CN105842535A - Harmonic wave main characteristic group screening method based on similar characteristic fusion - Google Patents

Abstract

The invention relates to a harmonic wave main characteristic group screening method based on similar characteristic fusion, comprising steps of performing decentralization on a monitoring sample of harmonic wave current, obtaining a decentralized numerical value, calculating a similarity index of the monitoring sample of the harmonic wave current, and screening out a main characteristic group on the basis of the similarity index distribution characteristics. The harmonic wave main characteristic group screening method can fast determine the main harmonic wave pollution times of a monitoring point.

Description

A kind of harmonic wave main syndrome screening technique merged based on similar features

Technical field

The present invention relates to Harmonious Waves in Power Systems and pollute field, a kind of main feature of harmonic wave merged based on similar features Group's screening technique.

Background technology

Along with being incorporated into the power networks and the increase of other nonlinear-load quantity, in power system of a large amount of power electronic equipments Harmonic pollution is increasingly severe.At present, the most complete quality of power supply on-line monitoring network has been had been built up, it is possible to monitoring electrical network electricity Pressure total harmonic distortion factor, each harmonic voltage containing ratio and phase angle, current total harmonic distortion rate, individual harmonic current containing ratio, The harmonic information such as virtual value and phase angle.Substantial amounts of on-line monitoring information contributes to harmonic pollution user modeling, obtains user's fortune The master data of row.But complete harmonic-model comprises whole harmonic voltage, harmonic current monitoring index, and mutual shadow between index Ring so that model is extremely complex and cannot realize parameter identification.Which in engineering practice, need to pick out Detecting Power Harmonics index Should comprise in a model, which variable should be rejected from model, namely needs to determine from a large amount of Historical Monitoring data The main syndrome of harmonic wave, in order to set up Practical model for the main syndrome of harmonic wave.

The most generally monitor user's voltage and current at points of common connection (Point of Common Coupling, PCC), Calculate the harmonic components of voltage and current, use the value greatly of maximum, meansigma methods or 95% in the detection time period to evaluate user couple The impact of the electrical network quality of power supply, and the foundation of harmonic wave control is carried out in this, as user.This way uses harmonic wave the most exactly Current source model characterizes user and pollutes, and using Monitoring Data as user model parameter, and do not account for different overtone order it Between influence each other, be not to user's harmonic pollution characteristic essence reflection.

Owing to harmonic source produces the principle complexity of harmonic wave, it tends to be difficult to set up general mathematical model.Harmonic source can at present To use the model such as equivalent source, crossover frequency admittance matrix, application independent component analysis, least square approximation and neutral net Etc. method from monitor sample data identification model parameter.Wherein, Harmonic source model based on crossover frequency admittance matrix considers The harmonic voltage impact on harmonic current, but model parameter to be recalculated under different operating modes.Based on least square approximation Harmonic current is expressed as first-harmonic, each harmonic component of voltage and is not divided by the current constant of voltage variations affect by Harmonic source model The expression formula of amount, utilizes least square approximation to ask for model parameter, and degree of accuracy is higher, but there is model parameter and ask for difficulty etc. Problem.Harmonic Source Modeling based on neutral net need not understand the internal structure of harmonic-producing load, but model accuracy is trained Sample number restricts.

If operating mode to be analyzed is close with the sample operating mode of parameter identification, then calculate error mainly by parameter identification precision Determine, and the most unrelated with the types of models selected.When operating mode to be analyzed differs greatly with sample operating mode, different harmonic sources Model is bigger on calculating error impact.

The thinking of Harmonic Source Modeling is to simplify model structure at present, thus error is bigger.If able to from history Monitoring Data determines harmonic wave main feature number of times, sets up detailed model for main feature number of times, it is possible in reserving model precision While a large amount of reduce parameter identification difficulty.

Summary of the invention

In view of this, the purpose of the present invention is to propose to a kind of harmonic wave main syndrome screening side merged based on similar features Method, it is possible to quickly determine that the major harmonic of monitoring point pollutes number of times.

The present invention uses below scheme to realize: a kind of harmonic wave main syndrome screening technique merged based on similar features, tool Body comprises the following steps:

Step S1: deduct individual harmonic current benchmark limit value C from harmonic current class monitor sample₀, harmonic current class is supervised This decentration of test sample, obtains the numerical value C of decentration_*；

Step S2: calculate the index of similarity of harmonic current class monitor sample: remember S_24*24For C_*Between middle each harmonic Index of similarity matrix, S (i, j) represents the similarity between i & lt harmonic current and jth subharmonic current monitor sample, its In, 1≤i, j≤25, be calculated as follows S (i, j), result is between 1 and-1:

$S(i,j)=\frac{\underset{k=1,m}{\Σ}{C}_{*}(k,i)\·C_{*}(k,j)}{\sqrt{\underset{k=1,m}{\Σ}{C}_{*}^2(k,i)}\·\sqrt{\underset{k=1,m}{\Σ}{C}_{*}^2(k,j)}};$

Step S3: screen main syndrome based on index of similarity distribution character.

