CN106646043A - Ferromagnetic resonance online monitoring system and ferromagnetic resonance classification recognition method for power distribution network - Google Patents

Abstract

The invention discloses a ferromagnetic resonance online monitoring system for a power distribution network, and the system comprises a system initialization module, a database module, a data communication module, a clustering recognition module, a knowledge base module, and a main control unit. The clustering recognition module carries out the clustering of a certain number of training samples comprising different types of ferromagnetic resonance through calculating the clustering analysis indexes, obtains the ferromagnetic resonance classification indexes, and stores the ferromagnetic resonance classification indexes into the knowledge base module; and then the clustering recognition module calculates the clustering analysis index of an actual sample, and compares the clustering analysis index of the actual sample with the ferromagnetic resonance classification indexes, thereby recognizing the type of ferromagnetic resonance. The invention also discloses a ferromagnetic resonance classification recognition method for the power distribution network. The system carries out the clustering analysis index calculation of the certain number of training samples, carries out the clustering to obtain the ferromagnetic resonance classification indexes, compares the clustering analysis index of the actual sample with the ferromagnetic resonance classification indexes during a fault to recognize the type of ferromagnetic resonance of the actual sample, brings convenience to a power maintainer to timely take corresponding measures for inhibiting the resonance, and prevents an accident from being further expanded.

Description

Power distribution network ferromagnetic resonance on-line monitoring system and ferromagnetic resonance classifying identification method

Technical field

The present invention relates to power distribution network ferromagnetic resonance technology of identification field, and in particular to a kind of power distribution network ferromagnetic resonance is supervised online Examining system, and ferromagnetic resonance classifying identification method.

Background technology

There is many inductance and capacity cell, such as transformer, transformer, reactor, line conductor in power system Can be used as inductance element；Line conductor over the ground can be used as capacity cell with capacitive coupling etc..Operated or sent out in system During raw failure, these inductance and capacity cell, it is possible to create a variety of oscillation circuits in any case, produce resonance Phenomenon, causes resonance overvoltage.Non-linear exciter inductance in voltage transformer can produce saturated phenomenon because band is cored, make electricity Sense parameter is no longer constant, but is changed with the change of electric current or magnetic flux.This circuit containing nonlinear inductance element, When certain condition is met, ferromagnetic resonance can be caused, produce very big overvoltage and overcurrent, seriously threaten power system Safe and stable operation.

At present for ferromagnetic resonance is all to take unified braking measure, but change pole of the ferromagnetic resonance to systematic parameter For sensitivity, frequency dividing, fundamental frequency, high frequency three types can be divided into, certain braking measure is effective, but right to a kind of resonance of frequency The resonance of another kind of frequency may be like water off a duck's back, therefore to ferromagnetic resonance does not carry out Classification and Identification, many times can not be fine Over-voltage suppression, accident can further expand.

The content of the invention

In order to overcome the deficiencies in the prior art, the invention discloses a kind of power distribution network ferromagnetic resonance on-line monitoring system System and ferromagnetic resonance classifying identification method.

For achieving the above object, the technical solution used in the present invention is：

Power distribution network ferromagnetic resonance on-line monitoring system, including system initialization module, DBM, data communication mould Block, clustering recognition module, base module and main control unit；

The system initialization module, when whole system puts into operation, is carried out to system structure parameter and operational factor Initialization, and the initialization information of each module is sent to into main control unit；

The DBM, including line mutual-ground capacitor database, PT magnetizing inductance databases, line-to-ground capacitor value Database and PT excitation reactance Value Datas storehouse；

The data communication module, connects the PMU PMUs in WAMS and is led to by Ethernet Letter, receives message and is parsed, and logging data is simultaneously stored；

The clustering recognition module, by calculating cluster analysis index, to a number of comprising different ferromagnetic resonance classes The training sample of type is clustered, and is obtained ferromagnetic resonance classification indicators and is stored in base module；The poly- of actual sample is calculated again Alanysis index, with ferromagnetic resonance classification indicators contrast identification ferromagnetic resonance type is carried out.

