CN115099345A - MMC sub-module fault monitoring method based on adaptive clustering algorithm - Google Patents

Abstract

The invention discloses an MMC sub-module fault monitoring method based on a self-adaptive clustering algorithm, which relates to the technical field of multi-level power electronic converters and specifically comprises the following steps: setting sampling time, extracting actual capacitance voltage values of all sub-modules, and establishing a fault characteristic quantity data prediction model based on a cubic exponential smoothing method to obtain a fault characteristic coefficient CVVSM of the MMC system; obtaining a neighborhood radius parameter EPS and a neighborhood density threshold parameter MinPts according to a fault characteristic coefficient CVV [ SM ] by using a K nearest neighbor algorithm and a mathematical expectation method, and further calculating a neighborhood density parameter Den to form a self-adaptively determined DBSCAN clustering algorithm; the invention utilizes self-adapting to determine the DBSCAN algorithm to analyze CVVSM, if a certain CVVSM displays abnormity in three continuous sampling periods, the corresponding sub-module is a fault sub-module, and the invention improves the efficiency of fault diagnosis on the basis of reducing cost.

Description

MMC submodule fault monitoring method based on self-adaptive clustering algorithm

Technical Field

The invention belongs to the technical field of multilevel power electronic converters, and particularly relates to an MMC sub-module fault monitoring method based on a self-adaptive clustering algorithm.

Background

With the continuous development of the current science and technology, the Modular Multilevel Converter (MMC) has the advantages of modularization and expandability, and shows wide research values in renewable energy source grid connection, electrified railway supply and motor driving. However, since the number of sub-modules (SMs) in an MMC is large, each SM includes a plurality of semiconductor devices, operational reliability is one of the core problems of the MMC in medium-voltage high-power system applications.

SM faults include Short Circuit Faults (SCF) and Open Circuit Faults (OCF), both of which distort voltage and current and disrupt stable operation of the MMC system, the latter being more difficult to detect, and thus a fast and effective open circuit fault location method is of great significance to the safe operation of the MMC system.

At present, methods for detecting and positioning open-circuit faults of MMC sub-modules can be roughly divided into three types: additional sensor based methods; a mathematical model-based approach; artificial intelligence based methods; however, the method adds additional equipment or requires an accurate mathematical model, so that the labor cost and the material cost of detecting and positioning the fault sub-module are increased, and some methods need to manually set a threshold value, so that the robustness of the method is reduced; therefore, an MMC sub-module fault monitoring method based on an adaptive clustering algorithm is provided.

Disclosure of Invention

Aiming at the defects of the prior art, the invention aims to provide an MMC sub-module fault monitoring method based on a self-adaptive clustering algorithm, and solves the technical problem that the method in the prior art needs to construct a complex mathematical model, so that the labor cost and the material cost of detecting and positioning a fault sub-module are increased.

The purpose of the invention can be realized by the following technical scheme: an MMC sub-module fault monitoring method based on a self-adaptive clustering algorithm comprises the following steps:

setting sampling time, extracting actual capacitance voltage values of all sub-modules, and establishing a fault characteristic quantity data prediction model based on a cubic exponential smoothing method to obtain a fault characteristic coefficient CVVSM of the MMC system;

obtaining a neighborhood radius parameter EPS and a neighborhood density threshold parameter MinPts according to a fault characteristic coefficient CVV [ SM ] by using a K nearest neighbor algorithm and a mathematical expectation method, and further calculating a neighborhood density parameter Den to form a self-adaptively determined DBSCAN algorithm;

and analyzing the CVV [ SM ] by using a self-adaptively determined DBSCAN algorithm, and if one CVV [ SM ] displays abnormity in three continuous sampling periods, taking the corresponding sub-module as a fault sub-module.

