CN114004512A - Multi-unit operation state outlier analysis method and system based on density clustering - Google Patents

Multi-unit operation state outlier analysis method and system based on density clustering Download PDF

Info

Publication number
CN114004512A
CN114004512A CN202111300977.1A CN202111300977A CN114004512A CN 114004512 A CN114004512 A CN 114004512A CN 202111300977 A CN202111300977 A CN 202111300977A CN 114004512 A CN114004512 A CN 114004512A
Authority
CN
China
Prior art keywords
vector
temperature
generator
phase winding
point set
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202111300977.1A
Other languages
Chinese (zh)
Inventor
高晨
童博
赵勇
程方
韩毅
宋子琛
张宝锋
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xian Thermal Power Research Institute Co Ltd
Original Assignee
Xian Thermal Power Research Institute Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Xian Thermal Power Research Institute Co Ltd filed Critical Xian Thermal Power Research Institute Co Ltd
Priority to CN202111300977.1A priority Critical patent/CN114004512A/en
Publication of CN114004512A publication Critical patent/CN114004512A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Electricity, gas or water supply

Abstract

The invention provides a density clustering-based multi-unit operation state outlier analysis method and a density clustering-based multi-unit operation state outlier analysis system, which comprise the following steps: step 1, acquiring state monitoring parameters corresponding to each wind turbine of a wind power plant at a time t, and combining to form a data set corresponding to the wind power plant, wherein the state monitoring parameters comprise motor rotating speed, torque, power, current effective value, generator U-phase winding temperature, generator V-phase winding temperature, generator W-phase winding temperature, gear box oil temperature, cooling system temperature, power electronic device temperature and cabin environment temperature; step 2, processing the acquired data by adopting a DBSCAN density clustering calculation method; the method is beneficial to early finding of the state deviation unit, meanwhile, the density clustering method is high in clustering speed, noise points can be effectively processed, spatial clusters of any shape can be found, the number of clusters to be divided does not need to be input, complicated training does not need to be carried out according to historical data, and the method has high universality.