The further scheme of the present invention is that the DBM is realized using SQL Serve.

The further scheme of the present invention is that the communication of the data communication module meets IEEE C37.94 agreements.

The further scheme of the present invention is that the main control unit is ZigBee central coordinators.

The further scheme of the present invention is, also including detection alarm module, when the cluster analysis index of actual sample falls When entering in ferromagnetic resonance classification indicators, alarm module start triggering circuit is detected, trigger is sent to into main control unit and is started Warning device, points out generation ferromagnetic resonance and shows ferromagnetic resonance type.

The further scheme of the present invention is that the calculation procedure of the clustering recognition module is as follows：

Step 1：By formula (1) ratio calculated parameter γ, formula (1) is：γ=X_c0/X_m, wherein X_c0For distribution network capacity Anti- value, X_mFor PT excitation reactance values；

Step 2：Each harmonic voltage magnitude when obtaining failure by Fourier decomposition, integral frequency harmonizing wave is designated as U_1m, U_2m,...,U_nm, corresponding harmonic frequency is designated as f₁,f₂,...,f_n；M-Acetyl chlorophosphonazo is designated asCorresponding harmonic frequency is designated as f_1/2,f_1/3,...,f_1/n；

Step 3：Frequency, amplitude according to integral frequency harmonizing wave and m-Acetyl chlorophosphonazo, by formula (2) resonant frequency factor-alpha is calculated, Formula (2) is：

Step 4：Using ratio parameter γ and resonant frequency factor-alpha as cluster analysis index；

Step 5：Using K mean cluster algorithm, a number of training sample comprising different ferromagnetic resonance types is entered Row cluster；

Step 6：Cluster result is compared with the ferromagnetic resonance type of training sample, the correctness of classification is verified, together When reject misclassification training sample, determine maximum distance d of the training sample to its cluster centre in i class ferromagnetic resonances_i；

Step 7：The cluster analysis index of actual sample is calculated, the actual sample is calculated to i further according to cluster analysis index Class ferromagnetic resonance cluster centre apart from L_i, when meeting condition：L_i＜ d_iWhen, determine that actual sample ferromagnetic resonance type is i classes.

The further scheme of the present invention is that the training sample includes frequency dividing, fundamental frequency, three kinds of ferromagnetic resonance classes of high frequency Type.

Power distribution network ferromagnetic resonance classifying identification method, comprises the following steps：

Step 1：By formula (1) ratio calculated parameter γ, formula (1) is：γ=X_c0/X_m, wherein X_c0For distribution network capacity Anti- value, X_mFor PT excitation reactance values；

Step 2：Each harmonic voltage magnitude when obtaining failure by Fourier decomposition, integral frequency harmonizing wave is designated as U_1m, U_2m,...,U_nm, corresponding harmonic frequency is designated as f₁,f₂,...,f_n；M-Acetyl chlorophosphonazo is designated asCorresponding harmonic frequency is designated as f_1/2,f_1/3,...,f_1/n；

Step 3：Frequency, amplitude according to integral frequency harmonizing wave and m-Acetyl chlorophosphonazo, by formula (2) resonant frequency factor-alpha is calculated, Formula (2) is：

Step 4：Using ratio parameter γ and resonant frequency factor-alpha as cluster analysis index；

Step 5：Using K mean cluster algorithm, a number of training sample comprising different ferromagnetic resonance types is entered Row cluster；

Step 6：Cluster result is compared with the ferromagnetic resonance type of training sample, the correctness of classification is verified, together When reject misclassification training sample, determine maximum distance d of the training sample to its cluster centre in i class ferromagnetic resonances_i；

Step 7：The cluster analysis index of actual sample is calculated, the actual sample is calculated to i further according to cluster analysis index Class ferromagnetic resonance cluster centre apart from L_i, when meeting condition：L_i＜ d_iWhen, determine that actual sample ferromagnetic resonance type is i classes.

The further scheme of the present invention is that the training sample includes frequency dividing, fundamental frequency, three kinds of ferromagnetic resonance types of high frequency.