Further, the obtaining process of the MMC system fault characteristic coefficient CVV [ SM ] comprises the following steps:

obtaining a single exponential prediction value u of the capacitor voltage in the kth sampling node based on the periodic variation trend of the capacitor voltage and a cubic exponential smoothing method _cp ⁽¹⁾ (k) And a predicted value u of the quadratic index _cp ⁽²⁾ (k) Third order exponential predicted value u _cp ⁽³⁾ (k)：

In the formula u _cp ⁽¹⁾ (k-1) is a single exponential predictor in the k-1 th sampling node, u _c (k-1) is the actual value in the k-1 th sampling node, u _cp ⁽²⁾ (k-1) is a quadratic exponential predictor in the k-1 th sampling node, u _cp ⁽³⁾ (k-1) is a cubic exponential predictor in the (k-1) th sampling node, and α is a smoothing factor coefficient;

obtaining the capacitor voltage u in the (k + T) th sampling node based on a cubic exponential smoothing method _cp Predicted value of (k + T):

u _cp (k+T)＝A _k +B _k ·T+C _k ·T ²

wherein

A _k ＝3u _cp ⁽¹⁾ (k)-3u _cp ⁽²⁾ (k)+u _cp ⁽³⁾ (k)

Further obtaining the capacitor voltage u in the k +1 th sampling node _cp Predicted value of (k + 1):

u _cp (k+1)＝β ⁽¹⁾ ·u _cp ⁽¹⁾ (k)+β ⁽²⁾ ·u _cp ⁽²⁾ (k)+β ⁽³⁾ ·u _cp ⁽³⁾ (k)

wherein

In the formula, beta ⁽¹⁾ Is a single exponential prediction coefficient, beta ⁽²⁾ Is a quadratic exponential prediction coefficient, beta ⁽³⁾ Predicting coefficients for cubic exponentials;

setting an initial value u _cp ⁽¹⁾ (1),u _cp ⁽²⁾ (1) And u _cp ⁽³⁾ (1)：

Obtaining a predicted value u based on a genetic algorithm and the periodic fluctuation stationarity of the capacitor voltage value during the normal operation of the MMC system _cp (k) And the actual value u _c (k) Mean square minimum of error (MSE) (alpha) _k ) Corresponding alpha _k ，：

In the formula, N _T The number of sampling periods participating in iteration;

and further obtaining a fault characteristic coefficient CVV [ SM ] of the ith sub-module of the MMC system:

CVV[SM _i ]＝u _ci (k)-u _cpi (k)。

further, the EPS acquiring process includes:

calculating a distance distribution matrix of CVV [ SM ] values of the N submodules:

D _N×N ＝{Dist(i,j)|1≤i≤N,1≤j≤N}

wherein

Dist(i,j)＝|CVV[SM _i ]-CVV[SM _j ]|

In the formula, N represents the number of sub-modules in a bridge arm, Dist (i, j) represents CVV [ SM ] _i ]And CVV [ SM _j ]The Euclidean distance between;

distance distribution matrix D _N×N This may be obtained by sorting the elements in each row in ascending order:

D _N×N ＝{D ₁ ,D ₂ ,...,D _K ,...,D _N-1 ,D _N }

wherein the elements in the K-th column constitute two CVVs SM]K nearest neighbor distance column vector D of all euclidean distances between them _K ；

By mixing D _K To obtain a vector D by taking the arithmetic mean of the elements in (1) _K K-th average nearest neighbor distance D of _AVK (ii) a And obtaining a parameter list D of the neighborhood radius EPS _EPS ：

D _Eps ＝{D _AVK |1≤K≤N}。

Further, based on the obtained EPS parameter list D _EPS When K is a fixed value, and EPS ═ D _AVK While, with CVV [ SM ] _i ]Data point as center, with D _AVK Drawing a circle for the radius, data CVV [ SM ] within the circle _j ]And CVV [ SM ] _i ]The Euclidean distance between Dist (i.j) is listed in the set N _EPS (i) In, N _EPS (i) The expression of (a) is:

N _Eps (i)＝{Dist(i,j)|Dist(i,j)≤D _AVK &1≤j≤N}

P _i ＝||N _Eps (i)||,i＝1,2,...,N-1,N

wherein, P _i Representative set N _EPS (i) The number of the data in the data list is,

calculate the MinPts value when K is a fixed value:

when K is 1,2,3, …, N, a parameter list D of neighborhood density threshold parameters MinPts is generated _MinPts ：

