A method and system for analyzing the operating state of a frequency converter

Abstract

The application provides a variable frequency device operation state outlier analysis method and system, including the following steps: step 1, obtaining original monitoring data of actual operation of a single variable frequency device in a SCADA system, denoted as matrix A; step 2, cleaning the data in the matrix A obtained in step 1 to obtain a matrix A'; step 3, segmenting the matrix A' obtained in step 2 to obtain L sub-matrices; step 4, calculating a state vector value corresponding to each sub-matrix; step 5, performing outlier analysis on the variable frequency device operation state according to the state vector value; the application can find the device with state deviation in advance and achieve early warning before failure. No pre-training is required according to historical data, and the application has strong universality.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of wind power generation, in particular to a variable frequency converter operation state outlier analysis method and system. BACKGROUND

[0002] The wind turbine converter is a key energy conversion device of the wind turbine, and the converter is high-frequency on-off during the operation of the wind turbine. The heat accumulates in the power module during the on-off process, which causes the temperature of the power module to rise and the failure rate to rise. In view of this, the module monitoring temperature becomes a parameter that can intuitively reflect the state of the device. However, the temperature change is a slow process, and the traditional method only focuses on the temperature change of a single device and sets a temperature alarm threshold, which often has a lag. Moreover, to avoid false alarms, the threshold setting value is generally large. SUMMARY

[0003] The present application provides a variable frequency converter operation state outlier analysis method and system, which solves the above-mentioned deficiencies in the prior art. Based on the original temperature analysis, the present application adds new state parameter analysis to realize comparative analysis of the device state and find the device with state deviation in advance to achieve early warning before failure occurs.

[0004] In order to achieve the above-mentioned purpose, the technical scheme adopted by the present application is:

[0005] The present application provides a variable frequency converter operation state outlier analysis method, which comprises the following steps:

[0006] Step 1: obtaining the original monitoring data of a single variable frequency converter actually running in the SCADA system, denoted as matrix A;

[0007] Step 2: cleaning the data in the matrix A obtained in step 1 to obtain matrix A';

[0008] Step 3: segmenting the matrix A' obtained in step 2 to obtain L sub-matrices;

[0009] Step 4: calculating the state vector value corresponding to each sub-matrix;

[0010] Step 5: performing outlier analysis on the variable frequency converter operation state according to the state vector value.

[0011] Preferably, in step 1, the matrix A is n rows and five columns, and the five columns are in turn the unit wind speed, the unit power, the converter machine side current, the converter grid side current and the converter outlet water temperature.

[0012] Preferably, in step 2, the specific method for cleaning the data in the matrix A obtained in step 1 is:

[0013] The out-of-limit point data, outlier data and stacked point data in the matrix A data are cleaned.

[0014] Preferably, in step 3, the matrix A' obtained in step 2 is segmented according to the first column data to obtain L sub-matrices; wherein the segmentation number L is calculated by the following formula:

[0015]

[0016] In the formula, s' j is the jth cleaned unit wind speed value in the first column vector; s'1 is the 1st cleaned unit wind speed value in the first column vector.

[0017] Preferably, in step 4, the state vector value corresponding to each sub-matrix is calculated, and the specific method is as follows:

[0018] S401, the distribution graph between power and frequency converter machine side current, power and grid side current, power and water temperature under different wind speed conditions in each sub-matrix is calculated;

[0019] S402, the edge point of each distribution graph obtained in S401 is identified;

[0020] S403, the fitting ellipse corresponding to each distribution graph is fitted according to the obtained edge point;

[0021] S404, the parameter value of each fitting ellipse is calculated;

[0022] S405, according to the obtained parameter value, the sum of the Euclidean distances between all distribution points on each distribution graph and the two foci of the corresponding fitting ellipse is calculated to obtain the Euclidean distance vector corresponding to each distribution graph;

[0023] S406, the distance distribution result corresponding to each Euclidean distance vector is calculated;

[0024] S407, the state value corresponding to each distribution graph is calculated according to the obtained distance distribution result;

[0025] S408, the state vector value corresponding to each sub-matrix is calculated according to the state value corresponding to each distribution graph.