Present invention advantage compared with prior art is：

By carrying out after cluster analysis index calculating to a number of training sample comprising different ferromagnetic resonance types Cluster, obtains ferromagnetic resonance classification indicators, then the cluster analysis index of actual sample when breaking down and ferromagnetic resonance are classified Index carries out contrasting the ferromagnetic resonance type of identification actual sample, is easy to electric power overhaul personnel to take corresponding suppression resonance in time Measure, Accident prevention further expands.

Description of the drawings

Fig. 1 is the functional block diagram of power distribution network ferromagnetic resonance on-line monitoring system of the present invention.

Fig. 2 is the power distribution network ferromagnetic resonance classifying identification method implementing procedure figure of the present invention.

Fig. 3 is the training sample K mean cluster analysis chart of the present invention.

Specific embodiment

Certain 10kV power distribution network ferromagnetic resonance on-line monitoring system as shown in Figure 1, including system initialization module, database Module, data communication module, clustering recognition module, base module, detection alarm module and ZigBee central coordinators.

The system initialization module, when whole system puts into operation, is carried out to system structure parameter and operational factor Initialization, and the initialization information of each module is sent to by association of ZigBee central authorities by the distinctive wireless senser of ZigBee technology Adjust device.

The DBM realizes possessing and data communication module communication function using SQL Serve, including circuit pair Ground capacitance data storehouse, PT magnetizing inductance databases, line-to-ground capacitor value database and PT excitation reactance Value Datas storehouse, it is main real The foundation of existing PT models and the foundation of 10kV isolated neutral system models.

The communication of the data communication module meets IEEE C37.94 agreements, including tcp data communication module and data solution Analysis module, it is main to realize and the communication between ZigBee central coordinators, and realize that data are parsed and electric network data writing function； Connect the PMU PMUs in WAMS by Ethernet to be communicated, receive TCP message, and parse its base This information, then by data inputting and is stored in tables of data.

The clustering recognition module, by calculating cluster analysis index, to a number of comprising different ferromagnetic resonance classes The training sample of type is clustered, and is obtained ferromagnetic resonance classification indicators and is stored in base module；The poly- of actual sample is calculated again Alanysis index, with ferromagnetic resonance classification indicators contrast identification ferromagnetic resonance type is carried out.

The calculation procedure of the clustering recognition module is as follows：

Step 1：By formula (1) ratio calculated parameter γ, formula (1) is：γ=X_c0/X_m, wherein X_c0For distribution network capacity Anti- value, X_mFor PT excitation reactance values；

Step 2：Each harmonic voltage magnitude when obtaining failure by Fourier decomposition, integral frequency harmonizing wave is designated as U_1m, U_2m,...,U_nm, corresponding harmonic frequency is designated as f₁,f₂,...,f_n；M-Acetyl chlorophosphonazo is designated asCorresponding harmonic frequency is designated as f_1/2,f_1/3,...,f_1/n；

Step 3：Frequency, amplitude according to integral frequency harmonizing wave and m-Acetyl chlorophosphonazo, by formula (2) resonant frequency factor-alpha is calculated, Formula (2) is：

Step 4：Using ratio parameter γ and resonant frequency factor-alpha as cluster analysis index；

Step 5：Using K mean cluster algorithm, to a number of comprising frequency dividing, fundamental frequency, three kinds of ferromagnetic resonance classes of high frequency The training sample of type is clustered；The training sample of each ferromagnetic resonance type can be by changing ground capacity C₀Parameter is simulating Obtain, such as：

1. high-frequency resonant

Make ground capacity parameter C₀=5 × 10^-5F, correspondence PT magnetizing inductance L=0.12H, can be by being calculated：X_c0= 63.66 Ω, X_m=35.37 Ω, and then be calculated：γ=1.8, α=3.39；

2. fundamental resonance

Make ground capacity parameter C₀=1 × 10^-3F, correspondence PT magnetizing inductances L=8.2 × 10^-3H, can be by being calculated： X_c0=0.96 Ω, X_m=2.58 Ω, and then be calculated：γ=0.37, α=1.07；