Description

Multi-unit operation state outlier analysis method and system based on density clustering
Technical Field
The invention relates to the field of wind power generation, in particular to a density clustering-based multi-unit operation state outlier analysis method and system.
Background
An existing fan operation monitoring system continuously monitors specified parameters of a single fan through setting a threshold value, and analyzes the operation state of the unit by comparing the monitored parameters with the threshold value. However, in order to reduce the false alarm rate, the threshold setting in this method is often too large, so that the hidden trouble of the fault cannot be found early, and the fault is only identified when the fault occurs. Therefore, on the basis of original analysis, the transverse comparison of the operation states of multiple fans in the wind power plant needs to be carried out, the fans of other units in the operation states are found, and early warning before faults occur is achieved.
Disclosure of Invention
The invention aims to provide a density clustering-based multi-unit operation state outlier analysis method and a density clustering-based multi-unit operation state outlier analysis system, which solve the defects in the prior art.
In order to achieve the purpose, the invention adopts the technical scheme that:
the invention provides a density clustering-based multi-unit operation state outlier analysis method, which comprises the following steps of:
step 1, acquiring state monitoring parameters corresponding to each wind turbine of a wind power plant at a time t, and combining to form a data set corresponding to the wind power plant, wherein the state monitoring parameters comprise motor rotating speed, torque, power, current effective value, generator U-phase winding temperature, generator V-phase winding temperature, generator W-phase winding temperature, gear box oil temperature, cooling system temperature, power electronic device temperature and cabin environment temperature;
step 2, processing the temperature data vectors in the data set in the step 1 to obtain a new data set matrix with temperature difference vectors;
step 3, converting the temperature vector of each row in the new data set matrix in the step 2 and the motor rotating speed vector in the data set in the step 1 into a group of scattered points on a two-dimensional plane, and recording a coordinate set of the group of scattered points on the two-dimensional plane as a two-dimensional point set;
step 4, performing DBSCAN density clustering calculation on the two-dimensional point set obtained in the step 3 to obtain a corresponding state value vector within the set time of the temperature vector of the row;
step 5, converting the torque vector and the power vector in the data set in the step 1 and the motor rotating speed vector in the data set in the step 1 into a group of scattered points on a two-dimensional plane respectively, and recording a coordinate set of the group of scattered points on the two-dimensional plane as a two-dimensional point set;
step 6, repeating the step 4, and combining the two-dimensional point sets corresponding to the torque vectors and the power vectors obtained in the step 5 to respectively obtain corresponding state value vectors within the set time of the torque vectors and the power vectors;
and 7, identifying the outlier state of each state monitoring parameter vector according to the corresponding state value vector within the set time.
Preferably, in step 2, the temperature data vector in the data set in step 1 is processed to obtain a new data set matrix with a temperature difference vector, and the specific method is as follows:
respectively calculating a generator U-phase winding temperature vector, a generator V-phase winding temperature vector, a generator W-phase winding temperature vector, a gearbox oil temperature vector, a cooling system temperature vector and a temperature difference vector between a power electronic device temperature vector and an engine room environment temperature vector in a data set to respectively obtain a generator U-phase winding temperature vector, a generator V-phase winding temperature vector, a generator W-phase winding temperature vector, a gearbox oil temperature vector, a cooling system temperature vector and a power electronic device temperature vector;
and forming the obtained generator U-phase winding temperature difference vector, the generator V-phase winding temperature difference vector, the generator W-phase winding temperature difference vector, the gearbox oil temperature difference vector, the cooling system temperature difference vector and the power electronic device temperature difference vector to obtain a new data set matrix.
Preferably, in step 3, the temperature vector of each column in the new data set matrix in step 2 and the motor rotation speed vector in the data set in step 1 are converted into a group of scatter points on a two-dimensional plane, and a coordinate set of the group of scatter points on the two-dimensional plane is recorded as a two-dimensional point set, and the specific method is as follows:
respectively taking the temperature vector of each column in the new data set matrix in the step 2 as an x axis of planar two-dimensional distribution;
taking the motor rotating speed vector as a y axis of planar two-dimensional distribution;
and recording a coordinate set corresponding to the scattered point on the two dimensions of the plane as a two-dimensional point set corresponding to the column of temperature vectors.
Preferably, in step 4, performing DBSCAN density clustering calculation on the two-dimensional point set obtained in step 3 to obtain a state value vector corresponding to the row of temperature vectors, and the specific method is as follows:
s41, setting an Eps parameter and a MinPts parameter of the DBSCAN density cluster;
s42, calculating to obtain an Eps parameter through a k-distance algorithm;
s43, calculating to obtain a MinPts parameter according to the two-dimensional point set;
s44, clustering the two-dimensional point set according to the obtained Eps parameters to obtain a density vector corresponding to the two-dimensional point set;
s45, classifying the obtained density vectors according to the obtained MinPts parameters, and dividing the two-dimensional point concentrated point into a cluster boundary point set and a cluster boundary point set;
s46, dividing the points in the two-dimensional point set into core points, cluster boundary points and data abnormal points according to the obtained density vectors, the cluster boundary point set and the cluster boundary point set;