[0026] Preferably, in S407, the state value corresponding to each distribution graph is calculated according to the obtained distance distribution result, and the specific method is as follows:

[0027]

[0028]

[0029]

[0030]

[0031]

[0032]

[0033] in, For the first Within the wind speed range, the state values ​​between power and inverter side current; They represent the first Various leading factors within the wind speed range; a′=0.51×a; d F1 +d F2 c is the sum of the Euclidean distances from each distribution point to the two foci; c is the focal length.

[0034] Preferably, in S408, the state vector value corresponding to each submatrix is ​​calculated based on the state value corresponding to each distribution map. The specific method is as follows:

[0035]

[0036] in, The state vector value corresponding to each submatrix; They represent the first Within the wind speed range, the state values ​​between power and inverter machine-side current, power and inverter grid-side current, and power and inverter outlet water temperature.

[0037] Preferably, in step 5, outlier analysis is performed on the inverter's operating state based on the state vector values. The specific method is as follows:

[0038] The status of the machine-side module is out of order and abnormal.

[0039] The network-side module is out of state and is abnormal.

[0040] The cooling system is malfunctioning or out of order.

[0041] in, They represent the first Within the wind speed range, the state values ​​between power and inverter machine-side current, power and inverter grid-side current, and power and inverter outlet water temperature.

[0042] A system for analyzing outliers in the operating status of a frequency converter, comprising:

[0043] The data acquisition unit is used to acquire the raw monitoring data of the actual operation of a single frequency converter in the SCADA system, denoted as matrix A;

[0044] a data cleaning unit, configured to clean data in the obtained matrix A to obtain a matrix A';

[0045] a data segmentation unit, configured to segment the obtained matrix A' to obtain L sub-matrices;

[0046] a data calculation unit, configured to calculate a state vector value corresponding to each sub-matrix;

[0047] a data analysis unit, configured to perform outlier analysis on the running state of the frequency converter according to the state vector value.

[0048] Compared with the prior art, the beneficial effects of the present application are:

[0049] The frequency converter running state outlier analysis method provided by the present application adds new state parameter analysis on the basis of original temperature analysis, realizes comparative analysis of device states, can find devices with state deviation in advance, and realizes early warning before failure occurs. No pre-training is required according to historical data, and the method has strong universality. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 is a flowchart of the present application;

[0051] Figure 2 is a power and frequency converter machine side current distribution graph under the condition of a wind speed of 6m / s-6.5m / s;

[0052] Figure 3 is a power and frequency converter machine side current distribution graph fitting result under the condition of a wind speed of 6m / s-6.5m / s. DETAILED DESCRIPTION

[0053] The present application will be further described in detail below with reference to the accompanying drawings.

[0054] As shown in the accompanying Figure 1 , the present application provides a frequency converter running state outlier analysis method, and the specific steps are as follows:

[0055] Step 1, collect original monitoring data related to the running of a single frequency converter in the SCADA system during actual operation (monitoring data for 2 consecutive months in this example), which covers various monitoring data under different operating conditions, denoted as matrix A. Matrix A is n rows and 5 columns, the first column of data is the unit wind speed, arranged in order from small to large, denoted as S=[s1, s2, …s i , …s n ]; the second column of data is the unit power monitoring value corresponding to the first column of unit wind speed data, denoted as P=[p1, p2, …p i , …p n]; the third column data is the inverter machine side current monitoring value corresponding to the first column unit wind speed data, denoted as Cr=[cr1, cr2, …cr i n ]; the fourth column data is the inverter grid side current monitoring value corresponding to the first column unit wind speed data, denoted as Cg=[cg1, cg2, …cg i n ]; the fifth column data is the inverter outlet water temperature monitoring value corresponding to the first column unit wind speed data, denoted as T=[t1, t2, …t i n ]; in the scheme, the vectors are column vectors, so:

[0056]

[0057] Step 2, the data of matrix A is cleaned, the out-of-limit points, outliers and stacking points in matrix A are removed, and the data matrix after data cleaning is denoted as A', and the A' is composed of column vectors S', P', Cr', Cg' and T':

[0058]

[0059] Among them, the column vectors S', P', Cr', Cg' and T' are j-dimensional vectors, respectively representing the unit wind speed, unit power monitoring value, inverter machine side current monitoring value, inverter grid side current monitoring value and inverter outlet water temperature monitoring value after cleaning.