3. Subharmonic Resonance

Make ground capacity parameter C₀=0.25F, correspondence PT magnetizing inductances L=7.39 × 10^-4H, can be by being calculated：X_c0 =0.013 Ω, X_m=0.232 Ω, and then be calculated：γ=0.056, α=0.72；

Training sample K mean cluster analysis result is as shown in figure 3, intercepted the cluster analysis of part training sample in table 1 Achievement data：

Table 1

In addition to discrete error data, the region of Subharmonic Resonance be with (0.0303, be 0.7441) center of circle, radius is less than In 0.4464 circle；The region of fundamental resonance be with (0.2062, be 1.0787) center of circle, radius less than 0.1393 circle in； The region of high-frequency resonant be with (1.2453, be 3.5313) center of circle, radius less than 0.7253 circle in；Other regions do not occur Ferromagnetic resonance.

Step 6：Cluster result is compared with the ferromagnetic resonance type of training sample, the correctness of classification is verified, together When reject misclassification training sample, determine maximum distance d of the training sample to its cluster centre in i class ferromagnetic resonances_i；

Step 7：The cluster analysis index of actual sample is calculated, the actual sample is calculated to i further according to cluster analysis index Class ferromagnetic resonance cluster centre apart from L_i, when meeting condition：L_i＜ d_iWhen, determine that actual sample ferromagnetic resonance type is i classes.

When the cluster analysis index of actual sample is fallen in ferromagnetic resonance classification indicators, detection alarm module starts triggering Circuit, is sent to trigger ZigBee central coordinators and starts warning device, lights alarm window, starts buzzer, points out Generation ferromagnetic resonance simultaneously shows ferromagnetic resonance type.

Claims (9)

1. power distribution network ferromagnetic resonance on-line monitoring system, it is characterised in that：Including system initialization module, DBM, number According to communication module, clustering recognition module, base module and main control unit；

The system initialization module, when whole system puts into operation, is carried out initially to system structure parameter and operational factor Change, and the initialization information of each module is sent to into main control unit；

The DBM, including line mutual-ground capacitor database, PT magnetizing inductance databases, line-to-ground capacitive reactance Value Data Storehouse and PT excitation reactance Value Datas storehouse；

The data communication module, connects the PMU PMUs in WAMS and is communicated by Ethernet, connects Parsed by message, logging data is simultaneously stored；

The clustering recognition module, by calculating cluster analysis index, to a number of comprising different ferromagnetic resonance types Training sample is clustered, and is obtained ferromagnetic resonance classification indicators and is stored in base module；The cluster point of actual sample is calculated again Analysis index, with ferromagnetic resonance classification indicators contrast identification ferromagnetic resonance type is carried out.

2. power distribution network ferromagnetic resonance on-line monitoring system according to claim 1, it is characterised in that：The DBM Realized using SQL Serve.

3. power distribution network ferromagnetic resonance on-line monitoring system according to claim 1, it is characterised in that：The data communication mould The communication of block meets IEEE C37.94 agreements.

4. power distribution network ferromagnetic resonance on-line monitoring system according to claim 1, it is characterised in that：The main control unit is ZigBee central coordinators.

5. power distribution network ferromagnetic resonance on-line monitoring system according to claim 1, it is characterised in that：Also include that detection is reported to the police Module, when the cluster analysis index of actual sample is fallen in ferromagnetic resonance classification indicators, detection alarm module starts triggering electricity Road, is sent to trigger main control unit and starts warning device, points out generation ferromagnetic resonance and shows ferromagnetic resonance type.