s47, calculating to obtain a state value vector corresponding to the row of temperature vectors at the time t according to the obtained core point, cluster boundary point and data anomaly point;
and S48, repeating S41 to S47, and calculating state value vectors corresponding to the row of temperature vectors at the t + delta t time, the t +2 delta t time, the t +3 delta t time and the t +4 delta t time respectively.
Preferably, in step 7, the outlier state of each state monitoring parameter vector is identified according to the state value vector corresponding to the obtained set time, and the specific method is as follows:
define OutljVariables and NegjA variable;
calculating Outl according to the corresponding state value vector within the set timejVariables and NegjA variable;
according to the calculated OutljVariables and NegjThe variables identify the outlier state of each corresponding state monitoring parameter vector.
Preferably, the Outl is obtained from the calculationjVariables and NegjThe variables identify the outlier state of each corresponding state monitoring parameter vector. The specific method comprises the following steps:
if OutljNot less than 4; the temperature difference of the u-phase winding of the generator of the j unit is normal;
if 2 is less than or equal to Outlj＜4，Negj0; the temperature difference of the u-phase winding of the generator of the j unit is normal;
if 2 is less than or equal to Outlj＜4，NegjIs greater than 0; the u-phase winding temperature difference of the generator of the j unit needs to be concerned;
if 0 is less than or equal to Outlj＜2，Negj0; the temperature difference of the u-phase winding of the generator of the j unit is normal;
if 0 is less than or equal to Outlj＜2，NegjIs greater than 0; the u-phase winding temperature difference of the generator of the j unit is separated;
if-2 is less than or equal to OutljLess than 0; the u-phase winding temperature difference of the generator of the j unit is separated;
if Outlj< -2 >; the u-phase winding temperature difference of the generator of the No. j unit is seriously separated.
A multi-unit operation state outlier analysis system based on density clustering comprises:
the system comprises a data acquisition unit, a data processing unit and a data processing unit, wherein the data acquisition unit is used for acquiring state monitoring parameters corresponding to each wind turbine generator set of a wind power plant at the time t and combining the state monitoring parameters to form a data set corresponding to the wind power plant, and the state monitoring parameters comprise motor rotating speed, torque, power, current effective value, generator U-phase winding temperature, generator V-phase winding temperature, generator W-phase winding temperature, gear box oil temperature, cooling system temperature, power electronic device temperature and cabin environment temperature;
the data matrix construction unit is used for processing the temperature data vectors in the data set to obtain a new data set matrix with temperature difference vectors:
the two-dimensional point set constructing unit is used for converting the temperature vector of each row in the new data set matrix and the motor rotating speed vector in the data set into a group of scattered points on a two-dimensional plane, and recording a coordinate set of the group of scattered points on the two-dimensional plane as a two-dimensional point set;
converting a torque vector and a power vector in a data set and a motor rotating speed vector in the data set into a group of scattered points on a two-dimensional plane respectively, and recording a coordinate set of the group of scattered points on the two-dimensional plane as a two-dimensional point set;
the state value vector construction unit is used for carrying out DBSCAN density clustering calculation on the obtained two-dimensional point set to obtain a corresponding state value vector within the set time of the temperature vector of the row;
combining the obtained two-dimensional point sets corresponding to the torque vector and the power vector to respectively obtain corresponding state value vectors within the set time of the torque vector and the power vector;
and the outlier identification unit is used for identifying the outlier state of each state monitoring parameter vector according to the corresponding state value vector within the set time.
Preferably, the state value vector construction unit includes:
the parameter setting unit is used for setting Eps parameters and MinPts parameters of the DBSCAN density clusters;
the parameter calculation unit is used for calculating an Eps parameter through a k-distance algorithm; calculating to obtain a MinPts parameter according to the two-dimensional point set;
the density vector calculation unit is used for clustering the two-dimensional point set according to the obtained Eps parameters to obtain a density vector corresponding to the two-dimensional point set;
the point set classification unit is used for classifying the obtained density vectors according to the obtained MinPts parameters and dividing the two-dimensional point set into a cluster boundary point set and a cluster boundary point set;
dividing the points in the two-dimensional point set into core points, cluster boundary points and data abnormal points according to the obtained density vectors, the cluster boundary point set and the cluster boundary point set;
and the state value vector calculation unit is used for calculating and obtaining a state value vector corresponding to the row of temperature vectors in the set time period according to the obtained core point, the cluster boundary point and the data abnormal point.
Compared with the prior art, the invention has the beneficial effects that:
the method and the system for analyzing the outlier of the running state of the multiple units based on the density clustering are used for realizing the state clustering of the multiple wind turbine units of the wind power plant in the same time period and identifying the center and the outlier of each type.
Drawings
FIG. 1 is a schematic flow diagram of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings.
Embodiment 1 is as shown in fig. 1, and the specific process of the method for analyzing the outlier of the operating state of multiple units based on density clustering provided by the invention includes the following steps:
step 1, because the area covered by a wind power plant is generally large, different units are located at different point locations, and the wind speed, the wind direction and the ambient temperature of different point locations are different, the running states of the whole wind turbine are also different even at the same moment; meanwhile, environmental parameters such as wind speed and wind direction change repeatedly, other unit state monitoring parameters need to respond to the change of the environmental parameters, and due to mechanical transmission, energy conversion responds to the change of the environmental parameters and lags. If the environmental parameters are directly set as independent variables and the unit state monitoring parameters are dependent variables, the real-time unit state monitoring parameters are discretely distributed near a stable state and are not beneficial to accurate classification. In actual operation, the generator is a direct device for converting mechanical energy converted from wind energy into electric energy, and the rotating speed is directly reflected, so that the rotating speed of the motor is selected as an independent variable for analysis, and state monitoring parameters of other units are taken as dependent variables.
A certain wind power plant has m units which are correspondingly numbered as 1, 2, …, j, … and m; the method comprises the following steps that n unit state monitoring parameters are set in the running process of each unit, wherein the n unit state monitoring parameters are respectively motor rotating speed, torque, power, current effective values, three-phase winding temperature, gear box oil temperature, cooling system temperature, power electronic device temperature, environment temperature and the like, and are correspondingly numbered 1, 2, …, i, … and n.
Acquiring n state monitoring parameters of m units of the wind power plant at the moment t, wherein the data set is recorded as D (t) and comprises the following steps:
Figure BDA0003338313370000061
wherein, D (t)1、D(t)2、...D(t)i、...D(t)nVector of monitoring parameters corresponding to n states at time t, e.g. D (t)1Is the motor speed vector, D (t)nThe vector is a cabin environment temperature monitoring vector, each vector is m-dimensional, and the vector corresponds to the units with the numbers of 1, 2, …, j, … and m.
D(t)i＝[d(t)i1 d(t)i2 ... d(t)ij ...d(t)im]T
Wherein, d (t)i1 d(t)i2 ... d(t)ij ... d(t)imAnd (4) the ith state monitoring parameters respectively corresponding to the m machine sets at the time T, wherein T represents transposition.
Step 2, processing the related temperature data vector in the data set matrix D (t) obtained in the step 1, and absolutely measuring the temperatureThe value vector is converted into a temperature difference vector of the monitored temperature and the environment, and D (t) in a known matrix D (t)nSetting original data set matrixes D (t) in D (t) for cabin environment temperature monitoring vectors corresponding to m sets at time tu1、D(t)u2、D(t)u3、D(t)u4、D(t)u5、D(t)u6The method is characterized in that the method comprises the following steps of respectively monitoring the U-phase winding temperature, the V-phase winding temperature, the W-phase winding temperature, the gearbox oil temperature, the cooling system temperature and the temperature monitoring vector of a power electronic device of m machine set generators at the time t, wherein U1, U2, U3, U4, U5 and U6 are positive integers, and the method comprises the following steps:
1＜u1＜u2＜u3＜u4＜u5＜u6＜n
then, the temperature monitoring vector D (t) will be specifieduaConverted into a temperature difference vector D' (t) with the environmentuaThe calculation method is as follows:
Figure BDA0003338313370000071
wherein, a is more than or equal to 1 and less than or equal to 6, D (t)uaAnd according to different values of a, respectively representing the U-phase winding temperature, the V-phase winding temperature, the W-phase winding temperature, the gearbox oil temperature, the cooling system temperature and the temperature monitoring vector of the power electronic device of the m machine set generators at the time t.
Related temperature data vector D (t) in original data set matrix D (t) item by itemu1、D(t)u2、D(t)u3、D(t)u4、D(t)u5、D(t)u6After processing, the temperature data monitoring vector D (t) in the original data set matrix D (t)u1、D(t)u2、D(t)u3、D(t)u4、D(t)u5、D(t)u6Become temperature difference vector D' (t)u1、D′(t)u2、D′(t)u3、D′(t)u4、D′(t)u5、D′(t)u6After transformation, a new data set matrix D' (t) is formed,
D′(t)＝[D(t)1 ... D′(t)u1、D′(t)u2、D′(t)u3、D′(t)u4、D′(t)u5、D′(t)u6 ... D(t)n]
step 3, for the matrix D' (t) obtained in the step 2, selecting D (t)1Namely, the rotating speed of the generator of each unit at the time t is taken as the x axis of the plane two-dimensional distribution, and the D '(t) of the matrix D' (t) obtained in the step 2 is useduaA y-axis as a planar two-dimensional distribution; in this example, with D' (t)u1Is the temperature difference vector of the generator u-phase winding temperature and the engine room temperature, then D (t)1And D' (t)u1The data is represented as a set of scattered points on a two-dimensional plane xoy. The coordinate set of the group of scattered points on the two-dimensional plane xoy is marked as Pt＝(Px，Py)tThen P istBy definition, as a two-dimensional set of points, there are:
Figure BDA0003338313370000081
wherein p ist1Representing the corresponding point of the temperature difference data of the motor speed of the No. 1 unit and the u-phase winding of the generator on a two-dimensional plane at the time t; similarly, ptj represents the corresponding point of the temperature difference data of the u-phase winding of the generator at the rotating speed of the motor of the unit of the number j at the time t on the two-dimensional plane.
Step 4, aiming at the two-dimensional point set P obtained in the step 3tDeveloping DBSCAN density clustering calculation, and setting two parameters Eps and MinPts required in the DBSCAN density clustering, wherein Eps represents the neighborhood radius of each cluster in the density clustering, and is marked as the element, and the k-distance algorithm is required by the calculation method of the element, and the method specifically comprises the following steps:
first, a parameter k, k being 2 × Dim-1,
where Dim is a set of points PtWhere Dim is 2, so k is 3;
calculate a set of points PtThe distance from each point to the point which is close to the kth point (here, 3), and then sorting the distances from large to small to obtain a curve which is marked as a k-distance curve; and identifying the position of the k-distance curve inflection point, wherein the distance length corresponding to the inflection point is the E.
MinPts represents a set of points PtIn (c) ptjWhen it becomes the center pointAnd in the minimum point number within the neighborhood radius epsilon, the calculation method is as follows:
MinPts＝Min(∈+1，γ)；
Figure BDA0003338313370000091
Figure BDA0003338313370000092
represents ln (m + e) rounding down; min (∈ +1, γ) means taking the smaller of ∈ +1 and γ.