[0060] Step 3, according to the value range of the first column S' of the matrix A' obtained in step 2, the matrix A' is segmented, and the number of segments is set to L, so:

[0061]

[0062]

[0063] It is indicated that the matrix A' can be divided into L sub-matrices, A'=[A1, A2, …, AU, …, AL], wherein AU=[SU, PU, CrU, CgU, TU]

[0064] Among them, the value range of S1 is:

[0065]

[0066] S1 min = s′1

[0067] P1, Cr1, Cg1 and T1 represent the corresponding data in the value range of S1, and m represents the number of data, so: ​​​

[0068] P1 = [p 11 , p 12 ,..., p 1k ,..., p 1m ]

[0069] Cr1 = [cr 11 , cr 12 ,..., cr 1k ,..., cr 1m ]

[0070] Cg1 = [cg 11 , cg 12 ,..., cg 1k ,..., cg 1m ]

[0071] T1 = [t 11 , t 12 ,..., t 1k ,..., t 1m ]

[0072] Similarly, the range of values of SU can be obtained as:

[0073]

[0074]

[0075] 1 < U < L

[0076] PU, CrU, CgU and TU represent the corresponding data in the range of values of SU, and f represents the number of data, so:

[0077] PU = [p U1 , p U2 ,..., p Uk ,..., p Uf ]

[0078] CrU = [cr U1 , cr U2 ,..., cr Uk ,..., cr Uf ]

[0079] CgU = [cg U1 , cg U2 ,..., cg Uk ,..., cg Uf ]

[0080] TU = [t U1 , t U2 ,..., t Uk ,..., tUf ]

[0081] Step 4, data analysis is carried out according to the sub-matrix obtained in step 3, and the distribution diagram between the corresponding power and the current on the machine side of the frequency converter, the power and the current on the grid side, and the power and the water temperature under different wind speed conditions in each sub-matrix is obtained; in this example, PU and CrU are taken as an example (taking PU and CrU as the power and the current on the machine side of the frequency converter corresponding to the wind speed of 6m / s-6.5m / s, and the other data is the same); the specific method is:

[0082] PU and CrU are taken as a group of two-dimensional variables x and y respectively, then according to the statistical method, the joint distribution of PU and CrU on the xoy plane can be obtained, and the distribution diagram between PU and CrU is obtained, as shown in Figure 2 .

[0083] Step 5, the edge points of each distribution diagram obtained in step 4 are identified by the convex hull identification method, and this group of edge points is recorded as:

[0084]

[0085] In the above formula, represents the vector composed of the coordinates of each edge point on the xoy plane, (p Uaa1 , cr Uaa1 ), (p Uaa2 , cr Uaa2 ), …, (p Uaaz , cr Uaaz ) represents the coordinates of the z edge points on the xoy plane.

[0086] The vector is fitted by using a nonlinear fitting method, and a fitting ellipse is obtained; the objective function of the fitting ellipse is set as:

[0087]

[0088]

[0089] Wherein, the objective function is the general form of the ellipse equation, λ1, λ2, λ3, λ4, λ5, 1 are the coefficients of each term of the ellipse equation.

[0090] According to the definition, the ellipse satisfying the fitting of this group of edge points is unique, that is, the vector [λ1, λ2, λ3, λ4, λ5, 1] is unique, and the ellipse is recorded as The vector is recorded as Then we have:

[0091]

[0092] Based on the geometric meaning of an ellipse, the data points on the xoy plane obtained in step 4, excluding edge points, are all distributed within the ellipse. internal.

[0093] Step 6, based on the ellipse obtained in Step 5 The coefficient vector [λ1, λ2, λ3, λ4, λ5, 1] ​​is used to calculate the coordinates of the center point of the ellipse, denoted as (Centroid). x Centroid y ) or (X C Y C The length of the major axis is denoted as majoraxis or a, and the length of the minor axis is denoted as minoraxis or b; the angle between the major axis and the positive x-axis of the xoy plane is denoted as θ; the focal length is denoted as c; and the coordinates of the focal point are denoted as (Fo). x Fo y ).

