A method for identifying abnormal data of synchronous line loss based on the approximate equality characteristic of variable weight rank sum

Abstract

The present invention discloses a method for identifying abnormal data of synchronous line losses based on the characteristic of variable weighted rank and approximate equality in the field of line loss data management of electric power systems. The method includes: obtaining a theoretical line loss power data sequence with the same time scale as the synchronous line loss data sequence, and performing power-to-electricity conversion; obtaining the average value of the difference between the synchronous line loss and the theoretical line loss, and forming a difference sequence set; using a discrete ranking method to rank the elements in the difference sequence set according to the absolute value; using a penalty variable weight method to weight the corresponding ranks of the difference elements, and calculating the variable weighted positive and negative rank sums; obtaining the abnormal data rank and difference threshold to identify whether there is abnormal data. The method can comprehensively consider the influence of the amplitude of abnormal data and the number of abnormal data, and does not rely on indicator selection for identification. It has high identification accuracy, strong adaptability, and fast identification speed.

Description

[0001] The present invention belongs to the field of abnormal data management and repair of power systems, and in particular relates to a method for identifying abnormal data of synchronous line losses. Background Art

[0002] In order to improve the line loss management level of power grid companies, State Grid Corporation of China has vigorously promoted the construction of an integrated power and line loss management system (synchronous line loss system). However, in actual operation, it was found that there were many abnormal data in the massive synchronous line loss system data, which seriously affected the line loss index assessment compliance rate and constrained the lean management level of line loss. Therefore, the identification of abnormal data in the synchronous line loss system is of great significance for line loss index assessment and optimizing the economic operation of the power grid.

[0003] At present, there have been some studies on data anomaly identification, but most of them focus on the anomaly identification of power system operation status data or load data. In terms of power system status abnormal data identification, it mainly includes low-rank model decomposition method, power flow tracking method, residual identification method, etc.; in terms of load abnormal data identification, it mainly includes low-rank matrix decomposition method, improved FCM algorithm, etc.; the above methods can significantly improve the identification accuracy and efficiency of power system data. However, the above methods are all abnormal data identification methods proposed for specific application scenarios of power systems. When they are directly applied to the abnormal data identification of synchronous line loss systems, the identification effect is not good, and the misjudgment rate and missed judgment rate are both high. With the development of fast calculation methods for theoretical line losses, it is possible to identify abnormal data by comparing theoretical line losses with synchronous line losses. Through research, it is found that the difference sequence between synchronous line losses and theoretical line losses has the characteristic of approximate rank sum equality. Using this rank sum approximation characteristic for abnormal data identification, the misjudgment rate and missed judgment rate are significantly reduced, and the identification effect is good. Summary of the invention

[0004] The purpose of the present invention is to provide a method for identifying abnormal data of synchronous line loss based on the variable weight rank and approximate equality characteristics, so as to solve the above problems existing in the existing methods. To achieve the above purpose, the technical solution provided by the present invention is a method for identifying abnormal data of synchronous line loss based on the variable weight rank and approximate equality characteristics, characterized in that the method comprises the following steps:

[0005] S1: Obtain the theoretical line loss power data sequence with the same time scale as the line loss data sequence of the same period, and perform power-to-electricity conversion;

[0006] S2: Calculate the average difference between the line loss in the same period and the theoretical line loss, and form a difference sequence set;

[0007] S3: Use discrete ranking method to rank the elements in the difference sequence set according to their absolute values;

[0008] S4: Use the penalty variable weight method to assign weights to the corresponding ranks of the difference elements, and calculate the sum of the variable weighted positive and negative ranks;

[0009] S5: According to the relative relationship between the positive and negative rank sums, analyze the approximate equality of the positive and negative rank sums when there is no abnormal data and the abnormal characteristics of the positive and negative rank sums when there is abnormal data;

[0010] S6: Based on this, the abnormal data rank and difference threshold are calculated to identify whether there is abnormal data, and the above steps are repeated to determine the location of the abnormal data.

[0011] 2. S1 comprises the following steps:

[0012] S101: Calculate a theoretical line loss power data sequence with the same time scale as the line loss data sequence of the same period;

[0013] Theoretical line loss is calculated based on a 15-minute time scale. Different voltage levels use different calculation methods for theoretical line loss. Voltage levels above 220kV use the power flow method, and voltage levels above 35kV and below 220kV use the forward-and-backward method.