Step 5, according to the two parameters Eps of the DBSCAN obtained in the step 4 and the Eps of the MinPts, starting a clustering algorithm to calculate the j point ptjFor example.
Introducing variable y, y epsilon PtI.e. y is the set of points ptjPoint of (1), calculate the jth point ptjSet of points contained in a neighborhood with e as radius, i.e. N∈(ptj)：
N∈(ptj)＝{y∈Pt：Eud(y，ptj)≤∈}；
Wherein Eud (y, p)tj) Representing point y and point ptjOf between, Euclidean distance, N∈(ptj) Representing a set of points ptjSatisfy Eud (y, p)tj) Set of all points y ≦ ε.
Step 6, repeating the calculation of step 5, and collecting the point set PtFirst point p oft1Start, to ptmStopping, calculating item by item to obtain N∈(pt1)、N∈(pt2)、...、N∈(ptj)、...、N∈(ptm) The result is a set of m points, N∈(Pt) The matrix that represents the set of points is:
N∈(Pt)＝[N∈(pt1) N∈(pt2) ... N∈(ptj) ... N∈(ptm)]
step 7, calculating m point sets obtained in step 6 item by itemDensity of (i.e. number of elements in the point set, N)∈(ptj) Is denoted as ρ∈(ptj)：
ρ∈(ptj)＝|N∈(ptj)|
Then a set of density vectors, using p, is obtained after this step is completed∈(Pt) Representing this set of density vectors, then:
ρ∈(Pt)＝[ρ∈(pt1) ρ∈(pt2) ... ρ∈(ptj) ... ρ∈(ptm)]
step 8, the elements of the density vector obtained in the step 7 are classified according to the two parameters Eps and MinPts of the DBSCAN obtained in the step 4, the core point (core) is firstly identified, and the arbitrary point p is selectedtjAt first, j is more than or equal to 1 and less than or equal to m, and the specific method is as follows:
for point ptj∈PtIf ρ∈(ptj) MinPts, i.e. point set N∈(ptj) If the number of the middle element is more than or equal to MinPts, the point p is indicatedtjIs the core point (core), and a new cluster C is formed at this timejSet points N∈(ptj) All points in the cluster Cj(ii) a If ρ∈(ptj) < MinPts, point p will betjFirstly, marking as a non-core point;
for point pt(j+1)∈PtIf ρ∈(pt(j+1)) Not less than MinPts, i.e. point set N∈(pt(j+1)) If the number of the middle element is more than or equal to MinPts, the point p is indicatedt(j+1)Also core point (core), when a new cluster C is formed(j+1)Set points N∈(pt(j+1)) In a cluster other than CjAll points falling into a cluster C(j+1)(ii) a In the same way, if ρ∈(pt(j+1)) < MinPts, point p will bet(j+1)First labeled as non-core points:
by analogy, comparing the density of each point with MinPts item by item to obtain all core points and non-core points, and recording the set formed by all the core points as PtCWhat is, what isThe set of non-core points is PtNCThen, there are:
Pt＝PtC+PtNC
simultaneously obtaining lambda clusters by calculation, wherein
Figure BDA0003338313370000101
Step 9, for all the non-core point sets P in step 8tNCClassifying the non-core points into a cluster boundary point set PtbAnd data anomaly set PtimThen, there are:
PtNC＝[ptNC_1 ptNC_2 ... ptNC_z]
wherein p istNC_1 ptNC_2 ... ptNC_zRepresenting a set of non-core points PtNCZ for a set of non-core points PtNCPoint p in (1)tNC_z，N∈(ptNC_z) A neighborhood of which is represented that is,
if N is present∈(ptNC_z) In which at least one point alpha is present and alpha is a core point, i.e. alpha ∈ PtCThen point ptNC_zIs a cluster boundary point;
if N is present∈(ptNC_z) If any one point is not the core point, the point ptNC_zIs a data anomaly point;
judging P item by itemtNCThe cluster boundary point set is marked as PtbdSet of cluster boundary points as PtimThen, there are:
Ptbd＝[ptbd_1 ptbd_2 ... ptbd_v]
Ptim＝[ptim_1 ptim_2 ... ptim_w]
PtNC＝Ptbd+Ptim
z＝v+w
wherein p istbd_1 ptbd_2 ... ptb_vDenotes cluster boundary points, v in total, ptim_1 ptim_2 ... ptim_wRepresenting data anomaly points. And the number of the devices is w.
Step 10, integrating the calculation classification results of step 8 and step 9, so as to set a group of point sets P at time ttPoints in (1) are classified into 3 types, and a variable val is introducedtjRepresents a point set PtAnd the data corresponding to the j-th point, namely the state value of the j-th unit at the time t.
Figure BDA0003338313370000111
The state value vector Val of the temperature difference of the u-phase winding of the generator set at the moment t can be obtained through calculationt，
Valt＝[valt1 valt2 ... valtj ... valtm]
Step 11, repeating the steps 4 to 10, and calculating the state value vector Val of the u-phase winding temperature difference of the generator at the t + delta t moment, the t +2 delta t moment, the t +3 delta t moment and the t +4 delta t momentt+Δt、Valt+2Δt、Valt+3ΔtAnd Valt+4Δt。
Valt+Δt＝[val(t+Δt)1 val(t+Δt)2 ... val(t+Δt)j ... val(t+Δt)m]
Valt+2Δt＝[val(t+2Δt)1 val(t+2Δt)2 ... val(t+2Δt)j ... val(t+2Δt)m]
Valt+3Δt＝[val(t+3Δt)1 val(t+3Δt)2 ... val(t+3Δt)j ... val(t+3Δt)m]
Valt+4Δt＝[val(t+4Δt)1 val(t+4Δt)2 ... val(t+4Δt)j ... val(t+4Δt)m]
According to Valt、Valt+Δt、Valt+2Δt、Valt+3ΔtAnd Valt+4ΔtIdentifying the outlier state of the u-phase winding temperature difference of each wind turbine generator in the period of time, wherein the specific method comprises the following steps:
two variables Outl are definedjAnd NegjOutlier state, Outl, for measuring u-phase winding temperature difference of generator of number j unitjAnd NegjThe calculation method of (2) is as follows:
Figure BDA0003338313370000112
step 12, according to the Outl calculated in step 11jAnd NegjMeasuring the outlier state of the u-phase winding temperature difference of the generator in the time period from t to t +4 delta t of the j-number unit, wherein the specific judgment method comprises the following steps:
Figure BDA0003338313370000121
after simplification, we obtain:
Figure BDA0003338313370000122
the 'abnormal occurrence' indicates that the temperature difference of the u-phase winding of the generator is already in an outlier state or is about to enter the outlier state compared with other units in the time period from t to t +4 delta t of the j unit. The 'normal' indicates that the temperature difference of the u-phase winding of the generator is not separated in the time period from t to t +4 delta t of the j unit. For a unit in the state of "abnormal occurrence", a patrol should be immediately scheduled.
And 13, repeating the steps 3 to 12, selecting the rotating speed of the motor as an x-axis coordinate of two-dimensional distribution, selecting other vectors (torque, power, current effective value, gear box oil temperature difference, power electronic device temperature difference and the like) in the D '(t) one by one as a y-axis coordinate of two-dimensional distribution, calculating the state of each wind turbine generator in each column vector in the data set matrix D' (t), and identifying the outlier. And arranging patrol for the unit with abnormal state.

