Collector state evaluation method based on support vector data descriptor

Abstract

The invention provides an acquisition terminal state evaluation method based on a support vector data descriptor, and relates to the field of electricity utilization information acquisition terminal state evaluation, and the method comprises the steps: obtaining five types of nine electric energy quality measurement indexes of a concentrator, and carrying out the normalization processing of the five types of nine electric energy quality measurement indexes; the method comprises the following steps: constructing a first hyper-sphere by using 9 normalized power quality measurement indexes of 5 classes; setting a scaling ratio, and scaling the first hyper-sphere to obtain a second hyper-sphere; introducing an abnormal sample set, iteratively correcting the second hyper-sphere based on a support vector data descriptor, and constructing a third hyper-sphere; and according to the positions of 9-dimensional data in the first hyper-sphere and the third hyper-sphere after normalization of the measured value of the acquisition concentrator, evaluating the state of the used acquisition device. According to the invention, the problem of image-based automatic judgment of the state of the collection concentrator is solved.

Description

Technical Field

[0001] The invention belongs to the field of power consumption information collection terminal state evaluation, and in particular relates to a power consumption information collection terminal state evaluation method based on a support vector data descriptor. Background Art

[0002] With the rapid development of new power systems, the dual-high characteristics of power distribution networks have become increasingly obvious, bringing new challenges to the collection of power consumption information. As a key device for realizing the transmission of power consumption information, the current method for status evaluation of power consumption information collection concentrators often judges the working status of the concentrator by monitoring single operating data such as voltage, current, and frequency, and mainly relies on preset thresholds and simple rules to judge the status of the concentrator. Although this type of method can effectively detect serious faults of the concentrator, it ignores the inherent coupling of various operating data, and it is difficult to discover complex subtle fault modes or abnormal conditions based on single data status evaluation.

[0003] To address this issue, power systems in some regions have introduced neural networks, which have, to a certain extent, achieved the application of the inherent correlation of data and partially improved the accuracy and adaptability of state assessment. However, due to the uninterpretability of neural networks and the large demand for computing resources, they are not suitable for fast computing and edge computing, and it is difficult to meet the technical requirements of existing concentrator state assessment. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, the present invention provides a method for evaluating the state of a collection concentrator based on a support vector data descriptor, which solves the problem of automatic judgment of the state of a collection concentrator based on an image.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is: a method for evaluating the state of a data acquisition device based on a support vector data descriptor, comprising the following steps:

[0006] S1. Obtain 9 power quality measurement indicators of 5 categories of the concentrator and perform normalization processing on them;

[0007] S2, constructing a first hypersphere using the normalized 9 power quality measurement indicators of 5 categories;

[0008] S3, setting a scaling ratio, scaling the first hypersphere to obtain a second hypersphere;

[0009] S4, introducing an abnormal sample set, and iteratively correcting the second hypersphere based on the support vector data descriptor to construct a third hypersphere;

[0010] S5. Evaluate the state of the collection concentrator according to the positions where the normalized 9-dimensional data of the measurement values ​​of the collection concentrator fall in the first hypersphere and the third hypersphere.

[0011] The beneficial effects of the present invention are as follows: the present invention uses the 9 power quality measurement indicators of the concentrator as the source data for state evaluation, and preliminarily constructs the first hypersphere by normalizing the data and limiting the measurement indicators. On this basis, the abnormal sample set of the collection concentrator whose measurement data does not exceed the limit is used as the iterative data, and further iterates on the basis of the preliminarily constructed first hypersphere to obtain the optimized third hypersphere. The state of the collection concentrator can be evaluated according to the relationship between the position of the 9-dimensional data in the space after normalization of the collection concentrator measurement value and each hypersphere. The present invention utilizes various types of power quality measurement data of various concentrators, and fully considers the coupling relationship between various types of data. At the same time, various power quality indicators are used to optimize the initial value of the support vector data descriptor, which reduces the iterative calculation amount and also avoids the occurrence of abnormal solutions during the iteration process. A large amount of computing power is only required when training the support vector data descriptor. After the construction of each hypersphere is completed based on the support vector data descriptor, the collected power quality measurement data only needs to be normalized in a relatively simple way, and the relationship between its coordinate space and each hypersphere can be compared to determine the concentrator status. The computing power requirement is low, the hypersphere construction can be completed in the cloud, and the concentrator status can be determined on the edge computing device. It is suitable for the construction of a new type of collection system application combining "cloud-edge".

