TMCnet News
Research for the Mixed Disturbance Detection of Power System Using LMD Algorithm [Sensors & Transducers (Canada)](Sensors & Transducers (Canada) Via Acquire Media NewsEdge) Abstract: In order to realize the accurate identification of mixed disturbance signal in power system, the local mean decomposition (LMD) algorithm is applied to the mixed disturbance detection in power system for the first time. The typical power quality mixed disturbance signal include harmonics and voltage flicker signal, harmonics and voltage swell signal, harmonics and voltage sag signal, harmonics and voltage interruption signal, as well as the actual mixed disturbance signals occurred in smart substation, are selected and analyzed by LMD algorithm. Disturbance signal is adaptively decomposed into a number of Product Function (PF for short) by the algorithm, and the PF is made of the envelope signal and pure Frequency Modulation signal. We can get the original signal of frequency and amplitude distribution curves. Simulation results show that LMD algorithm is better than HHT algorithm in the parameter fluctuation of transient characteristic parameter detection, the detection accuracy, the end effect and running time. Detection results of Smart Substation shows that, the amplitude, frequency, start and end time of disturbance signal can be accurately detected by LMD algorithm, proving the correctness of the LMD algorithm. Copyright © 2013 IFSA. Keywords: Local Mean Decomposition, Mixed disturbance, End effect, Harmonic, HHT, Power quality. (ProQuest: ... denotes formulae omitted.) 1. Introduction In twenty-first Century, the energy crisis has become increasingly prominent, energy saving and emission reduction, green energy, sustainable development has become the theme of the development of the electric power. With the arrival of the era of smart grid, power load become more and more complex and diverse. With a large number of applications of nonlinear, impact resistance, unbalanced load characteristics and harmonic serious device, Cause a series of power quality problems, such as voltage sags, voltage fluctuation and voltage flicker, waveform distortion and transient disturbance. It is particularly necessary and urgent for the scientific research on the power quality problems [1]. At present, the detection methods of power quality transient disturbance have many methods, including FFT algorithm, wavelet transform, S transform, Hilbert-Huang Transform (HHT) and other theoretical combination algorithm. Fourier transform can not handle non-linear, non-stationary signals, spectral leakage and fence phenomena also exist between treatments harmonic shortcomings; analysis of non-linear, non-stationary signals wavelet theory has many limitations, must construct divide strict and energy concentration wavelet basis [2-5], S transform is an inheritance and development of wavelet transform, S transform can analyze amplitude of each frequency disturbance signal, but there are some errors of S transform to detect the amplitude of Composite disturbance signal contains harmonics [6-10]; HHT power quality detection methods achieved good results, but the decomposition of the modal experience using a cubic spline interpolation fitting the envelope signal is easy to appear envelope, owe envelope phenomenon; HHT in the excessive number of "screening" led to the end effect of pollution throughout the data segment and the instantaneous frequency based HHT time frequency analysis methods often appear to be negative is a physical phenomena which is difficult to explain [11-12], Typical power quality disturbance signal include harmonics, voltage flicker signal, voltage swell signal, voltage sag signal, voltage interruption signal, transient oscillation signal, transient pulses signal and frequency fluctuation and mixed disturbance signal, which is composed of two or more single disturbance signal. Most of the research is to solve single disturbance problem, the study of mixed disturbance problem is less. Jonathan Smith proposed a new adaptive timefrequency analysis method local mean decomposition (local mean decomposition LMD) in 2005 [13], LMD time-frequency analysis method is to decompose the original signal into a series of product function group (the Product Function, PF), the layers PF by the envelope signal and pure FM signal is composed of two parts, they contain all of the instantaneous amplitude and instantaneous frequency information, the further combination when you can get the original signal frequency distribution. The LMD method has been successfully applied to the detection of EEG, the instantaneous frequency of the signal extraction and mechanical fault diagnosis [14-15]. LMD algorithm is applied to detection the disturbance time, frequency and amplitude of the mixed disturbance signal in power system for the first time. Simulation and experimental results show that LMD Algorithm is effective, and has better locate accuracy and computing speed. 2. LMD Principle 2.1. Local Mean Decomposition Process The basic calculation of the local mean decomposition flow chart shown in Fig. 1, the decomposition process of the LMD is a triple cyclic process: the first cyclic sliding smoothing strike a local mean function mi(t) and envelope estimation function ai (t), and the loop terminates conditions sliding smoothed signal adjacent dots are not equal; the two cyclic strike PFi (t) the process loop termination condition to strike out the sin (t) is a pure FM signal; the third cyclic process is for the purpose of obtaining all PFi (t), the loop terminates conditions residual component Uk (t) and only one extreme point. Pure FM signal and the envelope signal is separated from the original signal after three cycles, the pure FM signal and the envelope signal is obtained by multiplying the first PF component, and then gradually loop process, the decomposition of the PF component further determine the instantaneous frequency and instantaneous amplitude, we can obtain the complete time-frequency distribution of the original signal. 