A method for replenishing power based on clustering and PQUI identification algorithm

Abstract

The present invention discloses a method for compensating electricity consumption based on clustering and PQUI identification algorithms, and relates to the field of power grid operation and maintenance. At present, the amount of electricity stolen cannot be accurately calculated, and the calculated amount of electricity stolen differs too much from the actual amount of electricity stolen, making it difficult to maintain fairness between users and power companies. The present invention first determines the user's electricity theft time based on the user's load data, then determines the electricity theft method based on the relationship between the load data, and then determines the correction coefficient based on the electricity theft behavior, and finally calculates the compensatory electricity consumption based on the correction coefficient; the clustering algorithm and / or PQUI algorithm are used to calculate the number of days of electricity theft. The technical solution uses a clustering algorithm and a PQUI identification algorithm to calculate the corresponding amount of electricity stolen, and the calculation matching is good, so that the accuracy is high, and the collected data can meet the requirements, and the compensatory electricity consumption of the electricity theft user can be accurately, objectively and quickly calculated, so that it is consistent with the actual amount of electricity, thereby improving the fairness of electricity use.

Description

Technical Field

[0001] The present invention relates to the field of power grid operation and maintenance, and in particular to a method for replenishing electricity based on clustering and PQUI recognition algorithms. Background Art

[0002] Year-on-year increases in electricity production and consumption have significantly driven the development of electricity management technologies, but this has also led to an increasingly prominent problem of electricity theft. This problem not only raises concerns about electricity safety but also leads to unfairness issues among users and direct economic losses for power companies. Currently, the calculation of electricity replenishment is mostly based on manual analysis and estimation, but this method is too flexible and lacks persuasiveness, making it difficult to convince people. Therefore, a reasonable method is needed to identify a reasonable electricity theft range from historical load data and accurately calculate the replenishment electricity.

[0003] Combining multiple methods for determining the time of electricity theft and the user's electricity consumption curve can more accurately determine the time of electricity theft. Due to the large number of users, different electricity needs, and various theft methods, a single method for determining the time of electricity theft may not be able to accurately determine the time of electricity theft. Therefore, we need to use multiple methods.

[0004] After determining the time of electricity theft, the method of electricity theft needs to be confirmed. Due to the different methods of electricity theft, the correction coefficient of each user is also different, making it impossible to accurately calculate the amount of electricity stolen. The calculated amount of electricity stolen is too different from the actual amount of electricity stolen, making it difficult to maintain fairness between users and power companies. Summary of the Invention

[0005] The technical problem to be solved and the technical task proposed by the present invention are to improve and perfect the existing technical solutions and provide a method for replenishing the amount of electricity based on clustering and PQUI recognition algorithm to achieve the purpose of accurately calculating the replenished amount of electricity. To this end, the present invention adopts the following technical solutions.

[0006] A method for compensating electricity consumption based on clustering and PQUI recognition algorithms is implemented using a remote monitoring and diagnostic system for electric energy metering devices. The remote monitoring and diagnostic system for electric energy metering devices includes an electricity consumption information collection system, a data collection and analysis server, and a database server. The electricity consumption information collection system collects electricity consumption data and events from power users in real time. The data collection and analysis server analyzes and calculates electricity theft behaviors and corresponding compensatory electricity consumption based on the acquired data. The database server stores electricity consumption data, events, analysis and judgment thresholds, a library of electricity theft behavior characteristics, and correction coefficients for each electricity theft behavior. The data collection and analysis server includes the following steps when analyzing and calculating the compensatory electricity consumption:

[0007] 1) Determine whether the amount of electricity theft depends on the duration of the electricity consumption; if not, calculate directly based on the difference between the electricity consumption recorded on site and the electricity consumption finally settled by the power supply company; if it does, proceed to the next step;

[0008] 2) Calculate the number of days of electricity theft and determine whether the number of days of electricity theft can be determined; if so, directly calculate the amount of electricity theft based on the determined number of days; if not, calculate the amount of electricity theft based on the preset number of days; the calculation of the number of days of electricity theft adopts a clustering algorithm and / or a PQUI algorithm;

[0009] 3) Identify electricity theft behavior; obtain electricity usage data of suspected electricity theft users; compare the electricity theft behavior feature library under the corresponding metering method to match the most likely electricity theft behavior;

[0010] 4) Query the database server and check whether there is a correction coefficient for the electricity theft behavior. If so, determine the actual daily electricity consumption based on the corresponding correction coefficient, and then add the daily electricity consumption during the electricity theft time to obtain the actual electricity consumption; if not, proceed to the next step;

[0011] 5) Determine whether the daily electricity consumption forecast requirements are met; if not, replace the actual load with the capacity indicated by the billing electricity meter. Then, determine the daily electricity consumption time based on the production and operation electricity consumption and the daily electricity consumption of different electricity users, and multiply it by the electricity theft time to obtain the amount of electricity theft. If so, use the algorithm to predict the daily electricity consumption, and calculate the amount of electricity theft as of the day of the theft by multiplying the daily electricity consumption by the number of days of electricity theft.