[0012] Furthermore, the expression of the normalization process in S1 is as follows:

[0013]

[0014]

[0015] in, and The normalized results of voltage effective value, current effective value, frequency, three-phase current zero-sequence unbalance, three-phase current negative-sequence unbalance, three-phase voltage zero-sequence unbalance, three-phase voltage negative-sequence unbalance, 2-21st current total harmonic distortion rate and 2-21st voltage total harmonic distortion rate are respectively, i represents the i-th hypersphere, k represents the element of the k-th sample in the sample set, U RMS_M ,I RMS_M 、f M ,I 0unb_M ,I 2unb_M , U 0unb_M , U 2unb_M ,THDi M and THDu MRespectively represent the measured values ​​of voltage effective value, current effective value, frequency, three-phase current zero-sequence unbalance, three-phase current negative-sequence unbalance, three-phase voltage zero-sequence unbalance, three-phase voltage negative-sequence unbalance, 2nd to 21st current total harmonic distortion rate and 2nd to 21st voltage total harmonic distortion rate, U RMS_N ,I RMS_N ,I 0unb_N ,I 2unb_N , U 0unb_N , U 2unb_N ,THDi N and THDu N They represent the true values ​​of voltage RMS, current RMS, three-phase current zero-sequence unbalance, three-phase current negative-sequence unbalance, three-phase voltage zero-sequence unbalance, three-phase voltage negative-sequence unbalance, 2-21th current total harmonic distortion and 2-21th voltage total harmonic distortion, τ represents the relaxation coefficient, e represents the true value of 1_max 、e 2_max 、e 3_max 、e 4_max 、e 5_max 、e 6_max 、e 7_max 、e 8_max and e 9_max They respectively represent the error limits of voltage RMS, current RMS, frequency, three-phase current zero-sequence unbalance, three-phase current negative-sequence unbalance, three-phase voltage zero-sequence unbalance, three-phase voltage negative-sequence unbalance, 2nd to 21st current total harmonic distortion rate, and 2nd to 21st voltage total harmonic distortion rate.

[0016] f + , THDi + and THDu + They respectively represent the maximum allowable upper limits of the measured values ​​of voltage effective value, current effective value, frequency, three-phase current zero-sequence unbalance, three-phase current negative-sequence unbalance, three-phase voltage zero-sequence unbalance, three-phase voltage negative-sequence unbalance, 2nd to 21st current total harmonic distortion rate, and 2nd to 21st voltage total harmonic distortion rate. f - , THDi - and THDu - They respectively represent the minimum allowable lower limits of the measured values ​​of the effective value of voltage, effective value of current, frequency, three-phase current zero-sequence unbalance, three-phase current negative-sequence unbalance, three-phase voltage zero-sequence unbalance, three-phase voltage negative-sequence unbalance, 2nd to 21st current total harmonic distortion rate and 2nd to 21st voltage total harmonic distortion rate.

[0017] The beneficial effect of the above further scheme is that after normalization according to the above formula, the data expression can be enhanced, the calculation process can be made more accurate, and the influence between data of different magnitudes can be unified.

[0018] Furthermore, the S2 is specifically:

[0019] Generate a data set for constructing a first hypersphere according to the normalized 9 power quality measurement indicators of 5 categories;

[0020] Using the data set of the first hypersphere, construct the first hypersphere:

[0021] if

[0022] Where R1 represents the radius of the first hypersphere, c1 represents the center of the first hypersphere, Representation sample The distance to the center c1 of the first hypersphere, i1 represents the i1th data in the data set.

[0023] The beneficial effect of the above further scheme is that the first hypersphere is constructed to locate the multidimensional coordinate system for calculation, giving the entire SVDD an initial input, and locating the center of the hypersphere at zero reduces the complexity of the calculation.

[0024] Furthermore, the expression of the second hypersphere is as follows:

[0025] c2=(0,0,0,0,0,0,0,0,0),R1=η

[0026]

[0027] Where c2 represents the center of the second hypersphere, R1 represents the radius of the first hypersphere, η represents the calculated radius of the first hypersphere, and x (i') Represents the independent variable in the i'th dimension in the multidimensional coordinate system, i'=1,2,...,9.

[0028] The beneficial effect of the above further scheme is: the second hypersphere in the present invention is iterated using normal data, the hypersphere is roughly constructed based on the first hypersphere, the radius and center of the hypersphere are iteratively calculated to obtain a second hypersphere for judging normal data.

[0029] Furthermore, the S4 is specifically:

[0030] Obtain an abnormal sample set containing 9 items of power quality test data in 5 categories that do not exceed the maximum limit and have abnormalities at the acquisition terminal;

[0031] Normalizing the abnormal sample set to obtain a sample set for iteratively correcting the second hypersphere, wherein when performing the normalization process and selecting the relaxation variable, it should be ensured that the data in the sample set of the second hypersphere falls between the second hypersphere and the first hypersphere to the maximum extent;

[0032] Using the sample set of the second hypersphere, construct a third hypersphere;

[0033] Iterative calculation is performed on the third hypersphere to obtain a revised third hypersphere, thereby completing the construction of the third hypersphere.