Further, described step S1 specifically includes following steps:

Step S11: set measuring point harmonic current measurement value matrix as C_m*25, wherein C_m*25For m row, the matrix of 25 row；Wherein J row represent j subharmonic, 1≤j≤25, and the i-th row represents ith measurement value, 1≤i≤m；

Step S12: according to measuring point electric pressure, it is stipulated that individual harmonic current benchmark limit value is C₀, C₀Be 1*25 row to Amount, unit is A；

Step S13: to C_m*25Every a line, deduct harmonic current benchmark limit value C₀, obtain the numerical value C of decentration_*, C_* For m row, the matrix of 25 row, unit is A.Wherein, C₀Specify according to GB GB/T 14549-93.Such as, following table is injected exactly The allowable harmonic current of 10kV points of common connection.

Further, described step S3 specifically includes following steps:

Step S31: [-1,1] is divided into 10 minizones, according to S (i, j) (1 ＜ i≤25, i ＜ j≤25) value general It adheres to each interval separately, obtains 10 initial population；

Step S32: calculate the center c of each initial population_k, described c_kIt is (i, meansigma methods j) of S in each group；

Step S33: calculate S (i, j) (1 ＜ i≤25, i ＜ j≤25) and each group center distance l (i, j)=| S (i, j)-c_k|；

Step S34: (i j) belongs to that group closest with it by S；

Step S35: removing members is empty group, obtains new group and member thereof；

Step S36: return to step S32 and restart to hive off, until the grouping result between twice iteration no longer changes, or Till person's iterations is more than 100；

Step S37: (i, j) subscript, the most often group subscript represents a main syndrome of harmonic wave to list the S in each group.

Compared with prior art, the present invention has a following beneficial effect:

1, invention defines the index of similarity of harmonic current class monitor sample；Propose a kind of based on harmonic current class prison Survey the main syndrome screening technique of index similarity, it is possible to quickly determine that the major harmonic of monitoring point pollutes number of times.

2, the present invention uses a large amount of online monitoring data to carry out statistical computation, obtains monitoring point harmonic pollution feature number of times Statistical information, and it is not only the result of calculation under certain exceptional operating conditions, its conclusion is the most reasonable.

Accompanying drawing explanation

Fig. 1 is the method flow schematic diagram of the present invention.

Detailed description of the invention

Below in conjunction with the accompanying drawings and embodiment the present invention will be further described.

As it is shown in figure 1, present embodiments provide a kind of harmonic wave main syndrome screening technique merged based on similar features, tool Body comprises the following steps:

Step S1: deduct individual harmonic current benchmark limit value C from harmonic current class monitor sample₀, harmonic current class is supervised This decentration of test sample, obtains the numerical value C of decentration_*；

Step S2: calculate the index of similarity of harmonic current class monitor sample: remember S_24*24For C_*Between middle each harmonic Index of similarity matrix, S (i, j) represents the similarity between i & lt harmonic current and jth subharmonic current monitor sample, its In, 1≤i, j≤25, be calculated as follows S (i, j), result is between 1 and-1:

$S(i,j)=\frac{\underset{k=1,m}{\Σ}{C}_{*}(k,i)\·C_{*}(k,j)}{\sqrt{\underset{k=1,m}{\Σ}{C}_{*}^2(k,i)}\·\sqrt{\underset{k=1,m}{\Σ}{C}_{*}^2(k,j)}};$

Step S3: screen main syndrome based on index of similarity distribution character.

In the present embodiment, described step S1 specifically includes following steps:

Step S11: set measuring point harmonic current measurement value matrix as C_m*25, wherein C_m*25For m row, the matrix of 25 row；Wherein J row represent j subharmonic, 1≤j≤25, and the i-th row represents ith measurement value, 1≤i≤m；

Step S12: according to measuring point electric pressure, it is stipulated that individual harmonic current benchmark limit value is C₀, C₀Be 1*25 row to Amount, unit is A；

Step S13: to C_m*25Every a line, deduct harmonic current benchmark limit value C₀, obtain the numerical value C of decentration_*, C_* For m row, the matrix of 25 row, unit is A.Wherein, C₀Specify according to GB GB/T 14549-93.Such as, following table is injected exactly The allowable harmonic current of 10kV points of common connection.