[0012] 6) Storing the calculated amount of stolen electricity and the basis for the calculation, when the difference between the calculated amount of stolen electricity and the actual amount of stolen electricity exceeds a set threshold, modifying the correction coefficient of the database server.

[0013] As a preferred technical means: in step 2) calculating the number of days of electricity theft, the specific steps of using the clustering algorithm and / or the PQUI algorithm are as follows;

[0014] 201) Clustering algorithm: Electricity consumption is divided into three categories. Based on the cluster center, electricity consumption is divided into label 1, label 2, and label 3 from high to low. Label 3 is abnormal electricity. For label 3, if the time corresponding to the electricity consumption of label 3, that is, the abnormal time, occurs for more than 15 consecutive days, it is considered that the continuous abnormal time is electricity theft time.

[0015] 202) PQUI algorithm: performs relevant calculations based on the user's load and extracts useful features; the three-phase four-wire calculation formula is as follows:

[0016] S1=U a I a +U b I b +U c I c ……(1)

[0017]

[0018] K=(S1-S2) / W……(3)

[0019] Where U a , U b , U c , I a , I b , I c are three-phase voltage and three-phase current respectively; W is power consumption; P, Q are active power and reactive power respectively; S1, S2 are apparent power;

[0020] For normal users, their electricity usage does not change much, and the voltage and current values do not fluctuate too much, so the K value will not fluctuate too much and will remain in a stable range. When a user steals electricity, the change in electricity usage behavior directly leads to a change in the relationship between the loads. S1 and S2 decrease, the difference between the two increases, the power consumption W decreases, and the ratio K increases. Therefore, when the K value is much larger than the K value during normal electricity usage, it is considered that the electricity usage on that day is abnormal. If abnormal K values appear for 15 consecutive days, the user is considered to have stolen electricity.

[0021] The PQUI method requires electricity, voltage, current, and power for calculation, so it is not suitable for low-voltage users and can only be calculated for dedicated transformer users; clustering only requires electricity data, so it can be used by both low-voltage and dedicated transformer users; therefore, when the user is a low-voltage user, the clustering algorithm is used to determine the time of electricity theft; when the user is a dedicated transformer user, additional conditions are required; the PQUI identification algorithm uses the K value fluctuation to determine whether the electricity consumption is abnormal, so when the fluctuation is small, it cannot make an accurate judgment; the clustering algorithm directly judges based on the continuous time of the lowest class, avoiding the disadvantage of too small fluctuations, but because it requires multiple iterations, it takes a lot of time to calculate; in order to avoid the disadvantages of the two algorithms, the parameter D is added to calculate the fluctuation through distance, and then the appropriate algorithm is selected based on D; the specific calculation is as follows:

[0022] Normalize the data first:

[0023]

[0024] Where W i (i=1,2,3,…,n) is a set of power data of the user, W min is the minimum value in the data, W max is the maximum value in the data;

[0025] Then calculate the parameter D:

[0026] D=|W i* -mean(W * )| / n……(5)

[0027] Where W * W i * The sum of all data in, mean(W * ) is the average value of the normalized data, and n is the number of data;

[0028] The threshold is set to 0.2. When D≥0.2, the PQUI algorithm is used; when D<0.2, the clustering algorithm is used.

[0029] After a large number of experimental calculations and comparisons, the threshold value is 0.2, and the calculation result is more accurate and best matches the actual electricity theft time.

[0030] As a preferred technical means: the remote monitoring and diagnosis system for electric energy metering devices further includes a module for establishing a feature library of electric power theft behavior. When establishing the feature library of electric power theft behavior, the module includes the following steps:

[0031] A1) Obtain historical electricity theft data;

[0032] A2) Processing historical electricity theft data; obtaining electricity theft methods, assigning numbers to each theft method, and classifying them; the theft methods include: electricity theft without a meter, voltage circuit disconnection, poor voltage circuit contact, voltage circuit division, current circuit open circuit, current circuit short circuit, current circuit shunting, phase shifting, altering the internal structure of the meter, damaging the meter with high current or mechanical force, and external interference;

[0033] A3) Establish a corresponding electricity theft behavior feature database for the three metering methods: high supply high metering, high supply low metering, and low supply low metering; among which:

[0034] The data in the electricity theft behavior feature database under the high-supply high-metering metering mode includes the category, the output electricity theft method corresponding to the category, phase A voltage, phase C voltage, phase A current, phase C current and active power. When judging electricity theft behavior, the fault phase and the electricity theft method are determined based on the phase voltage, phase current and active power.