[0034] The beneficial effect of the above further scheme is that the third hypersphere in the present invention is iterated using a data set with abnormal data, and is finely constructed based on the second hypersphere, the radius and center of the third hypersphere are fine-tuned, and the edge of the third hypersphere is further refined to obtain a third hypersphere that can judge both normal data and abnormal data. If one or two hyperspheres are constructed less, the difficulty of calculation will be greatly increased, the required computing resources will also be greatly increased, and the simultaneous input of training data increases the possibility of data contradictions in the calculation process, which will increase the risk of calculation crash by about 25%.

[0035] Furthermore, the expression of the third hypersphere is as follows:

[0036] if

[0037] Where R3 represents the radius of the third hypersphere, c3 represents the center of the third hypersphere, k2 represents the data number in the second hypersphere, C represents the elastic coefficient, ξ i”' represents the deviation allowed for the ith sample, n2 represents the number of data sets whose sample clusters of the second hypersphere fall between the second hypersphere and the first hypersphere, represents the data input to the second hypersphere iteration formula, that is, the abnormal sample data set, Represents a slack variable.

[0038] Furthermore, the expression for iterative calculation of the third hypersphere is as follows:

[0039]

[0040] Wherein, k2 and k3 both represent the data numbers in the second hypersphere, n2 represents the number of data sets whose sample clusters in the second hypersphere fall between the second hypersphere and the first hypersphere, and They all represent the coefficients in the iterative process, and They are respectively represented as the k2th data and the k3th data in the second hypersphere, and C represents the intermediate parameter used to solve the optimization iterative process.

[0041] The beneficial effect of the above further scheme is that the third hypersphere constructed by the present invention can be added with new data for iterative training in future actual production work. In the process of continuous iteration with the addition of new data, the hypersphere for judging abnormal data will be continuously improved. Compared with other Deep-SVDDs, the introduction of new data requires re-combination with historical data for calculation, which requires a lot of time and computing power. The algorithm proposed by the present invention only needs to input new data for iteration, which greatly reduces the calculation time and computing power.

[0042] Furthermore, the S5 is specifically:

[0043] Collect the normalized 9-dimensional data of the concentrator measurement value, wherein the normalized 9-dimensional data is the normalized 9-item power quality measurement data of 5 categories;

[0044] In response to the normalized 9-dimensional data falling into the inner area of ​​the third hypersphere, that is, Then the collection concentrator is in a normal state, where R3 represents the radius of the third hypersphere, and c3 represents the center of the third hypersphere;

[0045] In response to the normalized 9-dimensional data falling within the inner region of the first hypersphere and the outer region of the third hypersphere, that is, Then the collection concentrator is in an abnormal state, where R1 represents the radius of the first hypersphere;

[0046] In response to the normalized 9-dimensional data falling into the outer area of ​​the first hypersphere, that is, The collection concentrator is in a fault state. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 The figure is a flow chart of the method of the present invention.

[0048] Figure 2 It is a projection diagram of the first hypersphere S1 in three-dimensional coordinates in this embodiment.

[0049] Figure 3 3 is a projection diagram of the second hypersphere S2 in this embodiment under three-dimensional coordinates.

[0050] Figure 4 FIG. 4 is a schematic diagram of the third hypersphere S3 in this embodiment.

[0051] Figure 5 Schematic diagram of the state of the collector used in this embodiment. DETAILED DESCRIPTION

[0052] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0053] Example 1

[0054] like Figure 1 As shown, the present invention provides a method for evaluating the state of a data acquisition device based on a support vector data descriptor, and the implementation method thereof is as follows:

[0055] S1. Obtain 9 power quality measurement indicators of 5 categories of the concentrator and perform normalization processing on them;

[0056] S2. Using the normalized 5 categories and 9 power quality measurement indicators, a first hypersphere is constructed, which is specifically:

[0057] Generate a data set for constructing a first hypersphere according to the normalized 9 power quality measurement indicators of 5 categories;

[0058] Using the data set of the first hypersphere, construct the first hypersphere;

[0059] S3, setting a scaling ratio, scaling the first hypersphere to obtain a second hypersphere;

[0060] S4, introducing an abnormal sample set, and iteratively correcting the second hypersphere based on the support vector data descriptor to construct a third hypersphere, which is specifically as follows:

[0061] Obtain an abnormal sample set containing 9 items of power quality test data in 5 categories that do not exceed the maximum limit and have abnormalities at the acquisition terminal;

[0062] Normalizing the abnormal sample set to obtain a sample set for iteratively correcting the second hypersphere, wherein when performing the normalization process and selecting the relaxation variable, it should be ensured that the data in the sample set of the second hypersphere falls between the second hypersphere and the first hypersphere to the maximum extent;