[0094] The coordinates of the center point, the length of the major axis, the length of the minor axis, the angle between the major axis and the positive x-axis of the xoy plane, the focal length, and the coordinates of the focal point are calculated as follows:

[0095]

[0096]

[0097]

[0098]

[0099]

[0100]

[0101] Fo x =X C ±c×cosθ

[0102] Fo y =Y C ±c×sinθ and Fo x Calculate the same sign

[0103] The ellipse fitting results and the above ellipse parameter calculation results are displayed in the distribution plot, as shown in the appendix. Figure 3 As shown.

[0104] Step 7, based on the ellipse obtained in Step 6 The parameters are used to establish a state analysis method, and the sum of the Euclidean distances d = d from the distribution points on the xoy plane to the two foci is calculated item by item in step 4. F1 +d F2The Euclidean distance vector can be obtained. Then we have:

[0105]

[0106] According to the geometric meaning of an ellipse, an ellipse The sum of the distances from any point on the curve to the two foci is 2a. (This is an example of an ellipse.) The sum of the distances from any point within the area enclosed by the curve to the two foci is less than 2a and greater than or equal to 2c; considering the error in the distribution of edge points used in the actual ellipse fitting process, 1.05 times 2a is taken as the reference coefficient, that is:

[0107] d F1 +d F2 >1.05×2a; This distribution point lies in an ellipse. Outside the corresponding range;

[0108] 2c≤d F1 +d F2 ≤1.05×2a; This distribution point lies in an ellipse Within the corresponding range;

[0109] Step 8, Statistics from Step 7 The intervals, frequencies, and percentages of each distance calculation value are shown in Table 1.

[0110]

[0111] Based on the calculation results of the ellipse parameters in step 6, we can conclude that:

[0112] 2c = 1769;

[0113] 2a′=1.05×2a=2181:

[0114] As can be seen from Table 1, approximately 85% of the distance calculation results d F1 +d F2 Distributed within the interval [1769, 1815], overall 98% of the distance calculation results d F1 +d F2 Distributed within the interval [1769, 1954], 100% of the distance calculation results d F1 +d F2 The distribution is within the interval [1769, 2184], which is basically consistent with 2a′=1.05×2a.

[0115] Step 9: Calculate the state value CV based on the distance distribution results obtained in Step 8. PCr (Condition Value), the specific calculation method is as follows:

[0116] CV PCr = 1; 2c < d F1 + d F2 ≤ 1.66a';

[0117] CV PCr = 0.9; 1.66a' < d F1 + d F2 ≤ 1.80a';

[0118] CV PCr = 0.7; 1.80a' < d F1 + d F2 ≤ 2a';

[0119] CV PCr = 0.6; 2a' < d F1 + d F2 ≤ 2.20a';

[0120] CV PCr = 0.4; 2.20a' < d F1 + d F2 ≤ 2.50a';

[0121] CV PCr = 0; 2.50a' < d F1 + d F2 ;

[0122] Step 10, repeat steps 4 to 9 to obtain the power and frequency converter machine side current state value CV calculation method under different wind speed conditions, the overall calculation method is as follows:

[0123]

[0124]

[0125]

[0126]

[0127]

[0128]

[0129] wherein, represents the power and frequency converter machine side current state value in the first wind speed interval, represents each pre-coefficient in the first wind speed interval, which is determined according to the different wind speed intervals and the focal distance distribution results of the sub-matrix elliptical fitting.

[0130] Step 11, repeat the previous steps, calculate the power and the inverter grid side current, power and inverter outlet water temperature monitoring values in different wind speed interval The calculation results are similar to step 10, then a group of state value vectors corresponding to each sub matrix can be obtained

[0131]

[0132] Among them, The state value of the power and the inverter grid side current, the power and the inverter grid side current, and the power and the inverter outlet water temperature monitoring value in the first wind speed interval. The state value of the power and the inverter grid side current, the power and the inverter grid side current, and the power and the inverter outlet water temperature monitoring value in the first wind speed interval.

[0133] Step 12, according to the state value vector obtained in step 11 The inverter running state outlier analysis is carried out, and different faults may occur:

[0134] The state of the grid side module is out of range, which is abnormal;

[0135] The state of the grid side module is out of range, which is abnormal;

[0136] The state of the cooling system is out of range, which is abnormal.