Overtone order (h)	2	3	4	5	6	7	8	9
									Allowable harmonic current (A)	26.0	20.0	13.0	20.0	8.5	15.0	6.4	6.8
Overtone order (h)	10	11	12	13	14	15	16	17
									Allowable harmonic current (A)	5.1	9.3	4.3	7.9	3.7	4.1	3.2	6.0
Overtone order (h)	18	19	20	21	22	23	24	25
									Allowable harmonic current (A)	2.8	5.4	2.6	2.9	2.3	4.5	2.1	4.1

In the present embodiment, described step S3 specifically includes following steps:

Step S31: [-1,1] is divided into 10 minizones, according to S (i, j) (1 ＜ i≤25, i ＜ j≤25) value general It adheres to each interval separately, obtains 10 initial population；

Step S32: calculate the center c of each initial population_k, described c_kIt is (i, meansigma methods j) of S in each group；

Step S33: calculate S (i, j) (1 ＜ i≤25, i ＜ j≤25) and each group center distance l (i, j)=| S (i, j)-c_k|；

Step S34: (i j) belongs to that group closest with it by S；

Step S35: removing members is empty group, obtains new group and member thereof；

Step S36: return to step S32 and restart to hive off, until the grouping result between twice iteration no longer changes, or Till person's iterations is more than 100；

Step S37: (i, j) subscript, the most often group subscript represents a main syndrome of harmonic wave to list the S in each group.

The foregoing is only presently preferred embodiments of the present invention, all impartial changes done according to scope of the present invention patent with Modify, all should belong to the covering scope of the present invention.

Claims (3)

1. the harmonic wave main syndrome screening technique merged based on similar features, it is characterised in that comprise the following steps:

Step S1: deduct individual harmonic current benchmark limit value C from harmonic current class monitor sample₀, to harmonic current class monitor sample Decentration, obtains the numerical value C of decentration_*；

Step S2: calculate the index of similarity of harmonic current class monitor sample: remember S_24*24For C_*Similar between middle each harmonic Degree index matrix, (i j) represents the similarity between i & lt harmonic current and jth subharmonic current monitor sample, wherein, 1 to S ≤ i, j≤25, be calculated as follows S (i, j), result is between 1 and-1:

$S(i,j)=\frac{\underset{k=1,m}{\Σ}{C}_{*}(k,i)\·C_{*}(k,j)}{\sqrt{\underset{k=1,m}{\Σ}{C}_{*}^2(k,i)}\·\sqrt{\underset{k=1,m}{\Σ}{C}_{*}^2(k,j)}};$

Step S3: screen main syndrome based on index of similarity distribution character.

A kind of harmonic wave main syndrome screening technique merged based on similar features the most according to claim 1, its feature exists Following steps are specifically included in: described step S1:

Step S11: set measuring point harmonic current measurement value matrix as C_m*25, wherein C_m*25For m row, the matrix of 25 row；Wherein jth row Representing j subharmonic, 1≤j≤25, the i-th row represents ith measurement value, 1≤i≤m；

Step S12: according to measuring point electric pressure, it is stipulated that individual harmonic current benchmark limit value is C₀, C₀It is the row vector of 1*25, single Position is A；

Step S13: to C_m*25Every a line, deduct harmonic current benchmark limit value C₀, obtain the numerical value C of decentration_*, C_*For m row, The matrix of 25 row, unit is A.

A kind of harmonic wave main syndrome screening technique merged based on similar features the most according to claim 1, its feature exists Following steps are specifically included in: described step S3:

Step S31: [-1,1] is divided into 10 minizones, according to S (i, j) (1 ＜ i≤25, i ＜ j≤25) value by its point Belong to each interval, obtain 10 initial population；

Step S32: calculate the center c of each initial population_k, described c_kIt is (i, meansigma methods j) of S in each group；

Step S33: calculate S (i, j) (1 ＜ i≤25, i ＜ j≤25) and each group center distance l (i, j)=| S (i, j)-c_k |；

Step S34: (i j) belongs to that group closest with it by S；

Step S35: removing members is empty group, obtains new group and member thereof；

Step S36: return to step S32 and restart to hive off, until the grouping result between twice iteration no longer changes, or repeatedly Till generation number is more than 100；

Step S37: (i, j) subscript, the most often group subscript represents a main syndrome of harmonic wave to list the S in each group.