6. power distribution network ferromagnetic resonance on-line monitoring system according to claim 1, it is characterised in that：The clustering recognition mould The calculation procedure of block is as follows：

Step 1：By formula (1) ratio calculated parameter γ, formula (1) is：γ=X_c0/X_m, wherein X_c0For power distribution network capacitor value, X_mFor PT excitation reactance values；

Step 2：Each harmonic voltage magnitude when obtaining failure by Fourier decomposition, integral frequency harmonizing wave is designated as U_1m,U_2m,..., U_nm, corresponding harmonic frequency is designated as f₁,f₂,...,f_n；M-Acetyl chlorophosphonazo is designated asCorresponding harmonic frequency is designated as f_1/2, f_1/3,...,f_1/n；

Step 3：Frequency, amplitude according to integral frequency harmonizing wave and m-Acetyl chlorophosphonazo, by formula (2) resonant frequency factor-alpha, formula are calculated (2) it is：

$\mathrm{\α}=\underset{\frac{1}{n}}{\overset{n}{\Σ}}(\frac{U_{im}}{\sqrt{U_{\frac{1}{n}m}^2+\mathrm{...}+U_{\frac{1}{2}m}^2+U_{1m}^2+\mathrm{...}+U_{nm}^2}}\×\frac{f_i}{f_1});$

Step 4：Using ratio parameter γ and resonant frequency factor-alpha as cluster analysis index；

Step 5：Using K mean cluster algorithm, a number of training sample comprising different ferromagnetic resonance types is gathered Class；

Step 6：Cluster result is compared with the ferromagnetic resonance type of training sample, the correctness of classification is verified, while picking Except the training sample of misclassification, maximum distance d of the training sample to its cluster centre in i class ferromagnetic resonances is determined_i；

Step 7：The cluster analysis index of actual sample is calculated, the actual sample is calculated to i class iron further according to cluster analysis index Magnetic resonance cluster centre apart from L_i, when meeting condition：L_i＜ d_iWhen, determine that actual sample ferromagnetic resonance type is i classes.

7. the power distribution network ferromagnetic resonance on-line monitoring system according to claim 1 or 6, it is characterised in that：The training sample This includes frequency dividing, fundamental frequency, three kinds of ferromagnetic resonance types of high frequency.

8. power distribution network ferromagnetic resonance classifying identification method, it is characterised in that comprise the following steps：

Step 1：By formula (1) ratio calculated parameter γ, formula (1) is：γ=X_c0/X_m, wherein X_c0For power distribution network capacitor value, X_mFor PT excitation reactance values；

Step 2：Each harmonic voltage magnitude when obtaining failure by Fourier decomposition, integral frequency harmonizing wave is designated as U_1m,U_2m,..., U_nm, corresponding harmonic frequency is designated as f₁,f₂,...,f_n；M-Acetyl chlorophosphonazo is designated asCorresponding harmonic frequency is designated as f_1/2, f_1/3,...,f_1/n；

Step 3：Frequency, amplitude according to integral frequency harmonizing wave and m-Acetyl chlorophosphonazo, by formula (2) resonant frequency factor-alpha, formula are calculated (2) it is：

$\mathrm{\α}=\underset{\frac{1}{n}}{\overset{n}{\Σ}}(\frac{U_{im}}{\sqrt{U_{\frac{1}{n}m}^2+\mathrm{...}+U_{\frac{1}{2}m}^2+U_{1m}^2+\mathrm{...}+U_{nm}^2}}\×\frac{f_i}{f_1});$

Step 4：Using ratio parameter γ and resonant frequency factor-alpha as cluster analysis index；

Step 5：Using K mean cluster algorithm, a number of training sample comprising different ferromagnetic resonance types is gathered Class；

Step 6：Cluster result is compared with the ferromagnetic resonance type of training sample, the correctness of classification is verified, while picking Except the training sample of misclassification, maximum distance d of the training sample to its cluster centre in i class ferromagnetic resonances is determined_i；

Step 7：The cluster analysis index of actual sample is calculated, the actual sample is calculated to i class iron further according to cluster analysis index Magnetic resonance cluster centre apart from L_i, when meeting condition：L_i＜ d_iWhen, determine that actual sample ferromagnetic resonance type is i classes.

9. power distribution network ferromagnetic resonance classifying identification method according to claim 8, it is characterised in that：The training sample bag Include frequency dividing, fundamental frequency, three kinds of ferromagnetic resonance types of high frequency.