諧波檢測(cè)方法-中英文

1、諧波檢測(cè)方法諧波檢測(cè)是諧波問題中的一個(gè)重要分支，對(duì)抑制諧波有著重要的指導(dǎo)作用， 對(duì)諧波的分析和測(cè)量是電力系統(tǒng)分析和控制中的一項(xiàng)重要工作，也是對(duì)繼電保 護(hù)、判斷故障點(diǎn)和故障類型等工作的重要前提。 準(zhǔn)確、實(shí)時(shí)地檢測(cè)出電網(wǎng)中瞬態(tài) 變化的畸變電流、電壓，是眾多國內(nèi)外學(xué)者致力研究的目標(biāo)。 常規(guī)的諧波測(cè)量方 法主要有三種：模擬帶通或帶阻濾波器的諧波測(cè)量；基于傅里葉變換的諧波測(cè)量；基于瞬時(shí)無功功率的諧波測(cè)量。一、模擬濾波器：最早的諧波檢測(cè)方法是采用模擬濾波器來實(shí)現(xiàn)的。輸入信號(hào)經(jīng)放大后送入一 組帶通濾波器，濾波器的中心頻率 fl、f2、fn是固定的，為工頻的整數(shù)倍， 且flf2fn,然后送多路顯示器中顯示測(cè)量

2、信號(hào)中所含諧波成分及其幅值。 該方法的實(shí)現(xiàn)電路簡(jiǎn)單，造價(jià)低，輸出阻抗低，品質(zhì)因數(shù)易于控制。但是電路較 難設(shè)計(jì)，誤差大，對(duì)電網(wǎng)頻率波動(dòng)和電路元件參數(shù)很敏感，使得檢測(cè)出的諧波中含有較多的基波分量，運(yùn)行損耗大。二、快速傅立葉變換：隨著計(jì)算機(jī)和微電子技術(shù)的發(fā)展，開始采用傅立葉分析的方法來檢測(cè)諧波和 無功電流，有離散傅立葉變換（DFT和快速傅立葉變換（FFT）兩種。這種方法 根據(jù)采集到的一個(gè)電源周期的電流值進(jìn)行計(jì)算，最終得出所需的諧波和無功電 流，它可以精確地分析和檢測(cè)整數(shù)次諧波，目前應(yīng)用比較廣泛，技術(shù)也相當(dāng)成熟。 但由于傅立葉變換要進(jìn)行兩次變換， 計(jì)算量大，計(jì)算時(shí)間長(zhǎng)，所以檢測(cè)時(shí)間比較 長(zhǎng)。三、瞬時(shí)無

3、功功率理論：1983年日本學(xué)者赤木泰文提出的瞬時(shí)無功功率理論，即p - q理論解決了 諧波和無功功率的瞬時(shí)檢測(cè)和不用儲(chǔ)能元件就能實(shí)現(xiàn)抑制諧波和無功補(bǔ)償?shù)葐?題，從而使得電力有源濾波理論由實(shí)驗(yàn)室的理論研究走向工作應(yīng)用。根據(jù)該理論，可以得到瞬時(shí)有功功率p和瞬時(shí)無功功率q , p和q中都含有直流分量和交流分 量。由此可得被檢測(cè)電流的基波分量，將基波分量與總電流相減即得相應(yīng)的諧波 電流。因?yàn)樵摲椒ê雎粤肆阈蚍至?，?duì)于不對(duì)稱系統(tǒng)，瞬時(shí)無功的平均分量不等 于三相的平均無功。所以，該方法只適用于三相電壓正弦、 對(duì)稱情況下的三相電 路諧波和基波無功電流的檢測(cè)。四、一種基于正交三角級(jí)數(shù)神經(jīng)網(wǎng)絡(luò)的諧波檢測(cè)方法。根