2.2. Local Mean Decomposition Algorithm Local mean decomposition can decompose any complicated signal into a number of the PF component which has a certain physical meaning and, each PF component by the plain envelope signal and FM signal integrated. For a signal x (t), the decomposition step is as follows: Step 1 : to determine the signal x (t) of all local extreme point ni. Step 2: through each extremum point ni, calculate any two adjacent local extreme point mean mi and the value of the envelope estimate ai ... (1) ... (2) Connect the adjacent local mean point mi and mi+i with broken line, and then conduct smooth handling by using sliding average algorithm to get the local mean function mn(t). Connect each adjacent envelope estimate values ai and ai+i with broken line, and then conduct smooth handling by using the moving average method to get the envelope estimate function an (t). Step 3: separate the local mean function mu (t) from the original signal x (t), and obtain the signal hn(t): ... (3) Step 4: divide hn(t) by the envelope estimate function an (t), get FM signal su (t): ... (4) Determine whether the Su (t) is a pure FM signal, the determination condition is to repeat the above steps for su (t), get the envelope estimation function ai2 (t) satisfies the ai2 (t)=l, and if not satisfy described and su (t) is not a pure FM signal and then repeat n times until si" (t) is a pure FM signal, i.e. Sin (t) the envelope of the estimation functions satisfy ai(n+i)(t)=l, so: ... (5) ... (6) Conditions for iterative terminated: ... (7) In practical application, in order to avoid excessive decomposition number, we can set a disturbance A, the iteration will end when 1-A<ain(t) <1+A. Step 5: multiply the iterative process envelope estimation function, get the envelope signal ai (t): ... (8) Step 6: obtain the formula (8) in the envelope signal ai (t) and pure FM signal Si" (t) multiplied, to obtain the original signal x (t), as a PF component: ... (9) The first PF component contains the highest frequency component of the original signal. Step 7: separate PF1(T) from the original signal x (t) to get ui (t) as a new data to repeat the above ... (10) As can be seen from the above steps, the original signal can be reconstructed by Uk(t) and all PF components, i.e.: ... (11) 2.3. Based on the Instantaneous Frequency of the LMD Strike By formula (11), the signal is decomposed into a number of PF component and, each PF components represented by the pure envelope signal a(t) and pure FM function s (t)=coscp(t), its frequency can be pure FM function s (t) directly solve, namely: ... (12) Expand the formula (12) and the derivative can be calculated the instantaneous frequency of s (t) the corresponding component of the instantaneous frequency of the PF. s(t) values between ± 1, if s(t) value is approximately equal to ± 1, ± 1 instead because it is derived by the derivative of the cosine function of the instantaneous frequency of the PF. This method of obtaining frequency is intuitive and simple, referred to as "direct method", and compared to the method of the instantaneous frequency of HHT transform strike to strike the instantaneous frequency of the "direct method" is always a positive value, does not appear HHT negative frequencies phenomenon. 3. Power Quality Disturbance Signal Model The utility grid ideally should provide a constant frequency, sine wave, standard voltage, stable electric service reliability for the power user. In a three-phase power system should meet the phase amplitude equal size, phase symmetry and mutual difference is 120°, but with the use of non-linear load and impact load growth and sensitive equipment, in addition influence on the external interference and all kinds of fault, this ideal state does not exist, the grid contains the disturbance signal, cause the signal waveform distortion. Typical power quality disturbance signal include harmonics, voltage flicker signal, voltage swell signal, voltage sag signal, voltage interruption signal, transient oscillation signal, transient pulses signal and frequency fluctuation and mixed disturbance signal, which is composed of two or more single. With the IEEE 1159-2009 standard as the basis, normalized mathematical model of power quality disturbance signal as shown in Table 1. In Table 1, s(t) is unit step function, tx, t2 are the beginning and ending time of disturbance signal disturbance, a is amplitude, ß is the frequency coefficient, c is the transient oscillation attenuation coefficient. 