D _MinPts ＝{MinPts(K)|1≤K≤N}。

Further, based on the obtained EPS parameter list D _EPS And MinPts parameter List D _MinPts To obtain the compound represented by EPS ═ D _AVK The average number of minpts (k) distributed in each unit area in a circle as a radius, i.e., the density threshold parameter Den:

further, a parameter list D of the density threshold parameter Den is obtained by the density threshold parameter Den _Den ：

D _Den ＝{Den(K)|1≤K≤N}。

Further, the step of analyzing CVV [ SM ] by using the adaptive DBSCAN algorithm is as follows:

selecting corresponding Den according to different K, wherein K is 1,2, …, N;

since different K's have different numbers of CVV [ SM ] s in the abnormal cluster set N, when the number of CVV [ SM ] s in the set N keeps its value stable, the maximum K value is selected as the K value accordingly, the density threshold is DenK corresponding to the K value, and the cluster set N of CVV [ SM ] and abnormal CVV [ SM ] in the core cluster set C is output, wherein the K value and DenK correspond to K;

the values in the cluster set N are judged as outliers and the outliers are detected and processed when 3 detection cycles at last, if no fault occurs, the CVV SM will be captured for fault localization at the next time period.

Further, after selecting the corresponding Den, the following operations are performed according to a clustering algorithm:

randomly selecting CVV [ SM ] _i ]Where i is 1,2, N, and is denoted by CVV [ SM ] _i ]Data points are centered on EPS ═ D _AVK Checking for the presence of CVV [ SM ] within the circle for radius]s number, denoted as Q _i ；

If Q _i ≧ MinPts (K), the CVV [ SM [) _i ]Put into the core data set C and marked as normalData;

if Q _i < MinPts (K), CVV [ SM _i ]Putting the data into an abnormal cluster set N, and recording the data as abnormal data;

randomly selecting CVV [ SM ] to repeat the operation until each CVV [ SM ] is selected;

let K be K +1 and then repeat the above operations.

The invention has the beneficial effects that:

in the using process, the capacitance voltage of each submodule required in the monitoring process is a parameter which is required to be obtained by the controller in the stable operation process of the modular multilevel converter, so that the additional hardware cost is not required to be increased, and meanwhile, the method is easy to implement in the conventional MMC system and has strong practicability; the fault diagnosis research is carried out only based on the interrelation of data characteristics without constructing an accurate mathematical model and manually setting a threshold value, the self-adaptability is strong, and the robustness of the method can be improved; the method has no need of changing the running state of the system, such as introducing circulating current, has generalizability, and is suitable for detecting and positioning the faults of the semiconductor device under any topology and any working condition.

Drawings

In order to more clearly illustrate the embodiments or technical solutions in the prior art of the present invention, the drawings used in the description of the embodiments or prior art will be briefly described below, and it is obvious for those skilled in the art that other drawings can be obtained based on these drawings without creative efforts.

FIG. 1 is a flow chart of the present invention;

FIG. 2 is a three-phase MMC and sub-module topology structure diagram of the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

As shown in fig. 1-2, an MMC sub-module fault monitoring method based on an adaptive clustering algorithm includes the following steps: setting a sampling moment, extracting actual capacitance voltage values of each sub-module, and establishing a fault characteristic quantity data prediction model based on a cubic exponential smoothing method to obtain a fault characteristic coefficient CVVSM of the MMC system; obtaining a neighborhood radius parameter EPS and a neighborhood density threshold parameter MinPts according to a fault characteristic coefficient CVV [ SM ] by using a K nearest neighbor algorithm and a mathematical expectation method, and further calculating a neighborhood density parameter Den to form a self-adaptively determined DBSCAN algorithm; and (3) analyzing the CVV [ SM ] by utilizing a self-adaptive determined DBSCAN algorithm, and if one CVV [ SM ] displays abnormity in three continuous sampling periods, taking the corresponding sub-module as a fault sub-module.