[0063] Using the sample set of the second hypersphere, construct a third hypersphere;

[0064] Iteratively calculate the third hypersphere to obtain a revised third hypersphere, thereby completing the construction of the third hypersphere;

[0065] S5. According to the positions of the normalized 9-dimensional data of the acquisition concentrator measurement value falling in the first hypersphere and the third hypersphere, the status of the acquisition concentrator is evaluated, which is specifically:

[0066] Collect the normalized 9-dimensional data of the concentrator measurement value, wherein the normalized 9-dimensional data is the normalized 9-item power quality measurement data of 5 categories;

[0067] In response to the normalized 9-dimensional data falling into the inner area of ​​the third hypersphere, that is, Then the collection concentrator is in a normal state, where R3 represents the radius of the third hypersphere, and c3 represents the center of the third hypersphere;

[0068] In response to the normalized 9-dimensional data falling within the inner region of the first hypersphere and the outer region of the third hypersphere, that is, Then the collection concentrator is in an abnormal state, where R1 represents the radius of the first hypersphere;

[0069] In response to the normalized 9-dimensional data falling into the outer area of ​​the first hypersphere, that is, The collection concentrator is in a fault state.

[0070] In this embodiment, the present invention uses 5 categories and 9 power quality measurement indicators of the concentrator as the source data for status evaluation, and evaluates the status of the concentrator into three categories: normal, abnormal, and fault. First, a nine-dimensional first hypersphere with the origin as the core is constructed through the standard values ​​of the 9 power quality measurement data, and the first hypersphere is used as the boundary to distinguish faults and non-faults (abnormal, normal); secondly, the scaling ratio is set, and the first hypersphere is scaled to obtain the second hypersphere, which is the original boundary between normal and abnormal; then the abnormal sample set D is introduced, and the abnormal sample set D is a concentrator with non-fault power quality measurement indicators but actually has defects. The abnormal sample set D is cleaned and the data items falling in the outer area of ​​the first hypersphere and the inner area of ​​the third hypersphere are eliminated; finally, the second hypersphere is used as the initial solution, and the abnormal sample set D is modified and iterated based on the support vector data descriptor (SVDD) to obtain the third hypersphere. The outer area of ​​the first hypersphere is the fault space, the inner area of ​​the first hypersphere and the area between the third hypersphere are the abnormal space, and the inner area of ​​the third hypersphere is the normal space. The state of the collection concentrator is determined by the position in the space where the normalized 9-dimensional data of the collection concentrator measurement value falls.

[0071] In this embodiment, the performance data of the concentrator evaluated in the present invention are: 1. The effective value of voltage U RMS ; 2. Current effective value I RMS ; 3. Frequency f; 4. Three-phase current zero-sequence unbalance I0 unb ; 5. Three-phase current negative sequence imbalance I 2_unb ; 6. Three-phase voltage zero sequence unbalance U 0_unb ; 7. Three-phase voltage negative sequence unbalance U 2_unb; 8. 2nd to 21st total harmonic distortion rate of current THDi; 9. 2nd to 21st total harmonic distortion rate of voltage THDu. The performance data of the concentrator must be normalized before being used to construct the first hypersphere. The above 9 types of data can be normalized to X in turn. i_1 ~X i_9 , the formula is shown in Formula 1.

[0072]

[0073] e 1_max ~e 9_max As shown in formula (2):

[0074]