Claims (8)

1. A multi-unit operation state outlier analysis method based on density clustering is characterized by comprising the following steps:
step 1, acquiring each state monitoring parameter corresponding to each wind turbine of a wind power plant at t moment, converting each state monitoring parameter into a corresponding state monitoring parameter vector, and combining to form a data set matrix corresponding to the wind power plant according to each obtained state monitoring parameter vector, wherein each state monitoring parameter vector is a motor rotating speed vector, a torque vector, a power vector, a current effective value vector, a generator U-phase winding temperature vector, a generator V-phase winding temperature vector, a generator W-phase winding temperature vector, a gearbox oil temperature vector, a cooling system temperature vector, a power electronic device temperature vector and a cabin environment temperature vector;
step 2, processing the temperature data vectors in the data set in the step 1 to obtain a new data set matrix with temperature difference vectors, wherein the temperature data vectors comprise a generator U-phase winding temperature vector, a generator V-phase winding temperature vector, a generator W-phase winding temperature vector, a gearbox oil temperature vector, a cooling system temperature vector, a power electronic device temperature vector and a cabin environment temperature vector;
step 3, converting the temperature vectors of each row in the new data set matrix in the step 2 and the motor rotating speed vectors in the data set in the step 1 into a group of scattered points on a two-dimensional plane, and recording a coordinate set of the group of scattered points on the two-dimensional plane as a temperature two-dimensional point set;
step 4, performing density clustering calculation on the temperature two-dimensional point set obtained in the step 3 to obtain a corresponding state value vector within the set time of the temperature vector of the row;
step 5, converting the torque vector and the power vector in the data set in the step 1 and the motor rotating speed vector in the data set in the step 1 into a group of scattered points on a two-dimensional plane, and respectively recording coordinate sets of the two groups of scattered points on the two-dimensional plane as a torque two-dimensional point set and a power two-dimensional point set;
step 6, respectively obtaining corresponding state value vectors within the torque vector setting time and the power vector setting time according to the torque two-dimensional point set and the power two-dimensional point set obtained in the step 5;
and 7, identifying the outlier state of each state monitoring parameter vector according to the corresponding state value vector within the set time.
2. The method for multi-unit operating state outlier analysis based on density clustering as claimed in claim 1, wherein in step 2, the temperature data vectors in the data set in step 1 are processed to obtain a new data set matrix with temperature difference vectors, and the specific method is as follows:
respectively calculating a generator U-phase winding temperature vector, a generator V-phase winding temperature vector, a generator W-phase winding temperature vector, a gearbox oil temperature vector, a cooling system temperature vector and a temperature difference vector between a power electronic device temperature vector and an engine room environment temperature vector in a data set to respectively obtain a generator U-phase winding temperature vector, a generator V-phase winding temperature vector, a generator W-phase winding temperature vector, a gearbox oil temperature vector, a cooling system temperature vector and a power electronic device temperature vector;
and forming the obtained generator U-phase winding temperature difference vector, the generator V-phase winding temperature difference vector, the generator W-phase winding temperature difference vector, the gearbox oil temperature difference vector, the cooling system temperature difference vector and the power electronic device temperature difference vector to obtain a new data set matrix.
3. The method for analyzing the multi-unit operating state outlier based on the density cluster as recited in claim 1, wherein in step 3, the temperature vector of each column in the new data set matrix in step 2 and the motor speed vector in the data set in step 1 are converted into a group of scattered points on a two-dimensional plane, and the coordinate set of the group of scattered points on the two-dimensional plane is recorded as a temperature two-dimensional point set, and the method comprises the following specific steps:
respectively taking the temperature vector of each column in the new data set matrix in the step 2 as an x axis of planar two-dimensional distribution;
taking the motor rotating speed vector as a y axis of planar two-dimensional distribution;
and recording a coordinate set corresponding to the scattered point on the two dimensions of the plane as a two-dimensional point set corresponding to the column of temperature vectors.
4. The method for analyzing the outlier of the multi-unit operation state based on the density clustering of claim 1, wherein in the step 4, the density clustering calculation is performed on the temperature two-dimensional point set obtained in the step 3 to obtain the corresponding state value vector of the row of temperature vectors within the set time, and the specific method is as follows:
s41, setting Eps parameters and MinPts parameters of density clustering calculation;
s42, calculating to obtain an Eps parameter through a k-distance algorithm;
s43, calculating to obtain a MinPts parameter according to the two-dimensional point set;
s44, clustering the two-dimensional point set according to the obtained Eps parameters to obtain a density vector corresponding to the two-dimensional point set;
s45, classifying the obtained density vectors according to the obtained MinPts parameters, and dividing the two-dimensional point concentrated point into a cluster boundary point set and a cluster boundary point set;
s46, dividing the points in the two-dimensional point set into core points, cluster boundary points and data abnormal points according to the obtained density vectors, the cluster boundary point set and the cluster boundary point set;