[0035] The data in the electricity theft behavior feature database under the high-supply, low-metering metering method includes the category, the output electricity theft method corresponding to the category, phase A voltage, phase B voltage, phase C voltage, phase A current, phase B current, phase C current, and active power. When judging electricity theft behavior, the fault phase and the electricity theft method are determined based on the phase voltage, phase current, and active power.

[0036] The data of the electricity theft behavior feature library under the low supply and low metering mode includes categories, output electricity theft methods and power consumption corresponding to the categories; when judging electricity theft behavior, the fault phase and electricity theft method are determined based on the power consumption.

[0037] This technical solution proposes establishing a comprehensive database of electricity theft behavior characteristics based on three typical electricity usage types: high-supply high-metering, high-supply low-metering, and low-supply low-metering. Based on this, a comprehensive matching mechanism is established to match the data of users suspected of electricity theft with electricity theft behavior, identify possible theft, and provide technical support for theft verification, effectively improving the efficiency of theft identification and reducing labor costs.

[0038] As the preferred technical means: electricity theft methods classified as non-meter electricity theft include: breaking the voltage transformer Q1, bypassing the transformer and bridging Q2, and adding a bypass to bypass the electricity meter Q3;

[0039] Theft methods classified as voltage circuit disconnection include: loosening the TV fuse Q4, breaking the fuse in the fuse tube Q5, loosening the voltage circuit terminal Q6, breaking the core of the voltage circuit wire Q7, and loosening the voltage connector Q8 of the energy meter;

[0040] Theft methods classified as poor voltage circuit contact include: loosening the voltage connector Q9 of the energy meter, loosening the voltage circuit terminal Q10, and loosening the low-voltage fuse Q11 of the TV;

[0041] The electricity theft methods classified as voltage circuit division include: inserting a resistor Q12 in series in the secondary circuit of the TV, disconnecting the neutral wire on the incoming line of a single-phase meter and inserting a voltage-reducing resistor Q13 in series between the outgoing line and the ground;

[0042] Theft methods classified as open current loop include: loosening the TA secondary output terminal (Q14), artificially creating a poor contact fault in the TA secondary circuit terminal (Q15), breaking the core of the current loop conductor (Q16), and breaking the neutral wire (Q17).

[0043] The electricity theft methods classified as current loop short-circuit electricity theft methods include: short-circuiting the current terminal Q18 of the energy meter, short-circuiting the terminal block Q19 in the current loop, and short-circuiting the primary or secondary side Q20 of the TA;

[0044] The electricity theft methods classified as current loop shunting include: replacing TAs with different transformation ratios Q21, changing the secondary tap of tapped TAs Q22, and changing the number of primary turns of through-core TAs Q23;

[0045] The electricity theft methods classified as phase-shifting theft include: swapping the phase and neutral wires of a single-phase meter, using the ground wire as the neutral wire (Q24); swapping the primary side input and output wires of a TA (Q25); swapping the same-name terminals of a TA (Q26); swapping the current terminals of a TA (Q27); swapping the phases of the TA-to-meter wiring (Q28); swapping the primary or secondary polarity of a TV (Q29); swapping the phases of the TV-to-meter wiring (Q30); using special inductors or capacitors for phase shifting (Q31);

[0046] Theft methods classified as altering the internal structure of the meter include: reducing the number of turns of the current coil (Q32); spot welding the manganese copper resistor and cutting the manganese copper signal line (Q33); connecting resistors in parallel or series with the current sampling circuit (Q34); replacing the graded sampling resistors in the voltage sampling circuit (Q35); connecting the voltage coil with resistors and other electronic components for voltage division (Q36); short-circuiting with copper wire hooks (Q37); and implanting remote control for current diversion (Q38).

[0047] Theft methods are classified as using large current or mechanical force to damage the meter. These methods include: burning the current coil Q39 with overload current, shocking the meter Q40 with the electric force of short-circuit current, and damaging the meter Q41 with mechanical external force.

[0048] The electricity theft methods classified as external interference include: strong magnetic interference theft Q42, high-frequency interference theft Q43, high-voltage pulse theft Q44, and short-circuiting the inlet and outlet lines of the meter box Q45.

