind generally occurs in the conditions of strong weather processes such as thunderstorms, squall lines, typhoons, sand dust, cold tides and the like, and the weather has strong nonlinear and non-stationary characteristics and has very important significance for monitoring and early warning of extreme disaster weather.

For destructive gusts, existing research has focused on weather scale gusts and paroxysmal atmospheric turbulence scale. For example, thunderstorm storm is one of the most frequently occurring and easily missed disastrous weather in summer. With the diversification of observation means, the development of novel detection technologies such as wind profile radar and boundary layer gradient wind provides further support for the fine research of the convection storm structure. The method comprises the following steps of coating duckweed and the like (coating duckweed, Yaohai, paint beam wave, and the like; 2014, analyzing the characteristics of a primary disastrous gale Doppler radar and a boundary layer in the north of Zhejiang province [ J ]. plateau meteorology, 33(6):1687 1696.), disclosing the cause of the primary disastrous gale in Zhejiang province based on various data such as 2 Doppler radars, a wind profile radar, a 370 m high tower and the like, and indicating that the combination of a strong convection monomer outflow boundary and a gust front leads to the disastrous gale and analyzing the characteristics of the boundary layer before and after the influence of the gust front. Chen Wen super et al (Chen Wen super, Liu Yi Jun, Song Li, etc.. 2019, an example analysis of wind characteristics of different strong wind weather systems [ J ] weather, 45(2): 251) and 262 ], analyzes the characteristics of average wind and pulsating wind of the boundary layer near the ground of different strong wind weather systems, and finds that typhoon, strong convection strong wind and cold air strong wind have different wind characteristics due to different generation mechanisms. The wind speed of strong convection changes most violently, the average turbulence intensity is far larger than that of typhoon and strong cold air strong wind, and the turbulence energy spectrum value of the typhoon strong wind is obviously higher than that of the strong convection and the strong cold air strong wind. If continuous and dense atmospheric turbulence data can be obtained, the statistical value of the Wind field characteristics can well reveal the continuous variation trend of the average speed, Wind direction angle, turbulence intensity, gust factor, turbulence integral scale and pulsating Wind speed power spectral density function and the strong Wind pulsation law within a short time period (Cao S Y, Tamura Y, Kikuchi N, et al 2009, Wind characteristics of a strong type Wind.

Generally, destructive strong winds are divided into two types, one type of strong wind is expressed in that the average wind speed of low frequency is far beyond a normal value, for example, the maximum wind speed of super-strong typhoon can exceed 12 levels, and the instantaneous wind speed of disastrous gust caused by thunderstorm is larger than 17m/s, so that a lot of research is already carried out on the strong winds; another type of strong wind is also very damaging, as indicated by its moderate average wind speed, but its high frequency pulsations can frequently give rise to extreme wind speeds that deviate significantly from the average. The more frequent the extreme wind occurs and the higher the energy intensity, the more destructive it is, which increases the fatigue load of the outdoor equipment, reduces the service life of the equipment and even directly causes permanent damage. For such high frequency gusts, existing research has focused on the scale of atmospheric turbulence at frequencies of 10-20Hz, while much less research has been done on gusts on the second to minute time scale. Meanwhile, as the ultrasonic anemoscope is required for acquiring turbulence data, the wind pulsation data of a turbulence scale is difficult to obtain by a common meteorological station.

Disclosure of Invention

The invention aims at providing a grading detection method for high-frequency strong wind of a non-stationary weather system, aims at the blank of research on a second-minute time scale high-frequency strong wind detection technology, and can perform more accurate early warning on high-frequency destructive strong wind.

In order to solve the technical problems, the invention adopts the following technical scheme:

a grading detection method for high-frequency strong wind of a non-stationary weather system comprises the following steps:

1) measuring the instantaneous wind speed;

2) extracting high-frequency wind pulsation by using a wind pulsation sequence represented by instantaneous wind speed;

3) detecting a threshold value of deviation of the high-frequency wind pulsation from Gaussian distribution;

4) and judging the high-frequency wind characteristics by utilizing the graded threshold value.

In the above technical scheme:

in the step 1), the instantaneous wind speed is measured by a cup anemometer.

In step 2), the process of extracting the high-frequency wind pulsation comprises the following steps:

21) the original wind pulsation sequence is decomposed into a series of internal model functions, which are recorded as:

where U (t) is the original sequence of wind pulsations, x_i(t) is the ith internal model function, N is the number of the internal model functions decomposed by the time series, and epsilon (t) is the residual low-frequency trend;

22) calculating each internal model function x_i(t) average frequency, adding the internal model functions in the order of seconds to minutes to obtain high-frequency wind pulsation U' (t), which is recorded as:

wherein M is the number of internal model functions in the high frequency interval.

In step 3), the process of extracting the high-frequency wind pulsation comprises the following steps:

by analyzing the probability density function of the high-frequency wind pulsation U' (t) and comparing the probability density function with the Gaussian distribution, the position where the probability density function begins to deviate from the Gaussian distribution is found and is set as a threshold value H₁Absolute value greater than threshold H₁Is considered to be a type of extreme wind U'_extreme1(t); the threshold for starting to deviate from the stable distribution is set to H₂Absolute value greater than threshold H₂Is considered to be a secondary extreme value wind U'_extreme2(t)。

In step 4), the method for judging the high-frequency wind characteristic comprises the following steps:

a) if the high-frequency wind pulsation completely falls within the range of Gaussian distribution, the wind is judged to be type A wind, and the wind is only provided with the base flow of the wind pulsation, is not high-frequency strong wind, and cannot harm outdoor equipment;

b) if the high-frequency wind pulsation falls between the Gaussian distribution and the stable distribution, judging the high-frequency wind pulsation to be B-type wind; at this time, according to the threshold value H₁Extracting extreme value wind U's'_extreme1(t) and calculating its Hilbert transient spectrum if one type of extreme windIf the amplitude of the pulsating Hilbert transient spectrum exceeds a threshold value, the wind of the type is quite strong, so that the wind is destructive and can possibly damage outdoor equipment;

c) if the high-frequency wind pulsation falls outside the stable distribution, judging the high-frequency wind pulsation to be C-type wind; at this time, according to the threshold value H₂Extracting two types of extreme value wind U'_extreme2(t) and harmful high frequency strong winds whose Hilbert spectra exceed a threshold can be a serious hazard to outdoor equipment.

The invention has the beneficial effects that:

the invention can detect and early warn destructive high-frequency strong wind in advance by utilizing instantaneous wind speed information obtained by the existing cup type anemometer of the automatic meteorological station, and fills the blank of research on high-frequency wind pulsation from second to minute. Meanwhile, advanced signal analysis methods such as Hilbert-Huang transform and the like are utilized, the latest research result of the field of atmospheric turbulence and wind engineering on high-frequency wind characteristics is combined, a set of destructive high-frequency strong wind grading detection technology is developed, and the result has wide application value in the fields of disaster weather early warning, building wind safety, wind engineering and the like.

Detailed Description

The invention firstly needs to extract the high-frequency information of the wind pulsation sequence. Destructive strong wind generally occurs in the case of strong weather processes such as thunderstorms, sand dust, cold tides and the like, and the weather has strong non-stationary characteristics, so that a time sequence analysis method suitable for a non-stationary weather system is required to be adopted. The Empirical Mode Decomposition (EMD) method is different from the traditional signal analysis method, and the method does not preset a basis function, but decomposes the signal according to the characteristic scale contained in the signal itself to obtain an internal model function of a finite order, so that each order of internal model function has a relatively clear physical meaning. The method has no requirements on data stationarity and linearity, has the characteristic of self-adaptability, can better retain the characteristics of the original signal, and is very suitable for extracting the high-frequency wind pulsation in a non-stationary weather system.

Whether the high-frequency wind pulsation reaches the standard of strong wind and whether the high-frequency wind pulsation is destructive or not is related to two characteristic quantities of the high-frequency wind pulsation. The extreme value characteristic of the pulsation; the second is the pulsating energy intensity. Studies have shown that the most damaging to buildings and outdoor equipment is not the average wind of high wind speeds, but the extreme wind speeds that deviate significantly from the average. The more frequently the extreme wind occurs and the higher the energy intensity, the stronger the destructiveness, which can increase the fatigue load of outdoor equipment, reduce the service life of the equipment, even directly cause permanent damage, and the large deviation theory can be used for engraving extreme characteristics of wind pulsation.

The central limit theorem considers random pulsations around the mean and the corresponding probability distribution is gaussian. The large deviation theory describes the characteristics of the tail of probability distribution, is also called as the theory of rare events, is suitable for describing extreme events of a random process, and meets the requirement of a type of distribution called as stable distribution. In fact, if the limit of variance in the central limit theorem is removed, the sum distribution of a plurality of independent identically distributed random variables still approaches a distribution which is a stable distribution^[5]Therefore, a gaussian distribution is a special case of a stable distribution with a finite variance. As the variance of the wind speed pulsations becomes larger and the extrema occur more and more frequently, their probability distribution begins to deviate from the gaussian distribution. In this case, the wind speed pulsations are not completely random and have a strong correlation.

In summary, whether the high-frequency wind pulsation satisfies the gaussian distribution or the stable distribution is one of the criteria for determining whether the wind pulsation is destructive; at the same time, extreme wind speeds also require a certain energy intensity, which is the second criterion for the generation of destructive strong winds. Generally speaking, the stronger the weather system, the more capable it is to sustain vortical motion that produces destructive weather, and we use Hilbert spectral analysis to calculate the energy level of extreme winds. The EMD method and Hilbert spectrum analysis are combined to obtain complete Hilbert-Huang transform (HHT for short), and the Hilbert transient spectrum has better local time-frequency drawing capability than a power spectrum, can perform intra-wave frequency modulation and inter-wave frequency modulation on nonlinear effects, and is suitable for processing nonlinear and non-stationary data. A disruptive high frequency gust can be considered to have occurred when the amplitude of the energy level of the Hilbert transient spectrum at frequencies on the order of seconds to minutes exceeds a threshold.