s47, calculating to obtain a state value vector corresponding to the row of temperature vectors at the time t according to the obtained core point, cluster boundary point and data anomaly point;
and S48, repeating S41 to S47, and calculating state value vectors corresponding to the row of temperature vectors at the t + delta t time, the t +2 delta t time, the t +3 delta t time and the t +4 delta t time respectively.
5. The method for analyzing the outlier of the multi-unit operation state based on the density clustering as claimed in claim 1, wherein in step 7, the outlier state of each state monitoring parameter vector is identified according to the corresponding state value vector obtained within the set time, and the specific method is as follows:
define OutljVariables and NegjA variable;
calculating Outl according to the corresponding state value vector within the set timejVariables and NegjA variable;
according to the calculated OutljVariables and NegjThe variables identify the outlier state of each corresponding state monitoring parameter vector.
6. The method as claimed in claim 5, wherein the outlier analysis method is based on the calculated OutljVariables and NegjIdentifying the outlier state of each corresponding state monitoring parameter vector by using the variable, wherein the specific method comprises the following steps:
if OutljNot less than 4; the temperature difference of the u-phase winding of the generator of the j unit is normal;
if 2 is less than or equal to Outlj＜4，Negj0; the temperature difference of the u-phase winding of the generator of the j unit is normal;
if 2 is less than or equal to Outlj＜4，NegjIs greater than 0; the u-phase winding temperature difference of the generator of the j unit needs to be concerned;
if 0 is less than or equal to Outlj＜2，Negj0; the temperature difference of the u-phase winding of the generator of the j unit is normal;
if 0 is less than or equal to Outlj＜2，NegjIs greater than 0; the u-phase winding temperature difference of the generator of the j unit is separated;
if-2 is less than or equal to OutljLess than 0; the u-phase winding temperature difference of the generator of the j unit is separated;
if Outlj< -2 >; the u-phase winding temperature difference of the generator of the No. j unit is seriously separated.
7. A multi-unit operation state outlier analysis system based on density clustering is characterized by comprising the following steps:
the data acquisition unit is used for acquiring each state monitoring parameter corresponding to each wind turbine generator set at the moment t of the wind power plant, converting each state monitoring parameter into a corresponding state monitoring parameter vector, and combining to form a data set matrix corresponding to the wind power plant according to each obtained state monitoring parameter vector, wherein each state monitoring parameter vector is a motor rotating speed vector, a torque vector, a power vector, a current effective value vector, a generator U-phase winding temperature vector, a generator V-phase winding temperature vector, a generator W-phase winding temperature vector, a gearbox oil temperature vector, a cooling system temperature vector, a power electronic device temperature vector and a cabin environment temperature vector;
the data matrix construction unit is used for processing the temperature data vectors in the data set to obtain a new data set matrix with temperature difference vectors;
the two-dimensional point set constructing unit is used for converting the temperature vector of each row in the new data set matrix and the motor rotating speed vector in the data set into a group of scattered points on a two-dimensional plane, and recording the coordinate set of the group of scattered points on the two-dimensional plane as a temperature two-dimensional point set;
converting the torque vector and the power vector in the data set and the motor rotating speed vector in the data set into a group of scattered points on a two-dimensional plane, and respectively recording coordinate sets of the two groups of scattered points on the two-dimensional plane as a torque two-dimensional point set and a power two-dimensional point set;
the state value vector construction unit is used for carrying out DBSCAN density clustering calculation on the obtained two-dimensional point set to obtain a corresponding state value vector within the set time of the temperature vector of the row;
respectively obtaining corresponding state value vectors within the torque vector setting time and the power vector setting time according to the obtained torque two-dimensional point set and the power two-dimensional point set;
and the outlier identification unit is used for identifying the outlier state of each state monitoring parameter vector according to the corresponding state value vector within the set time.
8. The system of claim 1, wherein the state value vector construction unit comprises:
the parameter setting unit is used for setting Eps parameters and MinPts parameters of the DBSCAN density clusters;
the parameter calculation unit is used for calculating an Eps parameter through a k-distance algorithm; calculating to obtain a MinPts parameter according to the two-dimensional point set;
the density vector calculation unit is used for clustering the two-dimensional point set according to the obtained Eps parameters to obtain a density vector corresponding to the two-dimensional point set;
the point set classification unit is used for classifying the obtained density vectors according to the obtained MinPts parameters and dividing the two-dimensional point set into a cluster boundary point set and a cluster boundary point set;
dividing the points in the two-dimensional point set into core points, cluster boundary points and data abnormal points according to the obtained density vectors, the cluster boundary point set and the cluster boundary point set;
and the state value vector calculation unit is used for calculating and obtaining a state value vector corresponding to the row of temperature vectors in the set time period according to the obtained core point, the cluster boundary point and the data abnormal point.
CN202111300977.1A 2021-11-04 2021-11-04 Multi-unit operation state outlier analysis method and system based on density clustering Pending CN114004512A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111300977.1A CN114004512A (en) 2021-11-04 2021-11-04 Multi-unit operation state outlier analysis method and system based on density clustering