The method comprises the following specific steps:

setting sampling time, extracting actual capacitance voltage values of all sub-modules, and establishing a fault characteristic quantity data prediction model based on a cubic exponential smoothing method to obtain a fault characteristic coefficient CVVSM of the MMC system;

obtaining a neighborhood radius parameter EPS and a neighborhood density threshold parameter MinPts according to a fault characteristic coefficient CVV [ SM ] by using a K nearest neighbor algorithm and a mathematical expectation method, and further calculating a neighborhood density parameter Den to form a self-adaptively determined DBSCAN algorithm;

and (3) analyzing the CVV [ SM ] by utilizing a self-adaptive determined DBSCAN algorithm, and if one CVV [ SM ] displays abnormity in three continuous sampling periods, taking the corresponding sub-module as a fault sub-module.

The MMC topology structure is composed of six bridge arms, as shown in fig. 2, each bridge arm includes N identical Submodules (SM) and a bridge arm inductor L _s The submodules adopt a half-bridge structure, and each submodule is composed of two power switches T ₁ 、T ₂ Two diodes D ₁ 、D ₂ And a DC capacitor, the capacitor voltage balancing method comprises: obtaining the number n of the submodules needing to be put into one bridge arm according to the comparison between the reference voltage of the bridge arm and the carrier wave _on The sub-module capacitor voltages are sorted in ascending order, and when the bridge arm current is positive, the capacitors are put inLowest voltage n _on A sub-module, when the bridge arm current is negative, the input capacitor has the highest voltage n _on And a sub-module.

An MMC sub-module fault monitoring method based on a self-adaptive clustering algorithm comprises the following steps: setting a sampling moment, extracting actual capacitance voltage values of each sub-module, and establishing a fault characteristic quantity data prediction model based on a cubic exponential smoothing method to obtain a fault characteristic coefficient CVVSM of the MMC system; obtaining a neighborhood radius parameter EPS and a neighborhood density threshold parameter MinPts according to a fault characteristic coefficient CVV [ SM ] by using a K nearest neighbor algorithm and a mathematical expectation method, and further calculating a neighborhood density parameter Den to form a self-adaptively determined DBSCAN algorithm; and (3) analyzing the CVV [ SM ] by utilizing a self-adaptive determined DBSCAN algorithm, and if one CVV [ SM ] displays abnormity in three continuous sampling periods, taking the corresponding sub-module as a fault sub-module.

The method specifically comprises the following steps:

(1) obtaining a single exponential prediction value u of the capacitor voltage in the kth sampling node based on the periodic variation trend of the capacitor voltage and a cubic exponential smoothing method _cp ⁽¹⁾ (k) And a predicted value u of the quadratic index _cp ⁽²⁾ (k) Third order exponential predicted value u _cp ⁽³⁾ (k)：

In the formula u _cp ⁽¹⁾ (k-1) is a single exponential predictor in the k-1 th sampling node, u _c (k-1) is the actual value in the k-1 th sampling node, u _cp ⁽²⁾ (k-1) is a quadratic exponential predictor in the k-1 th sampling node, u _cp ⁽³⁾ (k-1) is the cubic exponential predictor in the k-1 th sampling node, and α is the smoothing factor coefficient.

Obtaining a capacitor voltage u in the (k + T) th sampling node _cp Predicted value of (k + T):

u _cp (k+T)＝A _k +B _k ·T+C _k ·T ²

wherein

A _k ＝3u _cp ⁽¹⁾ (k)-3u _cp ⁽²⁾ (k)+u _cp ⁽³⁾ (k)

Further obtaining the capacitor voltage u in the (k +1) th sampling node _cp Predicted value of (k + 1):

u _cp (k+1)＝β ⁽¹⁾ ·u _cp ⁽¹⁾ (k)+β ⁽²⁾ ·u _cp ⁽²⁾ (k)+β ⁽³⁾ ·u _cp ⁽³⁾ (k)

wherein

In the formula, beta ⁽¹⁾ Is a single exponential prediction coefficient, beta ⁽²⁾ Is a quadratic exponential prediction coefficient, beta ⁽³⁾ The coefficients are predicted for cubic exponentials.

Setting an initial value u _cp ⁽¹⁾ (1),u _cp ⁽²⁾ (1) And u _cp ⁽³⁾ (1)：

Obtaining a predicted value u based on a genetic algorithm and the periodic fluctuation stability of the capacitor voltage value during the normal operation of the MMC system _cp (k) And the actual value u _c (k) A corresponding to MSE (α k) between the minimum squared error sum _k ，：

In the formula, N _T Refers to the number of sampling cycles involved in the iteration.