In the embodiment, the instantaneous wind speed is measured by a wind cup type anemometer, and the method further comprises the following steps:

1) extraction of high frequency wind pulsations using Empirical Mode Decomposition (EMD) method

Taking a section of wind pulsation sequence of 1-2 hours as an example, information of a second-minute magnitude is required to be extracted, and the wind pulsation of medium and low frequency which is larger than a minute level is close to average wind. Performing empirical mode decomposition on an original wind pulsation sequence U (t), and specifically comprising the following steps:

the EMD decomposition of the time series u (t) is achieved by a screening process. EMD initially extracts the highest frequency oscillations, defining two envelopes: one passing through all local maxima and the other passing through all local minima of the time series. The average of the two envelopes, called the mean envelope, is subtracted from the original time series to obtain a new series, and this process is repeated until the remaining signal satisfies the condition of becoming an internal model function, i.e. each pair of local maximum and minimum points is separated by a zero crossing. When the first internal model function x is obtained₁(t) subtracting it from the original time series U (t) to obtain U₁(t), repeating this process from U₁(t) extracting to obtain a second internal model function x₂(t) of (d). When the original time series is decomposed into a series of internal model functions, it can be written as:

where N is the number of time series decomposed as an internal model function and epsilon (t) is the remaining low frequency trend, called residual.

Each internal model function x is then calculated_i(t) average frequency, adding the internal model functions in the order of seconds to minutes, we get the high frequency wind pulsation U' (t):

where M is the number of internal model functions in the high frequency range.

2) Threshold for detecting deviation of dithering from Gaussian distribution

The high frequency wind pulsations extracted from step 1 also have different components, and the wind pulsations can be considered to be of a pulsating "base stream" U'_mean(t) and extreme value wind U'_extreme(t) composition. The two parts have different physical mechanisms and statistical properties and therefore different statistical distributions. The former only contributes to the central region of the probability density function and satisfies the Gaussian distribution; and extreme value wind with strong discreteness only contributes to the tail part of the probability density function, so that stable distribution is met. Reference (Uchaikin V, Zolotarev M.1999.chance and Stability: Stable Distributions and the same Applications [ M]VSP,570pp), the characteristic function obtained after fourier transformation of the probability density function of the stable distribution has the following form:

where Φ (k) is a stably distributed characteristic function, the parameters α and β are called shape parameters, and α ∈ (0, 2)],β∈[1,1]γ is called a scale parameter, δ is called a position parameter, and γ ∈ [0, ∞ ], δ ∈ (— ∞, ∞). When α is 2, the stable distribution is an average of 2 γ²(ii) a gaussian distribution of; when α ≠ 2, the variance of the stationary distribution diverges, and when α ≦ 1, the mean of the stationary distribution diverges.

Therefore, by analyzing the probability density function of the high-frequency wind pulsation U' (t), and comparing the probability density function with the gaussian distribution, the position where the probability density function starts to deviate from the gaussian distribution is found, and the threshold H is set₁Absolute value greater than threshold H₁Is considered to be a type of extreme wind U'_extreme1(t); the threshold for starting to deviate from the stable distribution is set to H₂Absolute value greater than threshold H₂Is considered to be a secondary extreme value wind U'_extreme2(t)。

3) Determining high frequency wind characteristics using a graded threshold

Then, the characteristic of the high-frequency wind pulsation is judged by using the classification threshold value obtained in the step 2), and the steps are as follows:

a) if the high-frequency wind pulsation completely falls within the range of Gaussian distribution, the high-frequency wind pulsation is A-type wind, only has the basic flow of the wind pulsation, is not high-frequency strong wind, and cannot cause harm to outdoor equipment;

b) falling between the gaussian and stationary distributions is the class B wind. If the pulsation is B-type wind, according to a threshold value H₁Extracting extreme value wind U's'_extreme1(t) and calculating its Hilbert transient spectrum^[4]If the amplitude of the Hilbert spectrum of extreme value wind pulsation exceeds a threshold value, the extreme value wind pulsation indicates that the wind has considerable strength and can generate destructiveness, and possibly the extreme value wind pulsation needs to attract attention to damage outdoor equipment;

c) outside the stable distribution is the class C wind. If the pulsation is C-type wind, according to a threshold value H₂Extracting two types of extreme value wind U'_extreme2(t) and harmful high frequency strong winds whose Hilbert spectra exceed a threshold can be a serious hazard to outdoor equipment.