[0075] in, and The normalized results of voltage effective value, current effective value, frequency, three-phase current zero-sequence unbalance, three-phase current negative-sequence unbalance, three-phase voltage zero-sequence unbalance, three-phase voltage negative-sequence unbalance, 2-21st current total harmonic distortion rate and 2-21st voltage total harmonic distortion rate are respectively, i represents the i-th hypersphere, k represents the element of the k-th sample in the sample set, U RMS_M ,I RMS_M 、f M ,I 0unb_M ,I 2unb_M , U 0unb_M , U 2unb_M ,THDi M and THDu M Respectively represent the measured values ​​of voltage effective value, current effective value, frequency, three-phase current zero-sequence unbalance, three-phase current negative-sequence unbalance, three-phase voltage zero-sequence unbalance, three-phase voltage negative-sequence unbalance, 2nd to 21st current total harmonic distortion rate and 2nd to 21st voltage total harmonic distortion rate, U RMS_N ,I RMS_N ,I 0unb_N ,I 2unb_N , U 0unb_N , U 2unb_N ,THDi N and THDu N They represent the true values ​​of voltage RMS, current RMS, three-phase current zero-sequence unbalance, three-phase current negative-sequence unbalance, three-phase voltage zero-sequence unbalance, three-phase voltage negative-sequence unbalance, 2-21th current total harmonic distortion and 2-21th voltage total harmonic distortion, τ represents the relaxation coefficient, e represents the true value of 1_max 、e 2_max 、e 3_max 、e 4_max 、e 5_max 、e 6_max 、e7_max 、e 8_max and e 9_max They respectively represent the error limits of voltage RMS, current RMS, frequency, three-phase current zero-sequence unbalance, three-phase current negative-sequence unbalance, three-phase voltage zero-sequence unbalance, three-phase voltage negative-sequence unbalance, 2nd to 21st current total harmonic distortion rate, and 2nd to 21st voltage total harmonic distortion rate. f + , THDi + and THDu + They respectively represent the maximum allowable upper limits of the measured values ​​of voltage effective value, current effective value, frequency, three-phase current zero-sequence unbalance, three-phase current negative-sequence unbalance, three-phase voltage zero-sequence unbalance, three-phase voltage negative-sequence unbalance, 2nd to 21st current total harmonic distortion rate, and 2nd to 21st voltage total harmonic distortion rate. f - , THDi - and THDu - They respectively represent the minimum allowable lower limits of the measured values ​​of the effective value of voltage, effective value of current, frequency, three-phase current zero-sequence unbalance, three-phase current negative-sequence unbalance, three-phase voltage zero-sequence unbalance, three-phase voltage negative-sequence unbalance, 2nd to 21st current total harmonic distortion rate and 2nd to 21st voltage total harmonic distortion rate.

[0076] In this embodiment, the above data are normalized to obtain a data set for constructing a first hypersphere. The data X used by the first hypersphere is 1_1 ,X 1_2 ,X 1_3 ,X 1_4 ,X 1_5 ,X 1_6 ,X 1_7 ,X 1_8 ,X 1_9 U RMS_M ,I RMS_M 、f M ,I 0unb_M ,I 2unb_M , U 0unb_M , U 2unb_M ,THDi M ,THDu M The values ​​are all allowed limits, that is, the one between the maximum allowable upper limit and the minimum allowable lower limit that has a larger absolute value of the measured value deviation. The slack variable τ is 0, and this is substituted into X 1_1 ,X 1_2 ,X 1_3 ,X 1_4 ,X 1_5 ,X 1_6 ,X1_7 ,X 1_8 ,X 1_9 are all 1, and this is used as the boundary length of each dimension to construct the data set Y1 for constructing the first hypersphere, Y1={x1,x2,x3,...,xn1}:

[0077] Among them, x i1 ∈R 9 (i.e. x i1 is a 9-dimensional real vector), n1=18, x1=(1,0,0,0,0,0,0,0,0), x2=(-1,0,0,0,0,0,0,0,0), x3=(0,1,0,0,0,0,0,0,0), x4=(0,-1,0,0,0,0,0,0,0), x5=(0,0,1,0,0,0,0,0,0), x6=(0,0,-1,0,0,0,0,0,0), ..., x 17 =(0,0,0,0,0,0,0,0,1), x 18 =(0,0,0,0,0,0,0,0,-1).

[0078] Then the first hypersphere can be constructed as:

[0079] if

[0080] Where R1 represents the radius of the first hypersphere, c1 represents the center of the first hypersphere, Representation sample The distance to the center c1 of the first hypersphere, i1 represents the i1th data in the data set.

[0081] The first hypersphere can be solved:

[0082] c1=(0,0,0,0,0,0,0,0,0),R1=1 (4)

[0083]

[0084] Among them, x (i') Represents the independent variable in the i'th dimension in the multidimensional coordinate system.

[0085] In this embodiment, the first hypersphere is scaled to obtain the second hypersphere, and the scaling ratio β is in the range of (0.5, 0.6), and the second hypersphere can be obtained:

[0086] c2=(0,0,0,0,0,0,0,0,0), R1=η(6)

[0087]

[0088] Where c2 represents the center of the second hypersphere, R1 represents the radius of the first hypersphere, η represents the calculated radius of the first hypersphere, and x (i') Represents the independent variable in the i'th dimension in the multidimensional coordinate system, i'=1,2,...,9.

[0089] In this embodiment, by means of data screening, an abnormal data set containing 9 items of power quality test data that do not exceed the maximum limit but the acquisition terminal has an abnormality is collected. in, It means that the i-th sample contains 9 types of power quality test data values, that is, d i is a 9-dimensional vector, n2′ represents the number of samples contained in the data set D, Represents the set of real numbers.

[0090] After normalizing the abnormal data set D according to formula (2), we get the sample set Y3 used to iteratively correct the second hypersphere. The slack variable τ should be appropriately adjusted to ensure that as many points as possible in the sample set Y3 fall between the second hypersphere and the first hypersphere. n2 is the number of data sets in the sample set Y3 that fall between the second hypersphere and the first hypersphere.