Claims

1. A method of operating condition outlier analysis for a frequency converter, characterized by, The method comprises the following steps: Step 1, get the original monitoring data of the actual operation of a single frequency converter in the SCADA system, denoted as matrix ; Step 2, clean the data in the matrix obtained in step 1 to obtain a matrix ; Step 3, segmenting the matrix obtained in Step 2 to obtain L sub-matrices segmenting to obtain L sub-matrices Step 4, calculating the state vector value corresponding to each sub-matrix; Step 5, performing outlier analysis on the operating state of the frequency converter according to the state vector value; In step 4, the state vector value corresponding to each sub-matrix is calculated, and the specific method is: S401, calculating the distribution graph between the power and the machine-side current of the frequency converter, the power and the grid-side current, and the power and the water temperature under different wind speed conditions in each sub-matrix; S402, identifying the edge points of each distribution graph obtained in S401; S403, fitting the fitting ellipse corresponding to each distribution graph according to the obtained edge points; S404, calculating the parameter value of each fitting ellipse; S405, calculating the sum of the Euclidean distances between all distribution points on each distribution graph and the two foci of the corresponding fitting ellipse according to the obtained parameter value, to obtain the Euclidean distance vector corresponding to each distribution graph; S406, counting the distance distribution result corresponding to each Euclidean distance vector; S407, calculating the state value corresponding to each distribution graph according to the obtained distance distribution result; S408, calculating the state vector value corresponding to each sub-matrix according to the state value corresponding to each distribution graph; In S407, the state value corresponding to each distribution graph is calculated according to the obtained distance distribution result, and the specific method is: ； ； ； ； ； ； wherein, is the first is the state value between the power and the inverter machine side current in the wind speed interval; respectively represent the first are the respective pre-coefficients in the wind speed interval; ; is the sum of the Euclidean distances between each distribution point and the two focal points; is the focal length; is the major axis length.

2. The method of claim 1, wherein, In step 1, the matrix is n rows by five columns, with the five columns in order being unit wind speed, unit power, frequency converter machine-side current, frequency converter grid-side current, and frequency converter outlet water temperature.

3. The method of claim 1, wherein, In step 2, the data in the matrix obtained in step 1 is cleaned. The specific method is as follows: ​ The matrix The out-of-limit point data, outlier data and stacking point data in the data are cleaned.

4. The method of claim 1, wherein, In step 3, the matrix obtained in step 2 is segmented according to the first column of data to obtain sub-matrices; wherein the number of segments is calculated by : In the formula, is the first cleaned unit wind speed value in the first column vector; is the first cleaned unit wind speed value in the first column vector; is the first cleaned unit wind speed value in the first column vector.

5. The method of claim 1, wherein, In S408, the state vector value corresponding to each sub-matrix is calculated according to the state value corresponding to each distribution graph, and the specific method is: wherein, is the state vector value corresponding to each sub-matrix; respectively represent the state value of the power and the current on the motor side of the frequency converter, the power and the current on the grid side of the frequency converter, and the power and the outlet water temperature of the frequency converter in the wind speed interval, respectively. is the state value of the power and the current on the motor side of the frequency converter, the power and the current on the grid side of the frequency converter, and the power and the outlet water temperature of the frequency converter in the wind speed interval, respectively.

6. The method of claim 1, wherein, In step 5, the operating state of the frequency converter is analyzed according to the state vector value, and the specific method is: Machine side module state outlier, anomaly; , net side module state outlier, abnormal; Cooling system out of limits, abnormal wherein, respectively represent the first the state values between the power and the inverter machine-side current, the power and the inverter grid-side current, and the power and the inverter outlet water temperature in the wind speed interval.

7. A variable frequency drive operating condition outlier analysis system, comprising: The analysis method based on claim 1 comprises: A data acquisition unit is configured to acquire original monitoring data of actual operation of a single frequency converter in a SCADA system, denoted as matrix ; A data cleaning unit is configured to clean data in the obtained matrix to obtain a matrix ; a data segmenting unit for segmenting the obtained matrix to obtain L sub-matrices; A data calculation unit for calculating the state vector value corresponding to each sub-matrix; A data analysis unit for performing outlier analysis on the operating state of the frequency converter according to the state vector value.

Images