4、據(jù)電力系統(tǒng)中非正弦周期電流的分解形式， 提出了一種基于正交三角級(jí)數(shù) 神經(jīng)網(wǎng)絡(luò)的諧波檢測(cè)方法。方法能同時(shí)檢測(cè)出非正弦周期電流中的基波分量與各 次諧波分量的幅值和相位以及有功電流和無功電流。通過仿真實(shí)例驗(yàn)證，該方法能夠把整數(shù)次諧波進(jìn)行有效分離，用相對(duì)較少的數(shù)據(jù)量達(dá)到了較高的檢測(cè)精度。 在電力系統(tǒng)中，由于非線性負(fù)載的廣泛應(yīng)用，向電網(wǎng)注入了大量的諧波電流，使 供電系統(tǒng)中的元件損耗增大，給電力系統(tǒng)中的設(shè)備運(yùn)行帶來很大危害。 為了防止 諧波危害系統(tǒng)安全運(yùn)行，就必須確切掌握電力系統(tǒng)中畸變波形含有諧波的實(shí)際情 況，采取相應(yīng)措施對(duì)其進(jìn)行抑制或補(bǔ)償。FFT法是當(dāng)今應(yīng)用得最多的諧波檢測(cè)方 法，但FFT法在實(shí)際應(yīng)用中

5、存在著頻譜泄漏問題，使得算出的各次諧波精度不高。 將神經(jīng)網(wǎng)絡(luò)方法應(yīng)用于電力系統(tǒng)諧波研究處于起步階段，在諧波源辨識(shí)、諧波預(yù)測(cè)與測(cè)量以及電力系統(tǒng)負(fù)荷預(yù)測(cè)等方面取得了一些成果。 根據(jù)非正弦周期電流的 分解式，提出了一種基于正交三角級(jí)數(shù)神經(jīng)網(wǎng)絡(luò)的諧波檢測(cè)方法，本質(zhì)上是利用三層神經(jīng)網(wǎng)絡(luò)的函數(shù)逼近性能給出了一種信號(hào)分離（分解）的方法，該方法能夠同時(shí)檢測(cè)出非正弦周期電流中的基波分量與各次諧波分量的幅值和相位以及有 功電流和無功電流，具有較高的檢測(cè)精度。所用神經(jīng)網(wǎng)絡(luò)結(jié)構(gòu)簡(jiǎn)單，激活函數(shù)采 用一組正交的三角函數(shù)，算法容易實(shí)現(xiàn)，網(wǎng)絡(luò)收斂速度較快。五、一種基于自適應(yīng)神經(jīng)元在線整定的諧波檢測(cè)方法，用于電流信號(hào)處理 技術(shù)

6、領(lǐng)域。步驟如下：將電壓作為參考輸入，負(fù)載電流作為原始輸入，通過實(shí)時(shí)運(yùn)算 輸出與負(fù)載電流基波有功分量幅值、 相位均相等的信號(hào)，將此信號(hào)從負(fù)載電流中 扣除后，得到諧波和無功電流分量的總和；同時(shí)為了提高計(jì)算的速度和精度，對(duì) 學(xué)習(xí)速率和積分環(huán)節(jié)的進(jìn)行在線整定： 通過對(duì)諧波電流的采樣值數(shù)據(jù)窗分析， 判 斷諧波是否發(fā)生或突變，在檢測(cè)的初始階段或者電流突變情況下，學(xué)習(xí)速率先取較大值，再取較小值；通過權(quán)值理論計(jì)算式可計(jì)算出權(quán)值初始值的估計(jì)值，從而以優(yōu)化的初始值重新進(jìn)行積分。本發(fā)明明顯提高了算法速度和精度， 優(yōu)化了算法 實(shí)時(shí)計(jì)算性能。六、基于線調(diào)頻小波變換的電機(jī)故障信號(hào)諧波檢測(cè)方法。線調(diào)頻小波變換統(tǒng)一了短時(shí)Fo

7、urier變換和小波變換的時(shí)頻分析，并能根據(jù) 信號(hào)的特點(diǎn)自適應(yīng)生成新的時(shí)頻窗口。本文首次將線調(diào)頻小波變換引進(jìn)電力系統(tǒng) 的突變信號(hào)處理中。分析了其消噪和濾除干擾的原理；構(gòu)造了線調(diào)頻小波變換的 算法。該算法不僅能解決消噪和濾除干擾的問題， 還能解決關(guān)于濾除整數(shù)（偶數(shù)） 次和分?jǐn)?shù)次諧波，并通過對(duì)電力系統(tǒng)突變信號(hào)處理的實(shí)例說明該算法的突出優(yōu) 點(diǎn)。近年來，短時(shí)Fourier變換和小波變換在電力系統(tǒng)故障診斷、檢測(cè)、定位、 識(shí)別以及信號(hào)消噪、重構(gòu)等方面的應(yīng)用也有很大的進(jìn)展。短時(shí) Fourier 變換 是一種使用固定大小的時(shí)頻分析窗口的 Fourier變換，適用于分析具有固定不變 帶寬的突變信號(hào)；小波變換使用

8、時(shí)間和頻率軸可伸縮的長(zhǎng)方形時(shí)頻分析窗口，適用于分析具有固定比例帶寬（恒Q,即濾波器品質(zhì)因數(shù)不變）的突變信號(hào)。這些使 得它們?cè)陔娏ο到y(tǒng)中信號(hào)處理某些方面如干擾、偶次諧波和非整數(shù)次諧波濾波等 的應(yīng)用受到一定的限制。因此尋找具有近似等Q的時(shí)頻窗口的時(shí)頻分析工具是非 常必要的，它除了時(shí)間平移，頻率平移和時(shí)頻拉伸外，還應(yīng)考慮矩形窗口的斜方 向的拉伸與旋轉(zhuǎn)變化。線調(diào)頻小波變換滿足上述要求，使用的時(shí)頻分析窗口除了時(shí)移、 頻移、尺度 變化以外，最主要的是包含了時(shí)頻窗口在時(shí)頻平面上的放置以及在傾斜方向上的 尺度變化（拉伸）。由于使用各種長(zhǎng)方形和各種平行四邊形的時(shí)頻窗口，所以線調(diào)頻小波變換可以分析具有非固定不變帶

9、寬和非固定比例帶寬（非恒Q）的突變信號(hào)。信號(hào)的消噪、濾除干擾、壓縮、恢復(fù)以及故障信號(hào)檢測(cè)、診斷、識(shí)別、定位 是電力系統(tǒng)信號(hào)處理的主要工作，其目的是盡可能地復(fù)原被噪聲或干擾污染的信 息源以及故障的特征和類型。嚴(yán)格地講，干擾和噪聲是兩個(gè)不同的概念。 干擾指 周期的、有規(guī)律的誤差信號(hào)（測(cè)量信號(hào)與真實(shí)信號(hào)的差）；而理論上不能預(yù)測(cè)的、 必須用概率統(tǒng)計(jì)刻畫劃的誤差信號(hào)定義為噪聲。電力系統(tǒng)在采樣信號(hào)時(shí)，現(xiàn)場(chǎng)存 在大量噪聲和干擾信號(hào)，嚴(yán)重影響了系統(tǒng)、設(shè)備監(jiān)測(cè)的靈敏度和可靠性，因此消 除干擾和濾掉噪聲是電力設(shè)備監(jiān)測(cè)的一個(gè)關(guān)鍵技術(shù)問題。在電力系統(tǒng)中，快速傅立葉（FFT）閾值濾波法和最小均方誤差（LMS）自適應(yīng)濾

10、波器是最常用的用來抑制干擾和消噪方法。但是，F(xiàn)FT閾值濾波不能消除平穩(wěn)隨 機(jī)型干擾，而LMS自適應(yīng)濾波器收斂性能受時(shí)延、 收斂因子等參數(shù)的影響，濾波 效果不穩(wěn)定，甚至有時(shí)不收斂?；谛〔ㄗ儞Q的干擾濾波器研究不多，文7在干擾濾波方面作了嘗試，它將干擾信號(hào)分為脈沖型干擾、連續(xù)周期型干擾和平 穩(wěn)隨機(jī)型干擾。主要討論連續(xù)周期型和平穩(wěn)隨機(jī)型干擾信號(hào)的抑制。 仔細(xì)分析后， 文中對(duì)平穩(wěn)隨機(jī)型干擾即白噪聲進(jìn)行了基于小波變換的分析處理，對(duì)有色噪聲未涉及。對(duì)連續(xù)周期型干擾濾波論及不多，小波變換對(duì)這類干擾應(yīng)該也不是有效的。 文10論及到連續(xù)周期型干擾濾波問題， 它提出了 3次B樣條小波對(duì)采樣信號(hào) 進(jìn)行預(yù)處理的方法，