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111300977.1A CN114004512A (en) 2021-11-04 2021-11-04 Multi-unit operation state outlier analysis method and system based on density clustering

Publications (1)

Publication Number Publication Date
CN114004512A true CN114004512A (en) 2022-02-01

Family

ID=79927342

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111300977.1A Pending CN114004512A (en) 2021-11-04 2021-11-04 Multi-unit operation state outlier analysis method and system based on density clustering

Country Status (1)

Country Link
CN (1) CN114004512A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116186634A (en) * 2023-04-26 2023-05-30 青岛新航农高科产业发展有限公司 Intelligent management system for construction data of building engineering
CN116662111A (en) * 2023-05-05 2023-08-29 浙江锐明智能控制技术有限公司 Intelligent network management control system for train
CN117436352A (en) * 2023-12-20 2024-01-23 聚合电力工程设计(北京)股份有限公司 Wind farm noise analysis method and system

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116186634A (en) * 2023-04-26 2023-05-30 青岛新航农高科产业发展有限公司 Intelligent management system for construction data of building engineering
CN116186634B (en) * 2023-04-26 2023-07-07 青岛新航农高科产业发展有限公司 Intelligent management system for construction data of building engineering
CN116662111A (en) * 2023-05-05 2023-08-29 浙江锐明智能控制技术有限公司 Intelligent network management control system for train
CN116662111B (en) * 2023-05-05 2023-11-17 浙江锐明智能控制技术有限公司 Intelligent network management control system for train
CN117436352A (en) * 2023-12-20 2024-01-23 聚合电力工程设计(北京)股份有限公司 Wind farm noise analysis method and system
CN117436352B (en) * 2023-12-20 2024-03-22 聚合电力工程设计(北京)股份有限公司 Wind farm noise analysis method and system

Similar Documents

Publication Publication Date Title
Liu et al. A small-sample wind turbine fault detection method with synthetic fault data using generative adversarial nets
Wang et al. An enhanced intelligent diagnosis method based on multi-sensor image fusion via improved deep learning network
CN114004512A (en) Multi-unit operation state outlier analysis method and system based on density clustering
Li et al. Self-attention ConvLSTM and its application in RUL prediction of rolling bearings
Li et al. A deep adversarial transfer learning network for machinery emerging fault detection
Gou et al. Aeroengine control system sensor fault diagnosis based on CWT and CNN
Liu et al. Data super-network fault prediction model and maintenance strategy for mechanical product based on digital twin
Tran et al. Effective fault diagnosis based on wavelet and convolutional attention neural network for induction motors
Zhang et al. Fault analysis and condition monitoring of the wind turbine gearbox
Chen et al. Multi-fault diagnosis study on roller bearing based on multi-kernel support vector machine with chaotic particle swarm optimization
Qu et al. A multi-fault detection method with improved triplet loss based on hard sample mining
Cao et al. Fault diagnosis of wind turbine gearbox based on deep bi-directional long short-term memory under time-varying non-stationary operating conditions
CN108072524B (en) Wind turbine generator gearbox bearing fault early warning method
CN110766175A (en) Pitch system fault detection method and device based on optimal interval distribution machine
CN111582392A (en) Multi-working-condition health state online monitoring method for key components of wind turbine generator
Saeki et al. Visual explanation of neural network based rotation machinery anomaly detection system
CN108564313A (en) The method and device of Wind turbines status monitoring and health evaluating based on fault tree
CN111538947A (en) Method for constructing wind power generator bearing fault classification model
Gong et al. A fast anomaly diagnosis approach based on modified CNN and multisensor data fusion
CN111709190A (en) Wind turbine generator operation data image identification method and device
Yang et al. Compound fault diagnosis of harmonic drives using deep capsule graph convolutional network
Peng et al. Deep unsupervised transfer learning for health status prediction of a fleet of wind turbines with unbalanced data
Li et al. A novel adaptive STFT-SFA based fault detection method for nonstationary processes
CN116467592A (en) Production equipment fault intelligent monitoring method and system based on deep learning
Pang et al. Multi kernel fusion convolutional neural network for wind turbine fault diagnosis

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination