A method for identifying users with faulty neutral and ground wire connections in leaky radio areas based on ridge regression analysis.

Abstract

A method for identifying users with incorrect neutral and ground wire wiring in low-voltage transformer substations based on ridge regression analysis is proposed to address the problem that incorrect neutral and ground wire wiring by low-voltage residential users causes the load current to be converted into residual current in the substation, leading to frequent tripping of the primary residual current protection of the distribution transformer and preventing the activation of leakage protection, thus endangering electrical safety. This method identifies low-voltage substations where excessive residual current causes frequent tripping of the primary residual current protection of the distribution transformer as leakage substations. It obtains time-by-time residual current data for the leakage substation and time-by-time load current data for each user in the substation on the same day, and establishes a ridge regression analysis model to identify users with incorrect neutral and ground wire wiring. This method can accurately identify users with incorrect neutral and ground wire wiring in leakage substations, significantly accelerating the troubleshooting of leakage faults caused by user wiring errors in low-voltage substations and ensuring electrical safety.

Description

Technical Field

[0001] This invention belongs to the field of low-voltage distribution network leakage detection and analysis, and relates to a user identification method for leakage transformer areas with incorrect neutral and ground wire wiring. Specifically, it relates to a user identification method for leakage transformer areas with incorrect neutral and ground wire wiring based on ridge regression analysis, in order to identify users with incorrect neutral and ground wire wiring in low-voltage transformer areas where leakage faults cause the primary residual current protection of the transformer to trip, thereby accelerating the investigation of leakage points and restoring the primary residual current protection of the transformer to operation. Background Technology

[0002] In low-voltage distribution transformer areas, to protect personnel and line safety, a three-level residual current protection system is typically installed to achieve leakage current protection: primary residual current protection (i.e., main transformer leakage current protection), secondary residual current protection (i.e., branch line leakage current protection), and tertiary residual current protection (i.e., terminal household leakage current protection). The operating current values ​​and operating times of each level of residual current protection device are coordinated. The rated residual current operating value of the next lower level should not exceed that of the previous level, while the rated breaking time and non-starting time of the previous level must be greater than the rated breaking time and extreme non-starting time of the next lower level, thus achieving graded protection with selectable operation. Among the three-level leakage current protection in the transformer area, the tertiary residual current protection for terminal households is the most sensitive. However, in many cases, due to user-related reasons, the tertiary residual current protection is not correctly configured or activated. In such cases, if the secondary residual current protection of a branch line malfunctions, it may also fail to operate when a leakage current fault occurs, causing the primary residual current protection of the transformer to trip due to the user's leakage current fault. If leakage faults cannot be located and cleared in a timely manner, the primary residual current protection of the low-voltage distribution transformer will trip frequently, affecting users' normal power supply. When leakage faults cannot be accurately located and cleared, the primary residual current protection of the distribution transformer is often forced to be deactivated, causing low-voltage users in the distribution area to lose the last level of protection in the three-level leakage protection system to prevent electric shock accidents, thus creating a major hidden danger for personal injury accidents caused by leakage.

[0003] Residual current protection (RCD) trips based on the effective value of the residual current, failing to provide richer electrical quantity information for low-voltage distribution areas during RCD fault analysis and localization. New-generation smart meters and distribution transformer smart terminals based on the IR46 standard support monitoring functions beyond legal metering tasks. Only the corresponding functional modules need to be configured to achieve online monitoring and analysis of electrical quantities such as residual current. Due to cost constraints, currently only distribution transformer smart terminals typically install residual current monitoring modules, enabling online monitoring of the low-voltage side residual current at the distribution transformer level. Because of the large number of smart meters in low-voltage residential users, residual current monitoring modules are generally not configured in these meters. Currently, many low-voltage residential user meters are equipped with high-performance liquid crystal communication (HPLC) modules, enabling metering data reporting at 15-minute intervals, greatly enriching the available user electricity consumption data and laying the foundation for further in-depth analysis. Since there may be a potential correlation between the user's load current and the residual current in the transformer area when a user experiences a leakage current fault, it is possible to identify and locate the leakage current fault by mining and analyzing the residual current in the transformer area recorded by the smart distribution terminal and the user's electricity consumption data recorded by the user's HPLC smart meter.

[0004] Low-voltage residential users typically use single-phase power supply, with a neutral wire, live wire, and grounding protection wire connected indoors. The live wire is one of the three phases (A, B, and C). Incorrect neutral and ground wire connections are a common fault among low-voltage users. When a single-phase low-voltage user incorrectly connects the neutral and ground wires at the distribution box, the load current flowing in from the phase wires will flow out through the grounding protection wire instead of the designated neutral wire. In this case, the user's load current will be entirely converted into residual current for themselves and the entire distribution area. When the terminal household's three-level residual current protection and the branch line's two-level residual current protection fail, the residual current in the distribution area caused by incorrect neutral and ground wire connections will cause frequent trips of the distribution transformer's first-level residual current protection. To ensure power supply reliability, it is often necessary to disable the distribution transformer's first-level residual current protection. At this time, the terminal household's three-level residual current protection, the branch line's two-level residual current protection, and the distribution transformer's first-level residual current protection all lose their protective function, and users can only rely on the grounding protection wire for electrical safety. However, the long-term carrying of large load currents on the grounding protection conductor will accelerate the aging of the terminals, leading to an increase in grounding resistance. When a large load current is carried again, a significant potential difference will be created, causing the casings of all household appliances connected to the grounding protection conductor to become dangerously voltaged, seriously endangering the personal safety of users. If the load current burns out the grounding terminals of the grounding protection conductor, users connected to that grounding conductor will completely lose grounding protection, and the metal casings of electrical equipment connected to that grounding conductor will become energized, easily causing electric shock accidents and resulting in personal injury or death. Therefore, identifying and locating incorrect neutral and ground wire wiring in low-voltage distribution areas is of significant practical importance for ensuring safe electricity use in low-voltage distribution areas.

[0005] Since the primary residual current protection of a distribution transformer is generally set to operate at 300mA, the residual current in a low-voltage distribution transformer normally fluctuates below 300mA. When a user's neutral or ground wire is incorrectly grounded, the user's load current is converted into the transformer's residual current, causing the residual current to fluctuate significantly with changes in the user's load current. It's possible to identify abnormal users with incorrect neutral or ground wire connections based on the linkage between the transformer's residual current and the user's load current. It should be noted that low-voltage residential users have similar daily electricity consumption behaviors and load compositions. When a large number of users are connected to a transformer where a leakage fault occurs, multiple users' load current curves are likely to be highly similar. In this case, identifying users with incorrect neutral or ground wire connections based on the similarity between the transformer's residual current and the user's load current is susceptible to multicollinearity caused by similar load current data from multiple residential users. Severe multicollinearity can lead to misjudgment or missed detection of users with incorrect connections. Existing literature also uses criteria such as correlation between user load current data and residual current data in the transformer area, or Euclidean distance, or Pearson correlation coefficient to identify abnormal users, such as CN114114074A and CN113933585A. However, as the number of users increases, the collinearity problem between data also increases, and the screening may result in missed or false positives. In addition, if there are two or more users with incorrect neutral or ground wire wiring in the transformer area, the residual current of the transformer area will be the vector sum of the natural leakage current of the transformer area and the residual current of each user with incorrect wiring. At this time, the impact of collinearity becomes very significant, which will significantly increase the probability of false positives and missed negatives. Summary of the Invention

[0006] The purpose of this invention is to address the problems existing in the prior art by proposing a method for identifying users with incorrect neutral and ground wire wiring in leakage current distribution areas based on ridge regression analysis. This method employs a ridge regression analysis model to eliminate the effects of multicollinearity, identifying suspected users with incorrect neutral and ground wire wiring. By mining and analyzing data resources in the distribution area, the scope of leakage current fault investigation is narrowed, and the utilization rate of primary residual current protection in the distribution transformer is increased. This solves the problem of low-voltage distribution areas losing leakage current protection due to the inability to promptly detect leakage current faults, which can easily lead to electrical safety accidents.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: a method for identifying users with wiring errors in the neutral and ground wires of a leaky radio station based on ridge regression analysis, the steps of which are as follows:

[0008] Step 1: Identify the low-voltage transformer area where the excessive residual current causes frequent tripping of the primary residual current protection of the transformer as the leakage transformer area.

[0009] Since the primary residual current protection of a distribution transformer is generally set to operate at 300mA, the residual current in a low-voltage distribution transformer area will normally fluctuate below 300mA. Although low-voltage distribution transformer areas without leakage faults will also have natural leakage current, it is generally below the 300mA setting value of the primary residual current protection. When a user in the distribution area makes a wiring error in the neutral or ground wire, the load current of the user with the incorrect wiring will be converted into the residual current of the distribution area, with a value ranging from several amperes to tens of amperes.

[0010] The residual current data of the low-voltage distribution area mentioned above is collected by the residual current detection module of the intelligent power distribution terminal of the distribution area, which is a conventional technology in this field.

[0011] Step 2: Obtain the residual current collected at the same time intervals on any day in the leakage transformer area, as well as the load current data collected from each user in the same time interval on the same day in the transformer area, and construct the time series of the residual current of the leakage transformer area and the time series of the load current of each user on the same day.

[0012] The aforementioned user load current is measured and reported to the metering center by a new type of HPLC smart meter. The data can be obtained from the electricity information collection system, which is a conventional technology in this field.

[0013] The acquisition interval for residual current and load current in the time series should not exceed 30 minutes. When constructing these two time series, the time interval for the metering data from the new HPLC smart meter is generally 15 minutes, meaning data is collected once every 15 minutes, resulting in 96 points of HPLC smart meter data collected per day. Therefore, it is best to divide the user load current data acquired in a single day into 15-minute intervals. For example, if the user's daily load current time series X... i ={X i1 ,X i2 ,…X in}, where i = 1, 2, ..., N, is the number of users; n is the number of elements in the time series; when data is collected every 15 minutes, n is 96, and at this time the daily residual current time series of the leakage radio station area is Y = {Y1, Y2, ..., Y}. n The number of elements n in} is also 96. Of course, this time length can be adjusted according to the different acquisition capabilities of different stations, and it is not necessarily 15 minutes. The lower limit of data acquisition density should not be less than 48 HPLC smart meter data points, that is, n is at least 48.

[0014] Step 3: Based on the constructed residual current time series of the leakage circuit area and the load current time series of each user, a ridge regression analysis method is used to model the model. The standardization coefficient and the penalty value when the model first tends to flatten out are determined to determine the final ridge regression model. The ridge regression coefficient of each user corresponding to the penalty value is calculated to determine whether the ridge regression coefficient value of each user is abnormal. If it is abnormal, the neutral and ground wires of the user corresponding to the ridge regression coefficient value are abnormal. The user is a user with incorrect neutral and ground wire connections, and an on-site power inspection is conducted.

[0015] The process of determining whether the ridge regression coefficient values ​​of each user are abnormal, as mentioned above, is as follows: determine whether the user's ridge regression coefficient value is greater than 0.1; if it is, then it is abnormal.

[0016] This invention addresses leakage faults in transformer substations experiencing tripped primary residual current protection. It establishes a ridge regression analysis model using load current data from each user within the substation and residual current data to identify users with incorrect neutral and ground wire connections. The ridge regression coefficients reflect the contribution of each user's load current to the substation's residual current. The significantly higher ridge regression coefficients for users with incorrect neutral and ground wire connections compared to normal users help identify these abnormal users. Since incorrect neutral and ground wire connections are relatively common faults in low-voltage substations, a considerable proportion of leakage faults in transformer substations where primary residual current protection cannot be activated are caused by incorrect neutral and ground wire connections. The residual current caused by these wiring errors is not an arc current, and its spectral characteristics do not change significantly. Existing methods based on residual current spectral characteristics struggle to accurately identify and locate faults. By leveraging the strong correlation between the load current of users with wiring errors and the substation's residual current, this invention accurately locates users with wiring errors, promptly eliminates faults, improves the activation rate of primary residual current protection in the substation, and ensures residential electricity safety.

[0017] In fact, monitoring devices can also be added to the branch lines to detect electrical quantities such as residual current of the branch lines. Similarly, the method proposed in this application can be used to analyze and identify users who have incorrect neutral and ground wire wiring or whose residual current of the branch lines exceeds the set threshold value (such as 300mA) when the load current changes. Attached Figure Description

[0018] Figure 1 This is a comparison chart of the residual current and load current curves of the transformer area in an embodiment of the present invention. Detailed Implementation

[0019] This invention relates to a method for identifying users with wiring errors in the neutral and ground wires of leaky radio stations based on ridge regression analysis. The steps of this method are as follows:

[0020] Step 1: Identify the low-voltage transformer area where the excessive residual current causes frequent tripping of the primary residual current protection of the transformer as the leakage transformer area.

[0021] Step 2: Obtain the residual current data collected at the same time intervals on any given day for the leakage transformer area, as well as the load current data collected from each user in the same time interval on the same day for that transformer area, and construct the time series of the residual current for that leakage transformer area and the time series of the load current for each user on that day.

[0022] When constructing a time series, the time period for collecting user load current data in the transformer area must be the same as the time period for collecting residual current in the transformer area, and the collection interval should not exceed 30 minutes, that is, the number of sequence elements should not be less than 48.

[0023] Step 3: Based on the constructed residual current time series of the leakage circuit area and the load current time series of each user, a ridge regression analysis method is used to model the model. The standardized coefficients and the penalty value at the first flattening point in the ridge regression model are determined to establish the final ridge regression model. The ridge regression coefficients for each user corresponding to the penalty value are calculated, and it is determined whether the ridge regression coefficient values ​​for each user are abnormal. If abnormal, the neutral and ground wire connections of the user corresponding to that ridge regression coefficient value are abnormal, and the user is considered to have incorrect neutral and ground wire connections. In the following embodiment, the abnormality of the neutral and ground wire connections for a user is determined by whether the ridge regression coefficient of that user is greater than 0.1; if it is greater than 0.1, it is considered abnormal.

[0024] The process of establishing a ridge regression analysis model is a current technique and can be calculated using SPSS Statistics software. Specifically:

[0025] Suppose we have a linear regression model: Y = β0 + β1X1 + ... + β n X n +ε (1) Its basic assumption is that the independent variables are X1, X2, ..., X n There is no strict linear relationship between them; otherwise, it would seriously affect the determination of the regression coefficient β. This linear regression model is written as...

[0026] Y=β01=Xβ+ε (2)

[0027] where ε follows a multivariate normal distribution N(0,σ). 2 I m Let matrix X be an m×n matrix with rank p. Then the LS estimate of parameter β0 is...

[0028]

[0029] The LS estimates of the regression coefficients are

[0030]

[0031] The LS estimate obtained at this point is unbiased, so the estimate is... The mean square error is

[0032]

[0033] Where λ1≥λ2≥...λ n ≥0 is (X) T The eigenvalues ​​of X). If (X) T If X) has at least one eigenvalue close to 0, then The value of β will increase, i.e., the estimated value of β. An increase in the error between β and X also indicates that there is an approximate linear relationship between the column vectors of matrix X.

[0034] If there exist numbers α1, α2, ..., αn that are not all zero... n , making

[0035] α1X1+α2X2+...+α n X n =0 (6)

[0036] This linear regression model exhibits perfect collinearity. If a random error v exists, then Ev = 0 and Ev 2 <∞, such that

[0037] α1X1+α2X2+...+α n X n +v=0 (7)

[0038] This indicates that the linear regression model is not perfectly collinear. If the linear regression model is perfectly collinear, then the LS estimate of the regression coefficients does not exist. Therefore, the regression model to be discussed is not perfectly collinear, also known as multicollinearity.

[0039] To address the multicollinearity problem in the data, ridge regression is employed. Ridge regression, also known as ridge ridge or Tikhonov regularization, is a regularization method for regression analysis of ill-posed problems. This method is relevant to data analysis problems.

[0040] Xθ=y (8)

[0041] When using the least squares method, the loss function is defined as ||Xθ-y|| 2 The solution to the above problem is

[0042] θ=(X T X) -1 X T y (9)

[0043] When the data matrix X is not of full column rank or there is strong correlation between some columns, it becomes an ill-posed problem. In this case, calculating (X) T X) -1The error can be large, making the traditional least squares method lack stability.

[0044] To solve the above problems, a regularization term is added to the loss function, namely ||Xθ-y||. 2 +||Γθ|| 2 Where Γ = αI, α is the ridge parameter, and I is the identity matrix. Therefore, the original solution becomes...

[0045] θ(α)=(X T X+αI) -1 X T y (10)

[0046] As shown in the above equation, as α increases, the absolute values ​​of each element in θ(α) decrease, and the deviation from the correct value also increases. The trajectory of θ(α) as α changes is called the ridge trace. In practical applications, the minimum α value corresponding to the earliest point when the ridge trace changes tends to stabilize can be selected, and the final ridge regression model can be determined based on this value.

[0047] Example 1

[0048] Select a low-voltage residential transformer substation; the residual current curve for this substation is as follows. Figure 1 As shown, the residual current in this transformer area frequently exceeds the 300mA setting value of the primary residual current protection of the transformer, indicating a faulty leakage transformer area. Continuous load current data from 80 users at 96 points and corresponding continuous residual current data from 96 points in the same area were selected and imported into SPSS Statistics software. Ridge regression analysis was performed on the user current data and the transformer area residual current data, resulting in the ridge regression model shown in Table 1 below.

[0049] Table 1 Ridge Regression Model

[0050] punish Standardized coefficients and punish Standardized coefficients and punish Standardized coefficients and 0.000 1.000 0.340 0.172 0.680 0.130 0.020 0.709 0.360 0.166 0.700 0.128 0.040 0.571 0.380 0.163 0.720 0.126 0.060 0.385 0.400 0.160 0.740 0.124 0.080 0.318 0.420 0.158 0.760 0.123 0.100 0.260 0.440 0.155 0.780 0.121 0.120 0.253 0.460 0.152 0.800 0.120 0.140 0.267 0.480 0.149 0.820 0.114 0.160 0.231 0.500 0.147 0.840 0.118 0.180 0.211 0.520 0.144 0.860 0.115 0.200 0.199 0.540 0.142 0.880 0.114 0.220 0.199 0.560 0.143 0.900 0.113 0.240 0.194 0.580 0.140 0.920 0.111 0.260 0.192 0.600 0.138 0.940 0.110 0.280 0.181 0.620 0.135 0.960 0.105 0.300 0.177 0.640 0.134 0.980 0.105 0.320 0.172 0.660 0.131 1.000 0.103

[0051] The ridge regression model represents the ridge trace, where the standardized coefficients describe the trend of the ridge trace. Table 1 shows that the penalty value corresponding to the first flattening of the ridge trace is 0.22, and the ridge regression coefficients at this point are shown in Table 2.

[0052] Table 2 Ridge Regression Coefficients

[0053]

[0054]

[0055] As shown in Table 2, the ridge regression coefficient for user 15# is 0.321, significantly greater than 0.1 and significantly greater than that of the other users. No other user has a ridge regression coefficient greater than 0.1. This indicates that user 15#'s load current has the greatest impact on the residual current in the transformer area, thus user 15# is identified as a suspected user with abnormal wiring. On-site inspection verified the analysis results as correct.

[0056] Example 2

[0057] Select 96 points of continuous current data from 80 users in another leakage transformer area and 96 points of continuous residual current data from the corresponding time in the same transformer area. Import the data into SPSS Statistics software and perform ridge regression analysis on the user current data and the transformer area residual current data. The ridge regression model is shown in Table 3 below.

[0058] Table 3 Ridge Regression Model

[0059]

[0060]

[0061] As shown in Table 3, the penalty value corresponding to the first time the change of the ridge trace tends to flatten out is 0.34, and the ridge regression coefficient at this time is shown in Table 4 below.

[0062] Table 4 Ridge Regression Coefficients

[0063] User x Ridge regression coefficient User x Ridge regression coefficient User x Ridge regression coefficient User x Ridge regression coefficient 1 -0.009 21 -0.036 41 0.008 61 0.050 2 -0.028 22 -0.040 42 -0.018 62 -0.070 3 -0.014 23 <![CDATA[ 0.116 ]]> 43 0.047 63 -0.031 4 0.031 24 -0.034 44 0.058 64 -0.028 5 0.042 25 0.025 45 0.032 65 -0.038 6 -0.020 26 -0.066 46 -0.055 66 0.034 7 -0.079 27 -0.008 47 0.069 67 -0.010 8 -0.051 28 -0.011 48 -0.038 68 0.002 9 0.045 29 -0.021 49 0.015 69 0.068 10 0.078 30 0.049 50 0.013 70 -0.036 11 0.013 31 0.013 51 0.043 71 -0.033 12 0.059 32 0.033 52 0.013 72 -0.054 13 -0.023 33 0.041 53 -0.097 73 -0.047 14 <![CDATA[ 0.250 ]]> 34 0.043 54 0.069 74 0.030 15 0.069 35 -0.035 55 0.065 75 0.014 16 -0.047 36 0.007 56 -0.038 76 -0.048 17 0.043 37 0.033 57 0.028 77 0.047 18 -0.031 38 0.024 58 -0.074 78 0.066 19 0.044 39 0.039 59 0.015 79 0.005 20 0.038 40 0.061 60 0.021 80 0.015

[0064] As shown in Table 4 above, the ridge regression coefficients for users 14# and 23# are both greater than 0.1, significantly different from those of other users. This indicates that the load current of users 14# and 23# has a significantly greater impact on the residual current in the transformer area than other users, thus identifying users 14# and 23# as suspected users with abnormal wiring. On-site inspection confirmed this result. Furthermore, since users 14# and 23# are connected to different phases, this demonstrates that ridge regression analysis is also effective in detecting abnormal neutral-to-ground wiring between two users in different phases.

[0065] Example 3

[0066] We selected 96 points of continuous current data from 80 users in a certain leakage transformer area and 96 points of continuous residual current data from the corresponding time in the same transformer area. We then imported the data into SPSS Statistics software and performed ridge regression analysis on the user current data and the transformer area residual current data. The ridge regression model is shown in Table 5 below.

[0067] Table 5 Ridge Regression Model

[0068] punish Standardized coefficients and punish Standardized coefficients and punish Standardized coefficients and 0.000 1.000 0.340 0.148 0.680 0.104 0.020 0.683 0.360 0.144 0.700 0.102 0.040 0.533 0.380 0.141 0.720 0.101 0.060 0.287 0.400 0.139 0.740 0.099 0.080 0.262 0.420 0.134 0.760 0.098 0.100 0.248 0.440 0.131 0.780 0.096 0.120 0.235 0.460 0.128 0.800 0.095 0.140 0.213 0.480 0.125 0.820 0.093 0.160 0.214 0.500 0.123 0.840 0.092 0.180 0.194 0.520 0.120 0.860 0.090 0.200 0.259 0.540 0.118 0.880 0.096 0.220 0.181 0.560 0.116 0.900 0.088 0.240 0.174 0.580 0.114 0.920 0.087 0.260 0.180 0.600 0.111 0.940 0.086 0.280 0.164 0.620 0.109 0.960 0.084 0.300 0.158 0.640 0.107 0.980 0.083 0.320 0.152 0.660 0.105 1.000 0.082

[0069] As shown in Table 5, the penalty value corresponding to the first time the change of the ridge trace tends to flatten out is 0.16, and the ridge regression coefficient at this time is shown in Table 6 below.

[0070] Table 6 Ridge Regression Coefficients

[0071] User x Ridge regression coefficient User x Ridge regression coefficient User x Ridge regression coefficient User x Ridge regression coefficient 1 -0.075 21 -0.074 41 0.024 61 -0.044 2 0.031 22 0.033 42 -0.101 62 -0.056 3 -0.061 23 0.082 43 0.014 63 0.11 4 0.093 24 -0.029 44 -0.085 64 0.016 5 0.011 25 -0.021 45 0.033 65 0.012 6 0.036 26 -0.056 46 0.031 66 0.037 7 0.035 27 -0.056 47 0.035 67 0.031 8 0.008 28 -0.034 48 0.017 68 -0.027 9 <![CDATA[ 0.355 ]]> 29 -0.081 49 0.047 69 -0.08 10 <![CDATA[ 0.259 ]]> 30 -0.017 50 -0.038 70 -0.015 11 0.032 31 0.063 51 -0.003 71 0.05 12 -0.056 32 0.031 52 -0.087 72 -0.009 13 -0.02 33 0.002 53 0.105 73 0.03 14 0.005 34 0.022 54 -0.094 74 -0.042 15 -0.024 35 0.028 55 -0.053 75 -0.012 16 0.031 36 -0.005 56 -0.046 76 0.046 17 -0.036 37 0.032 57 -0.024 77 -0.032 18 0.04 38 0.065 58 0.017 78 0.047 19 0.013 39 0.033 59 0.059 79 0.037 20 0.077 40 0.064 60 0.033 80 0.028

[0072] As shown in Table 6 above, the ridge regression coefficients for users #9 and #10 are significantly greater than 0.1, and differ significantly from those of other users. This indicates that the load current of users #9 and #10 has a significantly greater impact on the residual current in the transformer area than other users. Therefore, users #9 and #10 are suspected users with abnormal wiring. The on-site inspection confirmed the results.

[0073] Comparative Example 1

[0074] The Pearson correlation coefficient was calculated and analyzed using the method CN114114074A for the 96-point continuous current data of 80 users used in Example 3 and the corresponding 96-point residual current data sequence of the same transformer area. The results are shown in Table 7 below.

[0075] Table 7. Results of Pearson correlation coefficient

[0076] User x Correlation coefficient User x Correlation coefficient User x Correlation coefficient User x Correlation coefficient 1 -0.120 21 -0.081 41 0.092 61 -0.021 2 0.251 22 -0.084 42 -0.185 62 -0.044 3 -0.242 23 0.140 43 0.243 63 0.215 4 0.305 24 0.104 44 -0.114 64 0.329 5 0.205 25 0.051 45 0.148 65 0.021 6 0.089 26 -0.057 46 0.008 66 0.093 7 -0.014 27 -0.365 47 -0.002 67 0.051 8 -0.148 28 -0.028 48 0.248 68 -0.022 9 <![CDATA[ 0.858 ]]> 29 -0.267 49 0.050 69 -0.042 10 <![CDATA[ 0.479 ]]> 30 -0.132 50 -0.007 70 0.229 11 -0.073 31 0.024 51 0.003 71 0.336 12 -0.044 32 -0.069 52 -0.068 72 0.122 13 0.251 33 0.042 53 0.442 73 -0.098 14 -0.054 34 -0.001 54 0.028 74 -0.089 15 0.214 35 0.329 55 -0.072 75 -0.135 16 -0.138 36 0.039 56 -0.127 76 0.128 17 -0.162 37 -0.153 57 0.164 77 0.065 18 0.047 38 -0.013 58 -0.006 78 0.150 19 0.087 39 -0.028 59 0.284 79 0.206 20 0.314 40 0.182 60 0.049 80 -0.223

[0077] As shown in Table 7 above, the correlation coefficient for user #9 is greater than 0.7. According to the method in CN114114074A, there is a significant correlation between the load current and the residual current in the transformer area for this user, indicating an abnormal wiring situation. However, the correlation coefficient for user #10 is only 0.479, which is less than 0.7. This means that the correlation between the load current and the residual current in the transformer area for this user is weak, which is inconsistent with the inspection results of this invention.

[0078] Comparative Example 2

[0079] The Euclidean distance of the 96-point continuous current data of 80 users used in Example 3 and the corresponding 96-point residual current data sequence of the transformer area were analyzed according to the method of CN113933585A. The results are shown in Table 8 below.

[0080] Table 8. Euclidean distance calculation results

[0081] User x distance User x distance User x distance User x distance 1 77.125 21 83.364 41 77.899 61 92.279 2 74.703 22 106.432 42 75.815 62 70.337 3 82.533 23 113.854 43 66.720 63 77.064 4 73.625 24 82.184 44 80.013 64 73.128 5 74.118 25 115.614 45 74.296 65 96.755 6 86.541 26 73.861 46 96.668 66 80.960 7 85.100 27 174.378 47 123.520 67 72.618 8 172.055 28 78.685 48 62.325 68 77.391 9 <![CDATA[ 40.385 ]]> 29 110.231 49 74.552 69 82.208 10 <![CDATA[ 61.978 ]]> 30 75.607 50 77.593 70 71.875 11 82.397 31 83.482 51 103.361 71 71.862 12 82.142 32 152.476 52 83.437 72 80.565 13 68.194 33 76.784 53 57.720 73 253.917 14 113.553 34 87.176 54 72.792 74 76.272 15 86.720 35 69.188 55 78.401 75 79.292 16 106.880 36 73.850 56 81.711 76 73.491 17 79.708 37 87.317 57 58.539 77 64.865 18 73.650 38 128.045 58 72.293 78 74.901 19 74.948 39 84.085 59 72.138 79 75.011 20 51.851 40 87.373 60 76.688 80 118.648

[0082] As can be seen from the Euclidean distance calculation results in Table 8 above, user #9 has the smallest Euclidean distance value. According to the method in CN113933585A, user #9 is suspected of having a wiring abnormality, but user #10 does not have any suspected wiring abnormality, which is inconsistent with the analysis results of this invention. Therefore, when there are two abnormal users, the Euclidean distance calculation results for these two users are not significantly different from those for other users, and the Euclidean distance calculation results cannot be used to screen these two abnormal users.

[0083] As can be seen from Example 3 and comparative Examples 1 and 2, the collinearity problem among data increases with the number of users. Screening for abnormal users solely based on criteria such as the correlation between user load current data and residual current data in the transformer area, or Euclidean distance, may result in missed or false positives. This problem is even more pronounced when there are multiple abnormal users within the transformer area, as the vector superposition of user currents exacerbates the issue. However, this invention utilizes ridge regression analysis, which effectively addresses the collinearity problem, and the verification analysis has yielded excellent results.

Claims

1. A method for identifying users with wiring errors in the neutral and ground wires of a leaky radio station based on ridge regression analysis, characterized in that, The steps of this method are as follows: Step 1: Identify the low-voltage transformer area where the excessive residual current causes frequent tripping of the primary residual current protection of the distribution transformer as the leakage transformer area. Step 2: Obtain the residual current collected at the same time intervals on any day in the leakage transformer area, as well as the load current data collected from each user in the same time interval on the same day in the transformer area, and construct the time series of the residual current of the leakage transformer area and the time series of the load current of each user on the same day. Step 3: Based on the constructed residual current time series of the leakage circuit area and the load current time series of each user, the ridge regression analysis method is used to model the model. The standardization coefficient and the penalty value when the model first tends to flatten out are determined to determine the final ridge regression model. The ridge regression coefficient of each user corresponding to the penalty value is calculated to determine whether the ridge regression coefficient value of each user is abnormal. If it is abnormal, the neutral and ground wires of the user corresponding to the ridge regression coefficient value are abnormal, and the user is a user with incorrect neutral and ground wire connections. The process of determining whether the ridge regression coefficient value of each user is abnormal in step three is as follows: determine whether the ridge regression coefficient value of the user is greater than 0.1, and if it is, it is abnormal.

2. The method for identifying users with wiring errors in the neutral and ground wires of a leaky radio station based on ridge regression analysis as described in claim 1, characterized in that, In step two, the acquisition interval for residual current and load current in the time series should not exceed 30 minutes, meaning the number of sequence elements should not be less than 48.

Images