4. Detection and Analysis of Mixed Disturbance Signal Using LMD Algorithm 4.1. Analysis of the Mixed Disturbance Signal Detection Based on LMD Algorithm and HHT Algorithm The harmonic and voltage sag disturbance signal as an example, ... We use the LMD algorithm and HHT algorithm to detect the disturbance signal, can obtain the instantaneous amplitude and frequency curve as shown in Fig. 2. From Fig. 2 (a)-(g) and Table 2 shows that, LMD algorithm is better than HHT algorithm in the parameter fluctuation of transient characteristic parameter detection, the end effect, the detection accuracy and running time. 4.2. Detection of Mixed Disturbance Signal Based on LMD This algorithm uses the frequency of 4000 Hz. It meets the requirements of Smart Substation, namely each cycle sampling point 80. Each simulation data were analyzed from 960 sampling points. To detect six kinds of transient disturbance signal, generated by the mathematical model of disturbance signal as shown in Table 3, where T = 0.025. Mixed disturbance signal in Table 3 using LMD algorithm, waveform obtained as shown in Fig. 3 to Fig. 6. Each of these photos from top to bottom were the original transient disturbance signal waveform, amplitude function curve and frequency curve. Remove the endpoint, two extreme points of instantaneous frequency curve are the disturbance start and stop time. All sampling points of the instantaneous amplitude and frequency curve in the disturbance time are fitted by using Least Square Method, obtaining stable interval value is the detection results of disturbance amplitude and frequency. Mixed disturbance detection results based on LMD algorithm as shown in Table 4. 5. Experimental Verification Detection and identification of smart substation data collection using LMD algorithm, the results as shown in Fig. 7. We can see from the test results, the voltage signal of smart substation contains harmonics and voltage sag, voltage sags are the normalized amplitude of 0.1419 pu, start and stop time are 0.06775 s and 0.1943 s, harmonic ratio are shown in Table 5: 6. Conclusions The typical power quality mixed disturbance signal include harmonics and voltage flicker signal, harmonics and voltage swell signal, harmonics and voltage sag signal, harmonics and voltage interruption signal, as well as the actual mixed disturbance signals occurred in smart substation, are selected and analyzed by LMD algorithm. Disturbance signal is adaptively separate the disturbance signal. Simulation results show that LMD algorithm is better than HHT algorithm in the parameter fluctuation of transient characteristic parameter detection, the detection accuracy, the end effect and running time. Detection results of Smart Substation shows that, the amplitude, frequency, start and end time of disturbance signal can be accurately detected by LMD algorithm, proving the correctness of the LMD algorithm. Acknowledgment This work is supported by the Natural Science Foundation of the Education Department of Henan Province (2011 A470005,12A470005). References [1] . Xiao Xiangning, Power quality analysis and control, China Electric Power Press, Beijing, 2009, pp. 1-18. (in Chinese). [2] . Hu Ming, Chen Heng, Detection and Location of Power Quality Disturbances Using Wavelet Transform Modulus Maxima, Power System Technology, 25, 3,2001, pp. 12-16. (in Chinese). [3] . Liu Anding, Xiao Xianyong, Deng Wujun, Detection and analysis of power quality disturbances signal based on discrete cosine transform and wavelet transform, Power System Technology, 29, 10, 2005, pp. 70-74. (in Chinese). [4] . Zhao Fengzhan, Yang Rengang, Voltage sag disturbance detection based on short time fourier transform, in Proceedings of the CSEE, 27, 10, 2007, pp. 28-34. (in Chinese). [5] . Chen Xiangxun, Wavelet-based measurements and classification of short duration power quality disturbances, in Proceedings of the CSEE, 22, 10, 2002, pp. 1-6. (in Chinese). [6] . Xiao Xianyong, Xu Fangwei, Yang Honggen, Short duration disturbance classifying based on S-transform maximum similarity, International Journal of Electrical Power and Energy Systems, 31, 7,2009, pp. 374-378. [7] . Lee I. W. C., Dash P. K., S-transform-based intelligent system for classification of power quality disturbance signals, IEEE Transactions on Industrial Electronics, 50, 4, 2003, pp. 800-805. [8] . Dash P. K., Panigrahi B. K., Panda G., Power quality analysis using S-transform, IEEE Transactions on Power Delivery, 18, 2,2003, pp. 406-411. [9] . Yi Jiliang, Peng Jianchun, Classification of shorttime power quality disturbance signals based on generalized S-transform, Power System Technology, 33, 5,2009, pp. 22-26. (in Chinese). [10] . Yang Honggeng, Liu Shouliang, Xiao Xianyong, et al., S-transform-based expert system for classification of voltage dips, in Proceedings of the CSEE, 27, 1,2007, pp. 98-104. (in Chinese). [11] . Li Tianyun, Zhao Yan, Han Yongqiang, et al, Application of Hilbert-Huang transform method in detection of harmonic and voltage flicker, Power System Technology, 29, 2, 2005, pp. 73-77. (in Chinese). [12] . Li Tianyun, Zhao Yan, Li Nan, et al, A new method for power quality detection based on HHT, in Proceedings of the CSEE, 25, 17, 2005, pp. 52-56. (in Chinese). [13] . Smith J. S., The local mean decomposition and its application to EEG perception data, Journal of The Royal Society Interface, 2, 5,2005, pp. 443-454. [14] . Cheng Jun-Sheng, Yang Yu, Yu De-Jie, The local mean decomposition method and its application to gear fault diagnosis, Journal of Vibration Engineering, 22, 1,2009, pp. 76-84. (in Chinese). [15] . Cheng Junsheng, Zhang Kang, Yang Yu, Application of Local Mean Decomposition Method to the Processing of Modulated Signals, Journal of Vibration, Measurement & Diagnosis, 30, 4, 2010, pp. 362-367. (in Chinese). Cao Wensi, Xu Yan School of Electric Power, North China University of Water Resources and Electric Power, Zhengzhou, 450045, China Tel: 15890618228, fax: +86-371-65790043 E-mail: [email protected] Received: 18 September 2013 /Accepted: 22 November 2013 /Published: 30 December 2013 (c) 2013 International Frequency Sensor Association |