11、可基本消除偶次諧波和1.5次以上非整數(shù)次諧波。The method of The Harmonics examinationThe Harmonics examination is Harmonics an important branch within problem. which has an important instruction function to repress Harmonics. Is a power system analysis and control a key job to the harmonic analysis and the diagraph. Is al

12、so to after electricity protection, judgment the breakdown order and break down a type etc. work of important premise. Now how to accurate, quickly examine the distortion current, voltage is the target that numerous domestic and international scholars concentrate on a research. The Harmonics of norm

13、al measures method mainly have three methods: The Harmonics of measure according to the analogous filter; The Harmonics of measure according to Fourier transform; The Harmonics of measure according to the instantaneous reactive power.First、(The analogous filter) the earliest Harmonics examinations m

14、ethod.The Input signal sends into a set of band pass filter after amplify. The center frequency of filter f1 、f2、fn is fixed. Which is for the integral of run frequency doubly. f1 f2 fn. Then send many manifestation in and value. This method carries out electric circuit in brief, building pr the roa

15、d displays diagraph signal composition in the Harmonics contained ice low, output resistance is low.The article prime factor is easy to a control. But the circuit is more difficult design, the error margin is big, to motion and circuit component of the charged barbed wire net frequency parameter ver

16、y sensitive, Maketo there is more base frequency in the Harmonics of examining weight, circulate to exhaust greatly.Second FFT(Fast Fourier transform)Along with the technical development of the calculator and the micro-electronics, Start to adopting the method that the Fourier analysis to examine Ha

17、rmonics and reactive current. This kind of method according to collect of an electric current value of power supply period carry on a calculation. Finally get the Harmonics needed by reactive current. It can analyze by the square time with examination integral Harmonics. But because the Fourier tran

18、sform carries on two transformation. The calculation has great capacity, computing time is long, So examination time also is longer.Third、(Instantaneous reactive power)The Japanese scholar puts forward in 1983 of the moment has Instantaneous reactive power theories. Then thep q theories solvesHarmon

19、ics and reactive power of the moment examine with need not keep ability the component can carry out to repress Harmonics and reactive power compensation etc. problem. The theories research which makes active filter the theories from the laboratory thus heads for a work application. According to this

20、 theories. Wecan get to have a instantaneous active power p and instantaneous reactive power q . All imply direct current weight and exchanges weight in pandq. Fromhere the base frequency can be measured. The base frequency weights reducing mutually with total current have to correspond of Harmonics

21、 current. Because of that method neglected zero preface weights. The methods is applicable to three mutually voltage and under symmetry circumstance of three mutually electric circuit Harmonics.Fourth、( Method of Harmonics Measurement Based on Neural Network of Orthogonal Trigonometric Series. )Base

22、d on the neural network of orthogonal trigonometric series,an approach for h armonics measurement in power systems is presented in this paper.Using this method, the fundamental component and harmonics can be detected simultaneously with less data quantity. The simulation results validate that harmon

23、ics can be separa ted from a signal with high accuracy by the method developed in the paper. In the electrical power system, as a result of the misalignment loads widespread application, has poured into the massive harmonic current to the electrical network, causes in power supply systems part to lo

24、se increases, brings the very big harm for electrical power systems in equipment movement. In order to prevent the overtone harm system safety movement, must grasp in the electrical power system the distortion profile to include the overtone actual situation accurately, takes the corresponding measu

25、res to carry on to it suppresses or the compensation.The FFT law applies most overtoneexamination method now, but the FFT law has the frequency spectrum divulging problem in the practical application, causes various subharmonics precision which figures out not to be high. Applies the neural network

26、method in the electrical power system overtone research is at the start stage, in the overtone source identification, the overtone forecast and the survey as well as aspects and so on electrical power system load forecast has made some progresses. According to the non-sinusoidal periodic currents de

27、composition type, proposed one kind based on the orthogonal trigonometric series neural networks overtone examination method, essentially uses three neural networks the approximation of function performance has given one kind of signal separation (decomposition) the method, this method can simultane

28、ously examine in the non-sinusoidal periodic current the fundamental wave component with various subharmonics component peak-to-peak value and the phase as well as the wattful current and the idle current, has the high examination precision. Uses the neural network structure to be simple, the activa

29、tion function uses group of orthogonals easily the trigonometric function, the algorithm to realize, the network convergence rate is quick.Fifth ( An adaptive neuron-tuning the harmonic detection measurement.)The step is as follows : Takes the reference input the voltage, the load current takes the

30、primitive input, through the true-time operation output and the load current fundamental wave active component peak-to-peak value, the phase equal signal, this signal deducts after the load current, obtains the overtone and the idle current component sum total; Simultaneously to enhance the computat

31、ion the speed and the precision, to studies the speed and the integration element carries on the online installation:Through to harmonic current sampling value datawindow analysis, judgment overtone whether to have or the sudden change, in the examination preliminary stage or the electric current su

32、dden change situation, the study speed preoccupies the great value, takes the small value again; May calculate the weight starting value through the weight theory formula the estimated value, thus optimizes the starting value carries on the integral. This invention raised the algorithm speed and the

33、 precision obviously, optimized the algorithm real-time computation performance.Sixth、(Based OnThe Wavelet Transform Of The FMSignal Harmonic Motor Fault Detection Method. )LineFMwavelet transform unified short-time Fourier transform of thewavelet transform and time-frequency analysis, and according

34、 to the characteristics of signals generated new adaptive time-frequency window. This is the first line will be the introduction of FM wavelet transform power system mutant signal processing. Analysis of its denoising and filtering interference principle; constructed Line FMwavelet transform algorit

35、hm. The algorithm can not only solve the denoising and filtering interference issues, but also address the filtering integer (even), and scores of harmonics, and through mutation on the power signal processing system illustrated by the example of the outstanding merits of the algorithm.In recent yea

36、rs, the short-time Fourier transform and wavelet transform in the power system fault diagnosis, detection, location, identification and signal denoising, such as remodeling the application had a great progress. Short-termFourier transform is the use ofa fixed-size window of time-frequency analysis o

37、f the Fourier transform, apply to a fixed bandwidth of the mutant signal; wavelet transform the use of time and frequency axis scalable rectangular window of time-frequency analysis, applicable to Analysis of a fixed ratio bandwidth (Q constant, the quality factor is the same filter) mutant signal.

38、These allow them in the power system of certain areas such as signal processing interference, even harmonics and non-integer harmonics filtering, the application subject to certain restrictions. Therefore, such as Qfind a similar time-frequency window of time-frequency analysis tool is essential, in

39、 addition to its translation time, frequency, frequency translation and tension, we must also consider the rectangular window in the direction of the ramp stretching and rotation changes.Line FM wavelet transform to meet the above requirements, the use of time-frequency analysis window in addition t

40、o times, frequency shift, scale change, the most important thing is included in the time-frequency window on the time-frequency plane and placed in the direction of the tilt-scale changes ( tensile). Because of the use of rectangular parallelogram and the time-frequency window, so lines can be analy

41、zed FM wavelet transform a non-fixed bandwidth and bandwidth ratio of non-fixed (non-constant Q) mutant signal.Signal denoising, filtering interference, compression, recovery and fault signal detection, diagnosis, identification, localization signal processing power system is the main work, the aim

42、is to recover as much as possible interference by noise or pollution source of information as well as the characteristics of fault and type. Strictly speaking, interference and noise are two different concepts. Interference that cycle, the law of error signal (measuring signal and the signal real di

43、fference), and the theory can not be predicted, we must use probability and statistics portray zoned for the definition of the error signal noise. Power system in the sampling signal, the scene there are a lot of noise and interference signals, which has seriously affected the system, and the sensit

44、ivity of monitoring equipment reliability, eliminate the interference and noise is filtered out electrical equipment monitoring a key technical issue.In the power system, Fast Fourier (FFT) threshold filtering method and the minimum mean square error (LMS) adaptive filter is the most commonly used to inhibit interference and noise elimination method. However, the FFT threshold filtering can not remove stationary randomtype interference, and LMS adaptive filter convergence performance by the delay, the convergence of parameters, such as the impact of filtering effect of instability, and so