[0091] Y3={x1′,x2′,x3′,...,x n2 ′}(8)

[0092] Among them, the k1th sample 1≤k1≤n2, where k1 represents the data number in the second hypersphere.

[0093] The hypersphere that can be constructed is:

[0094] if

[0095] Where R3 represents the radius of the third hypersphere, c3 represents the center of the third hypersphere, ξ represents the relaxation variable, k2 represents the data number in the second hypersphere, ξ i”' represents the deviation allowed for the ith sample, n2 represents the number of data sets whose sample clusters of the second hypersphere fall between the second hypersphere and the first hypersphere, represents the data input to the second hypersphere iteration formula, that is, the abnormal sample data set, represents the relaxation variable, C is the elastic coefficient, and its value range is (0,1).

[0096] In the initial value setting of the iterative calculation, R3=R2, c3=(0,0,0,0,0,0,0,0,0). ξ i”'It represents the allowable deviation of the i'th sample. Usually, the allowable deviation of the sample is 2%.

[0097] Then the Lagrange multiplier is introduced and Construct the Lagrangian function:

[0098]

[0099] in, and They all represent the coefficients in the iterative process.

[0100] The Lagrangian function is respectively for R3, c3, Take the derivative and set it to zero:

[0101] Taking the derivative of R3:

[0102]

[0103] Taking the derivative of c3:

[0104]

[0105] Yes i To perform the derivation:

[0106]

[0107] Then introduce the Carlo-Kuhn-Tucker condition KKT, and construct the constraint condition as follows:

[0108]

[0109] In summary, the entire process of iterating the third hypersphere can be simplified to the following optimization problem:

[0110]

[0111] Wherein, k2 and k3 both represent the data numbers in the second hypersphere, n2 represents the number of data sets whose sample clusters in the second hypersphere fall between the second hypersphere and the first hypersphere, and They all represent the coefficients in the iterative process, and They are respectively represented as the k2th data and the k3th data in the second hypersphere, and C represents the intermediate parameter used to solve the optimization iterative process.

[0112] By solving this optimization problem, we can get the radius R3 and center c3 of the modified third hypersphere. The inclusion relationship of the three hyperspheres is that the first hypersphere contains the third hypersphere, and the third hypersphere contains the second hypersphere. Therefore, the state of the collection concentrator can be determined by the position of the normalized 9-dimensional data of the collection concentrator measurement value in space:

[0113] When the normalized 9-dimensional data falls within the inner region of the third hypersphere, that is, The collection concentrator is in normal state, in which the concentrator functions normally and the measured value does not exceed the limit;

[0114] When the normalized 9-dimensional data falls inside the first hypersphere and outside the third hypersphere, that is, The collection concentrator is in an abnormal state. In this state, the concentrator functions abnormally but the measured value does not exceed the limit;

[0115] When the normalized 9-dimensional data falls outside the first hypersphere, that is, The acquisition concentrator is in a fault state. In this state, the concentrator function fails and the measured value exceeds the limit.

[0116] That is: abnormal state: the area inside the first hypersphere and outside the third hypersphere; fault state: the area outside the first hypersphere, that is By calculating the position of the normalized 9-dimensional data in space, we can get a formula for determining the state of the collector:

[0117]

[0118] In summary, the present invention utilizes various types of power quality measurement data of various types of concentrators, and fully considers the coupling relationship between various types of data. At the same time, it optimizes the initial value of the support vector data descriptor using various power quality indicators, reduces the amount of iterative calculations, and also avoids abnormal solutions during the iteration process. Compared with the closest prior art in the third article, this method only requires a large amount of computing power when training the support vector data descriptor. After the construction of each hypersphere is completed based on the support vector data descriptor, the collected power quality measurement data only needs to be normalized relatively simply, and the relationship between its coordinate space and each hypersphere can be compared to determine the state of the concentrator. The computing power requirement is relatively low, the hypersphere construction can be completed in the cloud, and the concentrator state can be determined on the edge computing device. It is suitable for the construction of a new acquisition system application combining "cloud-edge", and the entire method process is explainable, and there is no black box model.

[0119] Example 2

[0120] To facilitate understanding of the present invention, the technical points in the above method are demonstrated by taking the 24-version acquisition terminal as an example, as follows:

[0121] S1: The maximum allowable upper and lower limits of the 9 types of power quality measurement values ​​of the 24-version acquisition terminal are:

[0122]

[0123] According to the above method, we can obtain:

[0124]

[0125] This is used as the reference value to normalize various measurement values ​​of the acquisition terminal.