Obtaining a fault characteristic coefficient CVV [ SM ] of the ith sub-module of the MMC system:

CVV[SM _i ]＝u _ci (k)-u _cpi (k)

(2) as shown in FIG. 2, according to the obtained fault characteristic coefficients CVV [ SM ], a distance distribution matrix of CVV [ SM ] values of the N sub-modules is calculated:

D _N×N ＝{Dist(i,j)|1≤i≤N,1≤j≤N}

wherein

Dist(i,j)＝|CVV[SM _i ]-CVV[SM _j ]|

In the formula, N represents the number of sub-modules in a bridge arm, Dist (i, j) represents CVV [ SM ] _i ]And CVV [ SM _j ]The euclidean distance between them.

Distance distribution matrix D _N×N This may be obtained by sorting the elements in each row in ascending order:

D _N×N ＝{D ₁ ,D ₂ ,...,D _K ,...,D _N-1 ,D _N }

wherein the elements in the K-th column constitute two CVVs SM]K nearest neighbor distance column vector D of all euclidean distances between them _K 。

By mixing D _K To obtain a vector D by arithmetically averaging the elements in (1) _K K-th average nearest neighbor distance D of _AVK . And obtaining a parameter list D of the neighborhood radius EPS _EPS ：

D _Eps ＝{D _AVK |1≤K≤N}

Based on the obtained EPS parameter list D _EPS When K is a fixed value, and EPS ═ D _AVK While, with CVV [ SM ] _i ]Data point as center, with D _AVK Drawing a circle for the radius, data CVV [ SM ] within the circle _j ]And CVV [ SM ] _i ]The Euclidean distance between Dist (i.j) is listed as set N _EPS (i) In (1). N is a radical of hydrogen _EPS (i) The specific expression is as follows:

N _Eps (i)＝{Dist(i,j)|Dist(i,j)≤D _AVK &1≤j≤N}

P _i ＝||N _Eps (i)||,i＝1,2,...,N-1,N

wherein, P _i Representative set N _EPS (i) The number of data in.

Calculate the MinPts value when K is a fixed value:

when K is 1,2,3, …, N, a parameter list D of neighborhood density threshold parameters MinPts is generated _MinPts ：

D _MinPts ＝{MinPts(K)|1≤K≤N}

Based on the obtained EPS parameter list D _EPS And MinPts parameter List D _MinPts To obtain the compound represented by EPS ═ D _AVK The average number of minpts (k) distributed in each unit area in a circle as a radius, i.e., the density threshold parameter Den:

further obtaining a parameter list D of density threshold parameters Den _Den ：

D _Den ＝{Den(K)|1≤K≤N}

(3) As shown in fig. 2, corresponding Den is selected according to different K (K ═ 1,2, …, N), and CVV is randomly selected according to a clustering algorithm SM _i ](i ═ 1,2,. cndot., N), and by CVV [ SM ] _i ]Data points are centered on EPS ═ D _AVK Checking for the presence of CVV [ SM ] in the circle for radius]s number, denoted as Q _i . If Q _i ≧ MinPts (K), the CVV [ SM ] _i ]And putting the data into a core data set C and recording the data as normal data. Otherwise, CVV [ SM ] _i ]And putting the data into an abnormal cluster N and recording the data as abnormal data. When K is a fixed value, randomly select CVV [ SM ]]Repeating the above operations until each CVV [ SM ]]Are selected.

Different K there are different numbers of CVV [ SM ] s in the abnormal cluster N. When the number of CVV [ SM ] s in set N keeps its value stable, the maximum K value is selected accordingly as the optimal K value, and the optimal density threshold is DenK corresponding to that K value. And outputting a cluster set N of the CVV [ SM ] and the abnormal CVV [ SM ] in the core cluster set C, wherein the K value and the DenK correspond to the K value.

The values in the cluster set N may be determined to be outliers, and the outliers may be detected and processed when lasting 3 sample periods. If no fault has occurred, the CVV [ SM ] will be captured for fault localization at the next time period.