[0014] S102: performing power-electricity conversion on the theoretical line loss power sequence to form a theoretical line loss electric quantity data sequence with the same time scale as the line loss data sequence of the same period;

[0015] 3. S2 comprises the following steps:

[0016] S201: Calculate the average difference between the line loss in the same period and the theoretical line loss element;

[0017] S202: construct a series of electric quantity ΔL of difference between line loss and theoretical line loss in the same period;

[0018] 4. S3 comprises the following steps:

[0019] Sort the elements in ΔL by absolute value from small to large and get new numbers. Let ΔL i The corresponding new number is SN(ΔL i ), then the rank is:

[0020] R(ΔL i )=sgn(ΔL i )·SN(ΔL i )

[0021] Where, ΔL i is an element in the difference electric quantity sequence set; sgn(x) is the sign function, and its calculation formula is:

[0022]

[0023] 5. S4 comprises the following steps:

[0024] S401: Calculate the penalty variable weight function value of the corresponding rank of the difference element, and the calculation formula is as follows:

[0025]

[0026] In the formula, is the penalty variable weight factor, which is related to the probability density function of ΔL. The calculation formula is:

[0027]

[0028] Easy to know, is the cumulative probability function of the line loss difference rate, and its value is [0, 1]. Therefore, 0≤Φ≤1, which satisfies the value condition of the variable weight factor in [0, 1].

[0029] S402: Calculate the variable weighted positive and negative “rank sums”, the calculation formula is as follows:

[0030]

[0031]

[0032] In the formula, is the difference element ΔL i The weight of .

[0033] 6. S5 comprises the following steps:

[0034] S501: Analyze the approximate equality characteristics of positive and negative ranks of time-varying weights in normal data without abnormalities.

[0035] If the quality of the data of the line loss system in the same period is high, the difference between the statistical line loss power in the same period and the theoretical line loss power should be randomly and evenly distributed in the neighborhood of zero, so the rank value of the difference sequence should be evenly distributed between [-n, +n], and the overall distribution of the difference should be symmetrical. The weight is related to the difference and is also symmetrical, so the difference between the variable weight positive and negative rank sums should not be too large, that is, in the case of normal data without abnormal data, the positive rank sum With negative rank sum The absolute values ​​of are approximately equal, that is,

[0036]

[0037] S502: Analyze the positive and negative ranks and abnormal characteristics when the line loss data set of the same period contains abnormal data.

[0038] When there are abnormal data in the same period system, it will cause the variable weighted positive rank sum Significantly greater than, less than, or not equal to the negative rank and absolute value The above-mentioned rank and approximate equality property no longer holds.

[0039] 7. S6 comprises the following steps:

[0040] If and only if the observed value r of the positive rank sum + When the following formula is satisfied, it is considered that the data quality of the line loss system in the same period is good and there is no abnormal data.

[0041] C L ≤r + ≤C U

[0042] In the formula, C L is a positive rank sum r + The lower threshold of C U is a positive rank sum r + The upper threshold of

[0043] The key to identifying abnormal data is to determine the above two thresholds.

[0044] S601: First, the α threshold considering only the misjudgment probability α is calculated, and the calculation formula is as follows:

[0045]

[0046] In the formula, When there is no abnormal data in the sample, R + The probability density distribution function of , which can be calculated based on historical data, is related to the sample size n. is a symmetric distribution function.

[0047] S602: Then calculate the β threshold considering the missed probability β, and the calculation formula is as follows;

[0048]

[0049] In the formula, When there are abnormal data in the sample, R + The probability density distribution function is calculated based on historical data and is related to the sample size n. It is a skewed bimodal symmetrical distribution function, which is far away from the middle value in most cases.

[0050] S603: Finally, the α∩β threshold is calculated by comprehensively considering the misjudgment probability α and the missed judgment probability β. The calculation is divided into three cases.

[0051] ① When When , the α∩β threshold needs to be compromised, and the weighted average of the α threshold and the β threshold is selected:

[0052]

[0053] ② When When , the α∩β threshold is:

[0054]

[0055] ③When When , both the α threshold and the β threshold just meet the requirements, and the α∩β threshold is:

[0056]

[0057] S604: Use the binary method to gradually reduce the sample size, repeat the above steps, and determine the location of abnormal data.

[0058] The present invention provides a method for identifying abnormal data of synchronous line loss based on the approximate equality characteristic of variable weighted rank sum, which is as follows: obtaining the theoretical line loss power data sequence with the same time scale as the synchronous line loss data sequence, and performing power-electricity conversion; obtaining the average difference between the synchronous line loss and the theoretical line loss, and forming a difference sequence set; using the discrete ranking method to rank the elements in the difference sequence set according to the absolute value; using the penalty variable weight method to weight the corresponding ranks of the difference elements, and calculating the variable weighted positive and negative rank sums; according to the relative size relationship of the variable weighted positive and negative rank sums, analyzing the approximate equality characteristics of the positive and negative rank sums when there is no abnormal data and the abnormal characteristics of the positive and negative rank sums when there is abnormal data; based on this, obtaining the abnormal data rank sum difference threshold, identifying whether there is abnormal data, and repeating the above steps to determine the position of the abnormal data. The method can comprehensively consider the influence of the abnormal data amplitude and the number of abnormal data, can be identified without relying on the selection of indicators, has high identification accuracy, strong adaptability, and fast identification speed, and is of great significance to the management of line loss data and improving the level of line loss management. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments.

[0060] Figure 1 The present invention provides a flow chart of a method for identifying abnormal data of synchronous line losses based on the variable weight rank and approximately equal characteristics. DETAILED DESCRIPTION

[0061] In order to clearly understand the technical solution of the present invention, its detailed structure will be presented in the following description. Obviously, the specific implementation of the embodiment of the present invention is not limited to the specific details familiar to those skilled in the art. Typical embodiments of the present invention are described in detail below. In addition to these embodiments described in detail, other implementation methods may also be provided.

[0062] The present invention is further described in detail below with reference to the accompanying drawings and embodiments.

[0063] Example 1

[0064] Figure 1The present invention is a flow chart of a method for identifying abnormal data of synchronous line loss based on the variable weight rank and approximately equal characteristics. Figure 1 The present invention provides a flowchart of a method for identifying abnormal data of synchronous line loss based on the variable weight rank and approximately equal characteristics, including:

[0065] S1: Obtain the theoretical line loss power data sequence with the same time scale as the line loss data sequence of the same period, and perform power-to-electricity conversion;

[0066] S2: Calculate the average difference between the line loss in the same period and the theoretical line loss, and form a difference sequence set;

[0067] S3: Use discrete ranking method to rank the elements in the difference sequence set according to their absolute values;

[0068] S4: Use the penalty variable weight method to assign weights to the corresponding ranks of the difference elements, and calculate the sum of the variable weighted positive and negative ranks;

[0069] S5: According to the relative relationship between the positive and negative rank sums, analyze the approximate equality of the positive and negative rank sums when there is no abnormal data and the abnormal characteristics of the positive and negative rank sums when there is abnormal data;

[0070] S6: Based on this, the abnormal data rank and difference threshold are calculated to identify whether there is abnormal data, and the above steps are repeated to determine the location of the abnormal data.

[0071] The S1 comprises the following steps:

[0072] S101: Calculate a theoretical line loss power data sequence with the same time scale as the line loss data sequence of the same period;

[0073] The theoretical line loss is calculated based on a 15-minute time scale. Different voltage levels use different methods to calculate the theoretical line loss.

[0074] ① The power flow method is used for voltage levels above 220kV. The line loss calculation formula based on power flow is:

[0075]

[0076] Where U i is the voltage at node i; G ij is the mutual admittance between node i and node j; δ ij is the phase angle difference between node i and node j.

[0077] Written in matrix form:

[0078]

[0079] ② For voltage levels above 35kV and below 220kV, the forward-backward method is used for calculation. The nth iteration formula of the forward-backward method for the distribution network is:

[0080] The current back-substitution formula of the branch is:

[0081]

[0082] Where: -The current vector flowing through branch ik at the nth iteration.

[0083] Node k voltage forward formula:

[0084]

[0085] Where: is the voltage vector of node k at the nth iteration; r ik +jx ik is the impedance of branch ik.

[0086] S102: performing power-electricity conversion on the theoretical line loss power sequence to form a theoretical line loss electric quantity data sequence with the same time scale as the line loss data sequence of the same period;

[0087] The above theoretical line loss calculation results in line loss power data, while the line loss obtained in the same period results in line loss electricity data. When performing difference data analysis, power-electricity conversion is required. The theoretical line loss calculation is performed on a 15-minute time scale and converted according to formula (5).

[0088]

[0089] Where N d is the time interval for difference comparison. For daily loss, N d =96.

[0090] The S2 comprises the following steps:

[0091] S201: Calculate the average difference between the line loss in the same period and the theoretical line loss

[0092] Let the synchronous line loss and theoretical line loss set be L s and L c , as shown in formula (6),

[0093]

[0094] The average difference between the line loss and the theoretical line loss in the same period is:

[0095]

[0096] S202: Construct a series of electric quantity difference between line loss and theoretical line loss in the same period. The construction formula is as follows:

[0097] ΔL=L s -Lc -E(L s -L c ) (8)

[0098] The S3 comprises the following steps:

[0099] Sort the elements in ΔL by absolute value from small to large and get new numbers. Let ΔL i The corresponding new number is SN(ΔL i ), then the rank is:

[0100] R(ΔL i )=sgn(ΔL i )·SN(ΔL i ) (9)

[0101] Where, ΔL i is an element in the difference electric quantity sequence set; sgn(x) is the sign function, and its calculation formula is:

[0102]

[0103] The S4 comprises the following steps:

[0104] S401: Calculate the penalty variable weight function value of the corresponding rank of the difference element, and the calculation formula is as follows:

[0105]

[0106] In the formula, is the penalty variable weight factor, which is related to the probability density function of ΔL. The calculation formula is:

[0107]

[0108] Easy to know, is the cumulative probability function of the line loss difference rate, and its value is [0, 1]. Therefore, 0≤Φ≤1, which satisfies the value condition of the variable weight factor in [0, 1].

[0109] S402: Calculate the variable weighted positive and negative “rank sums”, the calculation formula is as follows:

[0110]

[0111] The S5 comprises the following steps:

[0112] S501: Analyze the normal and non-abnormal data, and the characteristics of the time-varying weights of positive and negative ranks are approximately equal. According to formula (13),

[0113]

[0114] because Therefore

[0115]

[0116] If the quality of the data of the line loss system in the same period is high, the difference between the statistical line loss power in the same period and the theoretical line loss power should be randomly and evenly distributed in the neighborhood of zero, so the rank value of the difference sequence should be evenly distributed between [-n, +n], and the overall distribution of the difference should be symmetrical. The weight is related to the difference and is also symmetrical. Therefore, the difference between the positive and negative rank sums obtained according to formula (13) should not be too large, that is, in the case of normal data without abnormal data, the positive rank sum With negative rank sum The absolute values ​​of are approximately equal, that is,

[0117]

[0118] S502: Analyze the positive and negative ranks and abnormal characteristics when the line loss data set of the same period contains abnormal data.

[0119] When there are abnormal data in the data, the above-mentioned rank and approximate equality characteristics will be destroyed, and new difference characteristics will be presented, as shown in Table 1.

[0120]

[0121]

[0122] Table 1 Rank and difference characteristic table

[0123] Note: In the table, In this article, they respectively mean significantly greater than, significantly less than, and significantly not equal to.

[0124] Taking line loss as an example, Table 1 is explained as follows:

[0125] (1) When there are zero-value abnormal data in the line loss electricity of the same period, the line loss electricity of the same period will decrease, and the theoretical line loss electricity will remain unchanged. Then, the number of samples with negative difference between the line loss of the same period and the theoretical line loss will increase, and the negative rank sum will increase. Therefore,

[0126] (2) When there are constant abnormal data in the line loss electricity, the positive and negative deviations between the abnormal data and the actual data cannot be determined, so the relative relationship between the positive and negative rank sums is uncertain. However, the difference between the line loss electricity and the theoretical line loss electricity increases, which will destroy the approximate equality characteristic.

[0127] (3) When there are abnormal data in the line loss (generally abnormal increase), the line loss in the same period will increase significantly, and the theoretical line loss will remain unchanged. Then, the number of samples with positive difference between the line loss in the same period and the theoretical line loss will increase, and the positive rank sum will increase. Therefore, R +>|R - |.

[0128] The S6 comprises the following steps:

[0129] According to S6, under normal conditions with no abnormal data, So positive rank sum should be in the middle of formula (16), so for a given sample, when the calculated positive rank sum observation value r + If the value is too large or too small, it is considered that the line loss in the same period is too different from the theoretical line loss, and there are abnormal data in the sample. + When formula (17) is satisfied, it is considered that the data quality of the line loss system in the same period is good and there is no abnormal data.

[0130] C L ≤r + ≤C U (17)

[0131] In the formula, C L is a positive rank sum r + The lower threshold of C U is a positive rank sum r + The upper threshold of

[0132] The key to identifying abnormal data is to determine the above two thresholds.

[0133] S601: First, the α threshold considering only the misjudgment probability α is calculated, and the calculation formula is as follows:

[0134]

[0135] In the formula, When there is no abnormal data in the sample, R + The probability density distribution function of , which can be calculated based on historical data, is related to the sample size n. is a symmetric distribution function.

[0136] S602: Then calculate the β threshold considering the missed probability β, and the calculation formula is as follows;

[0137]

[0138] In the formula, When there are abnormal data in the sample, R + The probability density distribution function is calculated based on historical data and is related to the sample size n. It is a skewed bimodal symmetrical distribution function, which is far away from the middle value in most cases.

[0139] S603: Finally, the α∩β threshold is calculated by comprehensively considering the misjudgment probability α and the missed judgment probability β. The calculation is divided into three cases.

[0140] ① When When , the α∩β threshold needs to be compromised, and the weighted average of the α threshold and the β threshold is selected:

[0141]

[0142] When the actual requirement is to retain normal data as much as possible, that is, the misjudgment probability α is required to be smaller, α<β, the α∩β threshold in formula (20) is more inclined to the α threshold; conversely, if the actual requirement is to eliminate abnormal data as much as possible, that is, the missed judgment probability β is required to be smaller, β<α, the α∩β threshold in formula (20) is more inclined to the β threshold.

[0143] ② When When the threshold is α, the probability of missed judgment β will be reduced. When the threshold is β, the probability of misjudgment α will be reduced. Both requirements can be met. Through analysis, it can be seen that as long as the threshold is between the α threshold and the β threshold, the requirements are met. Therefore, the α∩β threshold is:

[0144]

[0145] ③When When , both the α threshold and the β threshold just meet the requirements, and the α∩β threshold is:

[0146]

[0147] S604: Use the binary method to gradually reduce the sample size, repeat the above steps, and determine the location of abnormal data.

[0148] Embodiment 2:

[0149] Taking the line loss system data of Gansu Province in the same period as an example, 10 330kV busbar daily line loss data were selected for abnormal data identification. The collection period was from January 1, 2017 to December 31, 2017. Each busbar had 365 data points, totaling 3650 data points. After statistical analysis, there were 42 zero-value abnormal data and 67 constant-value abnormal data in these data. Different thresholds ε were set. c , the mutation data obtained are different, as shown in Table 2.

[0150]

[0151]

[0152] Table 2 Mutation data table

[0153] From the above analysis, we can see that the mutation anomaly data obtained by setting different thresholds are not the same. Through logical relationship data verification, it is found that there are 73 non-zero and non-constant value anomaly data, which can be considered as mutation anomaly data, which is similar to the result of identifying anomaly quantity with a threshold of 75%, but the position of the anomaly data cannot correspond well, and there is a high misjudgment rate and missed judgment rate.

[0154] Taking this as an example, the present invention provides a method for identifying abnormal line loss data in the same period based on the variable weight rank sum approximate equality characteristic, which includes:

[0155] S1: Obtain the theoretical line loss power data sequence with the same time scale as the line loss data sequence of the same period, and perform power-to-electricity conversion;

[0156] According to formula (1), the theoretical line loss power data series of the same time period is calculated, where the power of each line part is shown in Table 3:

[0157]

[0158] Table 3 Theoretical line loss power table of each transmission line (kW)

[0159] According to the power-electricity data conversion of formula (5), the theoretical line loss electricity table of the same time scale (day) is obtained, as shown in Table 4:

[0160]

[0161] Table 4 Daily theoretical line loss of each transmission line (kWh)

[0162] S2: Calculate the average difference between the line loss in the same period and the theoretical line loss, and form a difference sequence set

[0163] From the synchronous line loss system, the daily synchronous line loss electricity table of each transmission line can be obtained, as shown in Table 5:

[0164]

[0165] 5. Electricity loss meter of each transmission line in the same period (kWh)

[0166] Then the average value of the difference between the line loss and the theoretical line loss of each transmission line in the same period is:

[0167]

[0168] Table 6 Average value of power consumption of the difference between line loss and theoretical line loss of each transmission line in the same period (kWh)

[0169] According to formula (8), the difference quantity series between the line loss and the theoretical line loss of each transmission line in the same period is constructed, as shown in Table 7:

[0170]

[0171] Table 7 The difference between the line loss and theoretical line loss of each transmission line in the same period (kWh)

[0172] S3: Use discrete ranking method to rank the elements in the difference sequence set according to their absolute values;

[0173] According to formula (9), the elements in the transmission line difference sequence ΔL in Table 7 are discretely ranked, and the ranking results are shown in Table 8:

[0174]

[0175]

[0176] Table 8 Discrete ranking results of each transmission line

[0177] S4: Use the penalty variable weight method to assign weights to the corresponding ranks of the difference elements, and calculate the sum of the variable weighted positive and negative ranks;

[0178] According to formula (11), the discrete ranking results of each transmission line in Table 8 are weighted, and the weighted results are shown in Table 9.

[0179]

[0180] Table 9 Weighting results

[0181] According to formula (13), the positive and negative ranks and absolute values ​​of each transmission line are calculated, and the results are shown in Table 10:

[0182]

[0183]

[0184] Table 10 Positive and negative rank sum results

[0185] S5: According to the relative relationship between the positive and negative rank sums, analyze the approximate equality of the positive and negative rank sums when there is no abnormal data and the abnormal characteristics of the positive and negative rank sums when there is abnormal data;

[0186] According to the analysis, when the data is normal and there is no abnormality, the absolute values ​​of the positive rank sum and the negative rank sum are approximately equal, otherwise the difference between the two is large.

[0187] S6: Obtain the abnormal data rank and difference threshold, identify whether there is abnormal data, and repeat the above steps to determine the location of the abnormal data.

[0188] According to formulas (20) to (22), the abnormal thresholds of each transmission line are calculated as shown in Table 11:

[0189]

[0190] Table 11 Positive rank and anomaly thresholds

[0191] Combining Table 10 and Table 11, we can see that the positive rank sum of Yuqing I, Hechuan II, Xianwu I, Xianwu II, and Liangwei II exceeds the threshold, and there are abnormal data. The binary method is used to repeat the above steps to determine the location of the abnormal data and perform statistics, as shown in Table 12:

[0192] Exception Type Abnormal number False positive rate Missed Detection Rate Identification accuracy Zero value exception 42 0 0 100％ Constant value abnormality 67 0 0 100％ Mutation Abnormality 71 4.11％ 2.74％ 94.52％

[0193] Table 12 Identification status table

[0194] According to Table 12, for zero-value abnormal data and constant-value abnormal data, the identification method in this paper can achieve a 100% accuracy rate. For mutation abnormal data, the misjudgment rate and missed judgment rate can be controlled below 5%, and the identification accuracy rate can reach 94.52%, which has a high identification accuracy.

[0195] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field can still modify or replace the specific implementation methods of the present invention with equivalents. Any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention are within the scope of protection of the pending claims.

Claims

1. A method for identifying abnormal line loss data in the same period based on the variable weight rank sum approximate equality characteristic, characterized in that: The method for identifying abnormal line loss data in the same period comprises the following steps: S1: Obtain the theoretical line loss power data sequence with the same time scale as the line loss data sequence of the same period, and perform power-to-electricity conversion; S2: Calculate the average difference between the line loss in the same period and the theoretical line loss, and form a difference sequence set; S3: Use discrete ranking method to rank the elements in the difference sequence set according to their absolute values; S4: Use the penalty variable weight method to assign weights to the corresponding ranks of the difference elements, and calculate the sum of the variable weighted positive and negative ranks; S5: According to the relative relationship between the positive and negative rank sums, analyze the approximate equality of the positive and negative rank sums when there is no abnormal data and the abnormal characteristics of the positive and negative rank sums when there is abnormal data; S6: Based on this, the abnormal data rank and difference threshold are calculated to identify whether there is abnormal data, and the above steps are repeated to determine the location of the abnormal data; The S5 comprises the following steps: S501: Analyze the time-varying weight positive and negative ranks of normal data without abnormalities and the approximate equality characteristics; If the quality of the data of the line loss system in the same period is high, the difference between the statistical line loss power in the same period and the theoretical line loss power should be randomly and evenly distributed in the neighborhood of zero, so the rank value of the difference sequence should be evenly distributed between [-n, +n], and the overall distribution of the difference should be symmetrical. The weight is related to the difference and is also symmetrical, so the difference between the variable weight positive and negative rank sums should not be too large, that is, in the case of normal data without abnormal data, the positive rank sum With negative rank sum The absolute values ​​of are approximately equal, that is, S502: Analyze the positive and negative ranks and abnormal characteristics when the line loss data set contains abnormal data in the same period; When there are abnormal data in the same period system, it will cause the variable weighted positive rank sum Significantly greater than, less than, or not equal to the negative rank and absolute value The phenomenon of , destroys the above-mentioned rank and approximate equality characteristics; The S6 comprises the following steps: If and only if the observed value r of the positive rank sum + When the following formula is satisfied, it is considered that the data quality of the line loss system in the same period is good and there is no abnormal data; C L ≤r + ≤C U In the formula, C L is a positive rank sum r + The lower threshold of C U is a positive rank sum r + The upper threshold of The key to abnormal data identification is to determine the two thresholds mentioned above; S601: First, the α threshold considering only the misjudgment probability α is calculated, and the calculation formula is as follows: In the formula, When there is no abnormal data in the sample, R + The probability density distribution function of , which can be calculated based on historical data, is related to the sample size n. is a symmetric distribution function; S602: Then calculate the β threshold considering the missed probability β, and the calculation formula is as follows; In the formula, When there are abnormal data in the sample, R + The probability density distribution function of is calculated based on historical data and is related to the sample size n; It is a skewed bimodal symmetrical distribution function, far away from the middle value in most cases; S603: Finally, the α∩β threshold which comprehensively considers the misjudgment probability α and the missed judgment probability β is calculated, and the calculation is divided into three cases; ① When When , the α∩β threshold needs to be compromised, and the weighted average of the α threshold and the β threshold is selected: ② When When , the α∩β threshold is: ③When When , both the α threshold and the β threshold just meet the requirements, and the α∩β threshold is: S604: Use the binary method to gradually reduce the sample size, repeat the above steps, and determine the location of abnormal data.

2. According to claim 1, a method for identifying abnormal line loss data in the same period based on the variable weight rank sum approximate equality characteristic is characterized in that: The S1 comprises the following steps: S101: Calculate a theoretical line loss power data sequence with the same time scale as the line loss data sequence of the same period; Theoretical line loss is calculated based on a 15-minute time scale. Different voltage levels use different calculation methods for theoretical line loss. Voltage levels above 220 kV use the power flow method, and voltage levels above 35 kV and below 220 kV use the forward-and-backward method. S102: performing power-electricity conversion on the theoretical line loss power sequence to form a theoretical line loss electric quantity data sequence having the same time scale as the line loss data sequence of the same period.

3. The method for identifying abnormal line loss data of the same period based on the variable weight rank sum approximate equality characteristic according to claim 1 is characterized in that: The S2 comprises the following steps: S201: Calculate the average difference between the line loss in the same period and the theoretical line loss element; S202: Construct a series of electric quantity difference between line loss in the same period and theoretical line loss.

4. The method for identifying abnormal line loss data of the same period based on the variable weight rank sum approximate equality characteristic according to claim 1 is characterized in that: The S3 comprises the following steps: Sort the elements in ΔL by absolute value from small to large and get new numbers. Let ΔL i The corresponding new number is SN(ΔL i ), then the rank is: R(ΔL i )=sgn(ΔL i )·SN(ΔL i ) Where, ΔL i is an element in the difference electric quantity sequence set; sgn(x) is the sign function, and its calculation formula is:

5. The method for identifying abnormal line loss data of the same period based on the variable weight rank sum approximate equality characteristic according to claim 1 is characterized in that: The S4 comprises the following steps: S401: Calculate the penalty variable weight function value of the corresponding rank of the difference element, and the calculation formula is as follows: In the formula, is the penalty variable weight factor, which is related to the probability density function of ΔL. The calculation formula is: Easy to know, is the cumulative probability function of the line loss difference rate, and its value is [0,1], so 0≤φ≤1, which satisfies the value condition of the variable weight factor in [0,1]; S402: Calculate the variable weighted positive and negative "rank sums", the calculation formula is as follows: In the formula, is the difference element ΔL i The weight of .

Images