[0126] S2: Construct the hypersphere S1, based on the above method, RMS_M ,I RMS_M 、f M ,I 0unb_M ,I 2unb_M , U 0unb_M , U 2unb_M ,THDi M ,THDu M The values ​​are all allowed limits, and the slack variable τ is 0. Substituting this into X 1_1 ,X 1_2 ,X 1_3 ,X 1_4 ,X 1_5 ,X 1_6 ,X 1_7 ,X 1_8 ,X 1_9 are all 1, and this is used as the boundary length of each dimension to construct the data set Y1 for constructing the first hypersphere. The first hypersphere can be solved by the method in S2 above:

[0127]

[0128] Since the nine-dimensional hypersphere cannot be displayed, the first hypersphere is in (x (1) ,x (2) ,x (3) ) The projection in three-dimensional coordinates is as follows Figure 2 shown.

[0129] S3: Scaling the first hypersphere to obtain the second hypersphere, the scaling ratio is 0.5, and the second hypersphere is obtained:

[0130]

[0131] The second hypersphere (x (1) ,x (2) ,x (3) ) The projection in three-dimensional coordinates is as follows Figure 3 shown.

[0132] S4: 163 data groups with measured data that do not exceed the maximum limit are selected from the 24th version of the acquisition terminal with abnormal functions by data screening, and based on this, an abnormal data set D = {d1, d2, ..., d 163}, the slack variable τ is set to 0.03, and after normalization, there are 161 groups of data items in Y3 that fall between the first hypersphere and the second hypersphere.

[0133] Y3={x1′,x2′,x3′,...,x 161 '}(twenty one)

[0134] Using this sample set as iterative data, the third hypersphere can be iterated based on the first hypersphere, such as Figure 4 shown.

[0135] like Figure 5 As shown, when the normalized 9-dimensional data of the acquisition concentrator measurement value falls within the third hypersphere, the state of the acquisition concentrator is in a normal state; when the normalized 9-dimensional data of the acquisition concentrator measurement value falls between the third hypersphere and the first hypersphere, the state of the acquisition concentrator is in an abnormal state; when the normalized 9-dimensional data of the acquisition concentrator measurement value falls outside the first hypersphere, the state of the acquisition concentrator is in a fault state, wherein, Figures 2 to 5 In the above, S1 represents the first hypersphere, S2 represents the second hypersphere, and S3 represents the third hypersphere.

Claims

1. A method for evaluating the state of a data acquisition device based on a support vector data descriptor, characterized in that: The following steps are involved: S1. Obtain 9 power quality measurement indicators of 5 categories of the concentrator and perform normalization processing on them; S2, constructing a first hypersphere using the normalized 9 power quality measurement indicators of 5 categories; S3, setting a scaling ratio, scaling the first hypersphere to obtain a second hypersphere; S4, introducing an abnormal sample set, and iteratively correcting the second hypersphere based on the support vector data descriptor to construct a third hypersphere; S5. Evaluate the state of the collection concentrator according to the positions where the normalized 9-dimensional data of the measurement values ​​of the collection concentrator fall in the first hypersphere and the third hypersphere.

2. The method for evaluating the state of a data acquisition device based on support vector data descriptor according to claim 1, characterized in that: The expression of the normalization process in S1 is as follows: in, and The normalized results of voltage effective value, current effective value, frequency, three-phase current zero-sequence unbalance, three-phase current negative-sequence unbalance, three-phase voltage zero-sequence unbalance, three-phase voltage negative-sequence unbalance, 2-21st current total harmonic distortion rate and 2-21st voltage total harmonic distortion rate are respectively, i represents the i-th hypersphere, k represents the element of the k-th sample in the sample set, U RMS_M ,I RMS_M 、f M ,I 0unb_M ,I 2unb_M , U 0unb_M , U 2unb_M ,THDi M and THDu M Respectively represent the measured values ​​of voltage effective value, current effective value, frequency, three-phase current zero-sequence unbalance, three-phase current negative-sequence unbalance, three-phase voltage zero-sequence unbalance, three-phase voltage negative-sequence unbalance, 2nd to 21st current total harmonic distortion rate and 2nd to 21st voltage total harmonic distortion rate, U RMS_N ,I RMS_N ,I 0unb_N ,I 2unb_N , U 0unb_N , U 2unb_N ,THDi N and THDu N They represent the true values ​​of voltage RMS, current RMS, three-phase current zero-sequence unbalance, three-phase current negative-sequence unbalance, three-phase voltage zero-sequence unbalance, three-phase voltage negative-sequence unbalance, 2-21th current total harmonic distortion and 2-21th voltage total harmonic distortion, τ represents the relaxation coefficient, e represents the true value of 1_max 、e 2_max 、e 3_max 、e 4_max 、e 5_max 、e 6_max 、e 7_max 、e 8_max and e 9_max They respectively represent the error limits of voltage RMS, current RMS, frequency, three-phase current zero-sequence unbalance, three-phase current negative-sequence unbalance, three-phase voltage zero-sequence unbalance, three-phase voltage negative-sequence unbalance, 2nd to 21st current total harmonic distortion rate, and 2nd to 21st voltage total harmonic distortion rate. THDi + and THDu + They respectively represent the maximum allowable upper limits of the measured values ​​of voltage effective value, current effective value, frequency, three-phase current zero-sequence unbalance, three-phase current negative-sequence unbalance, three-phase voltage zero-sequence unbalance, three-phase voltage negative-sequence unbalance, 2nd to 21st current total harmonic distortion rate, and 2nd to 21st voltage total harmonic distortion rate. THDi - and THDu - They respectively represent the minimum allowable lower limits of the measured values ​​of the effective value of voltage, effective value of current, frequency, three-phase current zero-sequence unbalance, three-phase current negative-sequence unbalance, three-phase voltage zero-sequence unbalance, three-phase voltage negative-sequence unbalance, 2nd to 21st current total harmonic distortion rate and 2nd to 21st voltage total harmonic distortion rate.