In the description herein, references to the description of "one embodiment," "an example," "a specific example" or the like are intended to mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

The foregoing shows and describes the general principles, essential features, and advantages of the invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, which are described in the specification and illustrated only to illustrate the principle of the present invention, but that various changes and modifications may be made therein without departing from the spirit and scope of the present invention, which fall within the scope of the invention as claimed.

Claims (8)

1. An MMC sub-module fault monitoring method based on a self-adaptive clustering algorithm is characterized by comprising the following steps:

setting sampling time, extracting actual capacitance voltage values of all sub-modules, and establishing a fault characteristic quantity data prediction model based on a cubic exponential smoothing method to obtain a fault characteristic coefficient CVVSM of the MMC system;

obtaining a neighborhood radius parameter EPS and a neighborhood density threshold parameter MinPts according to a fault characteristic coefficient CVV [ SM ] by using a K nearest neighbor algorithm and a mathematical expectation method, and further calculating a neighborhood density parameter Den to form a self-adaptively determined DBSCAN algorithm;

and analyzing the CVVSM by using a self-adaptively determined DBSCAN algorithm, and if three continuous sampling periods of a certain CVVSM show abnormity, the corresponding sub-module is a fault sub-module.

2. The MMC sub-module fault monitoring method based on adaptive clustering algorithm of claim 1, wherein the obtaining process of the MMC system fault characteristic coefficient CVV [ SM ] comprises:

obtaining a single exponential prediction value u of the capacitor voltage in the kth sampling node based on the periodic variation trend of the capacitor voltage and a cubic exponential smoothing method _cp ⁽¹⁾ (k) And a predicted value u of the quadratic index _cp ⁽²⁾ (k) Third order exponential predicted value u _cp ⁽³⁾ (k)：

In the formula u _cp ⁽¹⁾ (k-1) is a single exponential predictor in the k-1 th sampling node, u _c (k-1) is the actual value in the k-1 th sampling node, u _cp ⁽²⁾ (k-1) is a quadratic exponential predictor in the k-1 th sampling node, u _cp ⁽³⁾ (k-1) is a cubic exponential predictor in the k-1 th sampling node, and α is a smoothing factor coefficient;

obtaining the capacitor voltage u in the (k + T) th sampling node based on a cubic exponential smoothing method _cp Predicted value of (k + T):

u _cp (k+T)＝A _k +B _k ·T+C _k ·T ²

wherein

A _k ＝3u _cp ⁽¹⁾ (k)-3u _cp ⁽²⁾ (k)+u _cp ⁽³⁾ (k)

Further obtaining the capacitor voltage u in the k +1 th sampling node _cp Predicted value of (k + 1):

u _cp (k+1)＝β ⁽¹⁾ ·u _cp ⁽¹⁾ (k)+β ⁽²⁾ ·u _cp ⁽²⁾ (k)+β ⁽³⁾ ·u _cp ⁽³⁾ (k)

wherein

In the formula, beta ⁽¹⁾ Is a single exponential prediction coefficient, beta ⁽²⁾ Is a quadratic exponential prediction coefficient, beta ⁽³⁾ Predicting coefficients for cubic exponentials;

setting an initial value u _cp ⁽¹⁾ (1),u _cp ⁽²⁾ (1) And u _cp ⁽³⁾ (1)：

Obtaining a predicted value u based on a genetic algorithm and the periodic fluctuation stability of the capacitor voltage value during the normal operation of the MMC system _cp (k) And the actual value u _c (k) Mean square minimum of error (MSE) (alpha) _k ) Corresponding alpha _k ，：

In the formula, N _T The number of sampling periods participating in iteration is referred to;

and further obtaining a fault characteristic coefficient CVV [ SM ] of the ith sub-module of the MMC system:

CVV[SM _i ]＝u _ci (k)-u _cpi (k)。

3. the MMC sub-module fault monitoring method based on adaptive clustering algorithm of claim 1, characterized in that the obtaining process of EPS comprises:

calculating a distance distribution matrix of CVV [ SM ] values of the N submodules:

D _N×N ＝{Dist(i,j)|1≤i≤N,1≤j≤N}

wherein

Dist(i,j)＝|CVV[SM _i ]-CVV[SM _j ]|

In the formula, N represents the number of sub-modules in a bridge arm, Dist (i, j) represents CVV [ SM ] _i ]And CVV [ SM _j ]The Euclidean distance between;

distance distribution matrix D _N×N This may be obtained by sorting the elements in each row in ascending order:

D _N×N ＝{D ₁ ,D ₂ ,...,D _K ,...,D _N-1 ,D _N }

wherein the elements in the K-th column constitute two CVVs SM]K nearest neighbor distance column vector D of all euclidean distances between them _K ；

By mixing D _K To obtain a vector D by arithmetically averaging the elements in (1) _K K-th average nearest neighbor distance D of _AVK (ii) a And obtaining a parameter list D of the neighborhood radius EPS _EPS ：

D _Eps ＝{D _AVK |1≤K≤N}。

4. The MMC sub-module fault monitoring method based on adaptive clustering algorithm as claimed in claim 3, wherein based on the obtained EPS parameter list D _EPS When K is a fixed value and EPS ═ D _AVK While using CVV [ SM ] _i ]Data point as center, with D _AVK Drawing a circle for the radius, data CVV [ SM ] within the circle _j ]And CVV [ SM ] _i ]The Euclidean distance between Dist (i.j) is listed in the set N _EPS (i) In, N _EPS (i) The expression of (a) is:

N _Eps (i)＝{Dist(i,j)|Dist(i,j)≤D _AVK &1≤j≤N}

P _i ＝||N _Eps (i)||,i＝1,2,...,N-1,N

wherein, P _i Representative set N _EPS (i) The number of the data in the data list,

calculate the MinPts value when K is a fixed value:

when K is 1,2,3, …, N, a parameter list D of neighborhood density threshold parameters MinPts is generated _MinPts ：

D _MinPts ＝{MinPts(K)|1≤K≤N}。

5. The MMC sub-module fault monitoring method based on adaptive clustering algorithm as claimed in claim 4, wherein based on the obtained EPS parameter list D _EPS And MinPts parameter List D _MinPts Obtained in EPS ═ D _AVK The average number of minpts (k) distributed in each unit area in the circle as the radius, i.e., the density threshold parameter Den:

6. the MMC sub-module fault monitoring method based on adaptive clustering algorithm of claim 5, characterized in that, a parameter list D of a density threshold parameter Den is further obtained through the density threshold parameter Den _Den ：

D _Den ＝{Den(K)|1≤K≤N}。

7. The MMC sub-module fault monitoring method based on adaptive clustering algorithm of claim 1, wherein the step of analyzing CVV [ SM ] using the adaptively determined DBSCAN algorithm is as follows:

selecting corresponding Den according to different K, wherein K is 1,2, …, N;

since different K's have different numbers of CVV [ SM ] s in the abnormal cluster set N, when the number of CVV [ SM ] s in the set N keeps its value stable, the maximum K value is selected as the K value accordingly, the density threshold is DenK corresponding to the K value, and the cluster set N of CVV [ SM ] and abnormal CVV [ SM ] in the core cluster set C is output, wherein the K value and DenK correspond to K;

if the values in the cluster set N are judged to be outliers and the outliers are detected and processed for 3 detection periods Δ t, if no fault occurs, the CVV SM will be captured for fault localization at the next time period.

8. The MMC sub-module fault monitoring method based on adaptive clustering algorithm of claim 7, wherein after selecting the corresponding Den, according to clustering algorithm, the following operations are performed:

randomly selecting CVV [ SM ] _i ]Where i is 1,2, N, and is denoted by CVV [ SM ] _i ]Data points are centered on EPS ═ D _AVK Checking for the presence of CVV [ SM ] in the circle for radius]s number, denoted as Q _i ；

If Q _i ≧ MinPts (K), the CVV [ SM ] _i ]Putting the data into a core data set C and recording the data as normal data;

if Q _i < MinPts (K), CVV [ SM _i ]Putting the data into an abnormal cluster set N, and recording the data as abnormal data;

randomly selecting CVV [ SM ] to repeat the operation until each CVV [ SM ] is selected;

let K be K +1 and then repeat the above operations.

Images