3. The method for evaluating the state of a data acquisition device based on support vector data descriptor according to claim 1, characterized in that: The S2 is specifically: Generate a data set for constructing a first hypersphere according to the normalized 9 power quality measurement indicators of 5 categories; Using the data set of the first hypersphere, construct the first hypersphere: if Where R1 represents the radius of the first hypersphere, c1 represents the center of the first hypersphere, Representation sample The distance to the center c1 of the first hypersphere, i1 represents the i1th data in the data set.

4. The method for evaluating the state of a data acquisition device based on support vector data descriptor according to claim 1, characterized in that: The expression of the second hypersphere is as follows: c2=(0,0,0,0,0,0,0,0,0),R1=η Where c2 represents the center of the second hypersphere, R1 represents the radius of the first hypersphere, η represents the calculated radius of the first hypersphere, and x (i') Represents the independent variable in the i'th dimension in the multidimensional coordinate system, i'=1,2,...,9.

5. The method for evaluating the state of a data acquisition device based on support vector data descriptor according to claim 1, characterized in that: The S4 is specifically: Obtain an abnormal sample set containing 9 items of power quality test data in 5 categories that do not exceed the maximum limit and have abnormalities at the acquisition terminal; Normalizing the abnormal sample set to obtain a sample set for iteratively correcting the second hypersphere, wherein when performing the normalization process and selecting the relaxation variable, it should be ensured that the data in the sample set of the second hypersphere falls between the second hypersphere and the first hypersphere to the maximum extent; Using the sample set of the second hypersphere, construct a third hypersphere; Iterative calculation is performed on the third hypersphere to obtain a revised third hypersphere, thereby completing the construction of the third hypersphere.

6. The method for evaluating the state of a data acquisition device based on support vector data descriptor according to claim 5, characterized in that: The expression of the third hypersphere is as follows: Where R3 represents the radius of the third hypersphere, c3 represents the center of the third hypersphere, k2 represents the data number in the second hypersphere, C represents the elastic coefficient, ξ i”' represents the deviation allowed for the ith sample, n2 represents the number of data sets whose sample clusters of the second hypersphere fall between the second hypersphere and the first hypersphere, represents the data input to the second hypersphere iteration formula, that is, the abnormal sample data set, Represents a slack variable.

7. The method for evaluating the state of a data acquisition device based on support vector data descriptor according to claim 6 is characterized in that: The expression for iterative calculation of the third hypersphere is as follows: Wherein, k2 and k3 both represent the data numbers in the second hypersphere, n2 represents the number of data sets whose sample clusters in the second hypersphere fall between the second hypersphere and the first hypersphere, and They all represent the coefficients in the iterative process, and They are respectively represented as the k2th data and the k3th data in the second hypersphere, and C represents the intermediate parameter used to solve the optimization iterative process.

8. The method for evaluating the state of a data acquisition device based on support vector data descriptor according to claim 1, characterized in that: The S5 is specifically: Collect the normalized 9-dimensional data of the concentrator measurement value, wherein the normalized 9-dimensional data is the normalized 9-item power quality measurement data of 5 categories; In response to the normalized 9-dimensional data falling into the inner area of ​​the third hypersphere, that is, Then the collection concentrator is in a normal state, where R3 represents the radius of the third hypersphere, and c3 represents the center of the third hypersphere; In response to the normalized 9-dimensional data falling within the inner region of the first hypersphere and the outer region of the third hypersphere, that is, Then the collection concentrator is in an abnormal state, where R1 represents the radius of the first hypersphere; In response to the normalized 9-dimensional data falling into the outer area of ​​the first hypersphere, that is, The collection concentrator is in a fault state.

Images