[0049] As the preferred technical means: the characteristic database of electricity theft behavior under the high supply and high metering mode is:

[0050]

[0051]

[0052] The characteristic database of electricity theft behavior under the high supply and low metering mode is:

[0053]

[0054]

[0055]

[0056] The characteristic database of electricity theft behavior under the low supply and low metering mode is:

[0057]

[0058] As the preferred technical means: the judgment of electricity theft behavior includes the identification of electricity theft behavior under the high supply and low metering mode, the identification of electricity theft behavior under the high supply and high metering mode, and the identification of electricity theft behavior under the low supply and low metering mode;

[0059] The identification of electricity theft under the high-supply and low-metering method includes the following steps:

[0060] 3101) Inputting electricity usage data of a user suspected of electricity theft;

[0061] 3102) Determine whether the current and voltage of phases A, B, and C have values; if not, consider it impossible to determine and end; if so, proceed to the next step;

[0062] 3103) Determine whether the three-phase voltage and current are close to 0; if so, compare the electricity theft behavior feature library under the high-supply, low-metering metering mode, output 9 types of electricity theft reasons, and end; if not, proceed to the next step;

[0063] 3104) Determine whether one or more phases of the three-phase current are close to zero; if so, compare the power theft behavior feature library under the high-supply, low-metering metering mode, output Class 1 power theft cause and the power theft phase with output current close to zero, and end; if not, proceed to the next step;

[0064] 3105) Determine whether one or more phases of the three-phase voltage are close to zero; if so, compare the voltage with the electricity theft behavior feature library under the high-supply, low-metering metering mode, output two types of electricity theft reasons and the phases with the electricity theft voltage close to zero, and end; if not, proceed to the next step;

[0065] 3106) Call for phase power factor measurement;

[0066] 3107) Determine whether all three-phase power factors are normal; if not, proceed to step 2110; if so, proceed to the next step;

[0067] 3108) Determine whether all three-phase voltages are greater than or close to 220V; if not, compare the three-phase voltages with the electricity theft behavior feature library under the high-supply, low-metering metering mode, output the four types of electricity theft causes and the phase with the theft voltage that is too low, and end; if yes, proceed to the next step;

[0068] 3109) Determine whether the three-phase currents are equal. If not, compare the three-phase currents with the electricity theft behavior feature library under the high-supply, low-metering metering mode, output the three types of electricity theft causes and the phase with the smaller output current, and end; if so, output the three types of electricity theft causes and the phase with the smaller output current, and end;

[0069] 3110) Determine whether the user's daily electricity consumption is 0; if so, compare it with the electricity theft behavior feature library under the high-supply and low-metering metering method, output 6 or 8 types of electricity theft reasons and the electricity theft phase with abnormal power factor, and end; if not, proceed to the next step;

[0070] 3111) Determine whether one or more phases of the three-phase current are less than 0; if so, compare the three-phase current with the electricity theft behavior feature library under the high-supply, low-metering mode, output five types of electricity theft causes and the phases with negative current, and terminate; if not, output seven types of electricity theft causes and the phases with negative voltage, and terminate;

[0071] Identifying electricity theft under high-supply and high-metering mode includes the following steps:

[0072] 3201) Input the electricity usage data of the suspected electricity thief

[0073] 3202) Determine whether the current and voltage of phases A and C have values. If not, consider that the determination cannot be made and terminate.

[0074] 3203) Determine whether the voltage and current of phases A and C are both close to 0; if so, compare the electricity theft behavior feature library under the high-supply high-metering method, output 9 types of electricity theft causes, and end; if not, proceed to the next step;

[0075] 3204) Determine whether one or more phases of the voltage of phases A and C are close to zero; if so, compare the power theft behavior feature library under the high-supply high-metering metering mode, output Class 1 power theft cause and the power theft phase with voltage close to zero, and end; if not, proceed to the next step;

[0076] 3205) Determine whether one or more phases of the A and C currents are close to zero; if so, compare the power theft behavior feature library under the high-supply high-metering metering mode and output two types of power theft causes and the power theft phases with power close to zero; if not, proceed to the next step;

[0077] 3206) Call for phase power factor measurement;

[0078] 3207) Determine whether the power factors of phases A and C are all normal; if not, proceed to step 2210; if so, proceed to the next step;

[0079] 3208) Determine whether the voltages of phases A and C are both greater than or close to 220V; if not, compare the power theft behavior feature library under the high-supply high-metering metering mode, output the three types of power theft causes and the power theft phase with too low voltage, and end; if yes, proceed to the next step;

[0080] 3209) Determine whether the currents of phases A and C are equal; if so, compare the electricity theft behavior feature library under the high-supply, high-meter metering mode, output the four types of electricity theft causes and the electricity theft phases A and C, and end; if not, compare the electricity theft behavior feature library under the high-supply, high-meter metering mode, output the four types of electricity theft causes and the electricity theft phase with smaller current, and end;

[0081] 3210) Determine whether the voltage of phases A and C is less than 0; if so, compare the voltage of phases A and C with the voltage of phases C and C under the high-voltage power supply and high-voltage power metering mode, output the 7 types of power theft reasons and the power theft phases with voltage less than 0, and end; if not, proceed to the next step;

[0082] 3211) Determine whether the current of phases A and C is less than 0; if so, compare the power theft behavior feature library under the high-supply high-metering metering mode, output the five types of power theft causes and the power theft phases with current less than 0, and end; if not, proceed to the next step;

[0083] 3212) Determine whether the user's daily electricity consumption is 0; if so, compare the electricity theft behavior feature library under the high-supply and high-metering metering mode, output 6 or 8 types of electricity theft causes and the electricity theft phases AC or AB, and end; if not, compare the electricity theft behavior feature library under the high-supply and high-metering metering mode, output 8 types of electricity theft causes and the electricity theft phases BC, and end;

[0084] The identification of electricity theft under the low-supply and low-metering method includes the following steps:

[0085] 3301) Input the electricity usage data of the suspected electricity thief

[0086] 3302) Determine whether the user's daily electricity consumption has a value; if not, consider it impossible to determine and end; if so, proceed to the next step;

[0087] 3303) Determine whether the user's daily electricity consumption is close to 0; if so, compare the electricity theft behavior feature library under the low supply and low metering mode, output Class 1 electricity theft reason, and end; if not, proceed to the next step;

[0088] 3304) Determine whether the user's daily electricity consumption is negative; if so, compare the electricity theft behavior feature library under the low supply and low metering method, output two types of electricity theft causes, and end; if not, compare the electricity theft behavior feature library under the low supply and low metering method, output three types of electricity theft causes, and end.

[0089] As a preferred technical means: the database server stores electricity theft behaviors and correction coefficients for each electricity theft behavior, including correction coefficients for dedicated transformer high-voltage power supply and high-metering users, correction coefficients for dedicated transformer high-voltage power supply and low-metering users, and correction coefficients for low-voltage users;

[0090] The correction coefficients for high-power and high-meter users of dedicated transformers are shown in the following table:

[0091]

[0092]

[0093]

[0094] The correction coefficients for high-supply and low-cost users of dedicated transformers are shown in the following table:

[0095]

[0096]

[0097]

[0098] The correction factors for low-voltage users are shown in the following table:

[0099]

[0100] Beneficial Effects: This technical solution addresses the current drawbacks of relying solely on estimates for electricity consumption, which suffer from low accuracy, lack of persuasiveness, and significant labor consumption. This technical solution utilizes a clustering algorithm and a PQUI identification algorithm to calculate the corresponding amount of stolen electricity. This provides good matching and high accuracy, and the collected data meets the requirements. This allows for accurate, objective, and rapid calculation of the amount of electricity stolen by electricity users, ensuring consistency with the actual amount of electricity consumed, thereby improving fairness in electricity use. BRIEF DESCRIPTION OF THE DRAWINGS

[0101] Figure 1 It is a flow chart of the present invention.

[0102] Figure 2 It is a line graph of daily electricity consumption of a low-voltage user.

[0103] Figure 3 This is the result graph of the user's electricity consumption after the clustering algorithm.

[0104] Figure 4 This is a line graph of daily electricity consumption of a dedicated transformer user.

[0105] Figure 5 This is the PQUI result diagram for dedicated users. DETAILED DESCRIPTION

[0106] The technical solution of the present invention is further described in detail below with reference to the accompanying drawings.

[0107] like Figure 1 As shown, a method for compensating electricity consumption based on clustering and PQUI recognition algorithms is implemented using a remote monitoring and diagnostic system for electric energy metering devices. The remote monitoring and diagnostic system for electric energy metering devices includes an electricity consumption information collection system, a data collection and analysis server, and a database server. The electricity consumption information collection system collects electricity consumption data and events from power users in real time. The data collection and analysis server analyzes and calculates electricity theft behaviors and corresponding compensatory electricity consumption based on the acquired data. The database server stores electricity consumption data, events, analysis and judgment thresholds, a library of electricity theft behavior characteristics, and correction coefficients for each electricity theft behavior. The data collection and analysis server includes the following steps when analyzing and calculating the compensatory electricity consumption:

[0108] 1) Determine whether the amount of electricity theft depends on the duration of the electricity consumption; if not, calculate directly based on the difference between the electricity consumption recorded on site and the electricity consumption finally settled by the power supply company; if it does, proceed to the next step;

[0109] 2) Determine whether the number of days of electricity theft can be determined; if so, calculate directly based on the determined number of days; if not, calculate based on the preset number of days;

[0110] 3) Identify electricity theft behavior; obtain electricity usage data of suspected electricity theft users; compare the electricity theft behavior feature library under the corresponding metering method to match the most likely electricity theft behavior;

[0111] 4) Query the database server and check whether there is a correction coefficient for the electricity theft behavior. If so, determine the actual daily electricity consumption based on the corresponding correction coefficient, and then add the daily electricity consumption during the electricity theft time to obtain the actual electricity consumption; if not, proceed to the next step;

[0112] 5) Determine whether the daily electricity consumption forecast requirements are met; if not, replace the actual load with the capacity indicated by the billing electricity meter. Then, determine the daily electricity consumption time based on the production and operation electricity consumption and the daily electricity consumption of different electricity users, and multiply it by the electricity theft time to obtain the amount of electricity theft. If so, use the algorithm to predict the daily electricity consumption, and calculate the amount of electricity theft as of the day of the theft by multiplying the daily electricity consumption by the number of days of electricity theft.

[0113] 6) Storing the calculated amount of stolen electricity and the basis for the calculation, when the difference between the calculated amount of stolen electricity and the actual amount of stolen electricity exceeds a set threshold, modifying the correction coefficient of the database server.

[0114] In step 2), when determining the number of days of electricity theft, a clustering algorithm and / or a PQUI algorithm is used;

[0115] 201) When calculating the number of days of electricity theft in step 2), the specific steps of using the clustering algorithm and / or the PQUI algorithm are as follows:

[0116] 201) Clustering algorithm: Electricity consumption is divided into three categories. Based on the cluster center, electricity consumption is divided into label 1, label 2, and label 3 from high to low. Label 3 is abnormal electricity. For label 3, if the time corresponding to the electricity consumption of label 3, that is, the abnormal time, occurs for more than 15 consecutive days, it is considered that the continuous abnormal time is electricity theft time.

[0117] 202) PQUI algorithm: performs relevant calculations based on the user's load and extracts useful features; the three-phase four-wire calculation formula is as follows:

[0118] S1=U a I a +U b I b +U c I c ……(1)

[0119]

[0120] K=(S1-S2) / W……(3)

[0121] Where U a , U b , U c , I a , I b , I c are three-phase voltage and three-phase current respectively; W is power consumption; P, Q are active power and reactive power respectively; S1, S2 are apparent power;

[0122] For normal users, their electricity usage does not change much, and the voltage and current values do not fluctuate too much, so the K value will not fluctuate too much and will remain in a stable range. When a user steals electricity, the change in electricity usage behavior directly leads to a change in the relationship between the loads. S1 and S2 decrease, the difference between the two increases, the power consumption W decreases, and the ratio K increases. Therefore, when the K value is much larger than the K value during normal electricity usage, it is considered that the electricity usage on that day is abnormal. If abnormal K values appear for 15 consecutive days, the user is considered to have stolen electricity.

[0123] The PQUI method requires electricity, voltage, current, and power for calculation, so it is not suitable for low-voltage users and can only be calculated for dedicated transformer users; clustering only requires electricity data, so it can be used by both low-voltage and dedicated transformer users; therefore, when the user is a low-voltage user, the clustering algorithm is used to determine the time of electricity theft; when the user is a dedicated transformer user, additional conditions are required; the PQUI identification algorithm uses the K value fluctuation to determine whether the electricity consumption is abnormal, so when the fluctuation is small, it cannot make an accurate judgment; the clustering algorithm directly judges based on the continuous time of the lowest class, avoiding the disadvantage of too small fluctuations, but because it requires multiple iterations, it takes a lot of time to calculate; in order to avoid the disadvantages of the two algorithms, the parameter D is added to calculate the fluctuation through distance, and then the appropriate algorithm is selected based on D; the specific calculation is as follows:

[0124] Normalize the data first:

[0125]

[0126] Where W i (i=1,2,3,…,n) is a set of power data of the user, W min is the minimum value in the data, W max is the maximum value in the data;

[0127] Then calculate the parameter D:

[0128] D=|W i * -mean(W * )| / n……(5)

[0129] Where W* W i * The sum of all data in, mean(W * ) is the average value of the normalized data, and n is the number of data;

[0130] The threshold is set to 0.2. When D≥0.2, the PQUI algorithm is used; when D<0.2, the clustering algorithm is used.

[0131] After a large number of experimental calculations and comparisons, the threshold value is 0.2, and the calculation result is more accurate and best matches the actual electricity theft time.

[0132] The remote monitoring and diagnosis system for electric energy metering devices also includes a module for establishing a feature library of electric energy theft behavior. When establishing the feature library of electric energy theft behavior, the module includes the following steps:

[0133] A1) Obtain historical electricity theft data;

[0134] A2) Processing historical electricity theft data; obtaining electricity theft methods, assigning numbers to each theft method, and classifying them; the theft methods include: electricity theft without a meter, voltage circuit disconnection, poor voltage circuit contact, voltage circuit division, current circuit open circuit, current circuit short circuit, current circuit shunting, phase shifting, altering the internal structure of the meter, damaging the meter with high current or mechanical force, and external interference;

[0135] A3) Establish a corresponding electricity theft behavior feature database for the three metering methods: high supply high metering, high supply low metering, and low supply low metering; among which:

[0136] The data in the electricity theft behavior feature database under the high-supply high-metering metering mode includes the category, the output electricity theft method corresponding to the category, phase A voltage, phase C voltage, phase A current, phase C current and active power. When judging electricity theft behavior, the fault phase and the electricity theft method are determined based on the phase voltage, phase current and active power.

[0137] The data in the electricity theft behavior feature database under the high-supply, low-metering metering method includes the category, the output electricity theft method corresponding to the category, phase A voltage, phase B voltage, phase C voltage, phase A current, phase B current, phase C current, and active power. When judging electricity theft behavior, the fault phase and the electricity theft method are determined based on the phase voltage, phase current, and active power.

[0138] The data of the electricity theft behavior feature library under the low supply and low metering mode includes categories, output electricity theft methods and power consumption corresponding to the categories; when judging electricity theft behavior, the fault phase and electricity theft method are determined based on the power consumption.

[0139] The electricity theft methods classified as non-metered electricity theft include: disconnecting the voltage transformer Q1, bypassing the transformer and bridging Q2, and adding a bypass to bypass the electricity meter Q3;

[0140] Theft methods classified as voltage circuit disconnection include: loosening the TV fuse Q4, breaking the fuse in the fuse tube Q5, loosening the voltage circuit terminal Q6, breaking the core of the voltage circuit wire Q7, and loosening the voltage connector Q8 of the energy meter;

[0141] Theft methods classified as poor voltage circuit contact include: loosening the voltage connector Q9 of the energy meter, loosening the voltage circuit terminal Q10, and loosening the low-voltage fuse Q11 of the TV;

[0142] The electricity theft methods classified as voltage circuit division include: inserting a resistor Q12 in series in the secondary circuit of the TV, disconnecting the neutral wire on the incoming line of a single-phase meter and inserting a voltage-reducing resistor Q13 in series between the outgoing line and the ground;

[0143] Theft methods classified as open current loop include: loosening the TA secondary output terminal (Q14), artificially creating a poor contact fault in the TA secondary circuit terminal (Q15), breaking the core of the current loop conductor (Q16), and breaking the neutral wire (Q17).

[0144] The electricity theft methods classified as current loop short-circuit electricity theft methods include: short-circuiting the current terminal Q18 of the energy meter, short-circuiting the terminal block Q19 in the current loop, and short-circuiting the primary or secondary side Q20 of the TA;

[0145] The electricity theft methods classified as current loop shunting include: replacing TAs with different transformation ratios Q21, changing the secondary tap of tapped TAs Q22, and changing the number of primary turns of through-core TAs Q23;

[0146] The electricity theft methods classified as phase-shifting theft include: swapping the phase and neutral wires of a single-phase meter, using the ground wire as the neutral wire (Q24); swapping the primary side input and output wires of a TA (Q25); swapping the same-name terminals of a TA (Q26); swapping the current terminals of a TA (Q27); swapping the phases of the TA-to-meter wiring (Q28); swapping the primary or secondary polarity of a TV (Q29); swapping the phases of the TV-to-meter wiring (Q30); using special inductors or capacitors for phase shifting (Q31);

[0147] Theft methods classified as altering the internal structure of the meter include: reducing the number of turns of the current coil (Q32); spot welding the manganese copper resistor and cutting the manganese copper signal line (Q33); connecting resistors in parallel or series with the current sampling circuit (Q34); replacing the graded sampling resistors in the voltage sampling circuit (Q35); connecting the voltage coil with resistors and other electronic components for voltage division (Q36); short-circuiting with copper wire hooks (Q37); and implanting remote control for current diversion (Q38).

[0148] Theft methods are classified as using large current or mechanical force to damage the meter. These methods include: burning the current coil Q39 with overload current, shocking the meter Q40 with the electric force of short-circuit current, and damaging the meter Q41 with mechanical external force.

[0149] The electricity theft methods classified as external interference include: strong magnetic interference theft Q42, high-frequency interference theft Q43, high-voltage pulse theft Q44, and short-circuiting the inlet and outlet lines of the meter box Q45.

[0150] The electricity theft behavior is expressed in a table as follows:

[0151] Table 1 Electricity theft behavior table

[0152]

[0153]

[0154] The characteristic database of electricity theft behavior under the high supply and high metering mode is:

[0155]

[0156]

[0157]

[0158] The characteristic database of electricity theft behavior under the high supply and low metering mode is:

[0159]

[0160]

[0161] The characteristic database of electricity theft behavior under the low supply and low metering mode is:

[0162]

[0163]

[0164] The judgment of electricity theft includes the identification of electricity theft under high supply and low metering mode, the identification of electricity theft under high supply and high metering mode, and the identification of electricity theft under low supply and low metering mode;

[0165] The identification of electricity theft under the high-supply and low-metering method includes the following steps:

[0166] 3101) Inputting electricity usage data of a user suspected of electricity theft;

[0167] 3102) Determine whether the current and voltage of phases A, B, and C have values; if not, consider it impossible to determine and end; if so, proceed to the next step;

[0168] 3103) Determine whether the three-phase voltage and current are close to 0; if so, compare the electricity theft behavior feature library under the high-supply, low-metering metering mode, output 9 types of electricity theft reasons, and end; if not, proceed to the next step;

[0169] 3104) Determine whether one or more phases of the three-phase current are close to zero; if so, compare the power theft behavior feature library under the high-supply, low-metering metering mode, output Class 1 power theft cause and the power theft phase with output current close to zero, and end; if not, proceed to the next step;

[0170] 3105) Determine whether one or more phases of the three-phase voltage are close to zero; if so, compare the voltage with the electricity theft behavior feature library under the high-supply, low-metering metering mode, output two types of electricity theft reasons and the phases with the electricity theft voltage close to zero, and end; if not, proceed to the next step;

[0171] 3106) Call for phase power factor measurement;

[0172] 3107) Determine whether all three-phase power factors are normal; if not, proceed to step 2110; if so, proceed to the next step;

[0173] 3108) Determine whether all three-phase voltages are greater than or close to 220V; if not, compare the three-phase voltages with the electricity theft behavior feature library under the high-supply, low-metering metering mode, output the four types of electricity theft causes and the phase with the theft voltage that is too low, and end; if yes, proceed to the next step;

[0174] 3109) Determine whether the three-phase currents are equal. If not, compare the three-phase currents with the electricity theft behavior feature library under the high-supply, low-metering metering mode, output the three types of electricity theft causes and the phase with the smaller output current, and end; if so, output the three types of electricity theft causes and the phase with the smaller output current, and end;

[0175] 3110) Determine whether the user's daily electricity consumption is 0; if so, compare it with the electricity theft behavior feature library under the high-supply and low-metering metering method, output 6 or 8 types of electricity theft reasons and the electricity theft phase with abnormal power factor, and end; if not, proceed to the next step;

[0176] 3111) Determine whether one or more phases of the three-phase current are less than 0; if so, compare the three-phase current with the electricity theft behavior feature library under the high-supply, low-metering mode, output five types of electricity theft causes and the phases with negative current, and terminate; if not, output seven types of electricity theft causes and the phases with negative voltage, and terminate;

[0177] Identifying electricity theft under high-supply and high-metering mode includes the following steps:

[0178] 3201) Input the electricity usage data of the suspected electricity thief

[0179] 3202) Determine whether the current and voltage of phases A and C have values. If not, consider that the determination cannot be made and terminate.

[0180] 3203) Determine whether the voltage and current of phases A and C are both close to 0; if so, compare the electricity theft behavior feature library under the high-supply high-metering method, output 9 types of electricity theft causes, and end; if not, proceed to the next step;

[0181] 3204) Determine whether one or more phases of the voltage of phases A and C are close to zero; if so, compare the power theft behavior feature library under the high-supply high-metering metering mode, output Class 1 power theft cause and the power theft phase with voltage close to zero, and end; if not, proceed to the next step;

[0182] 3205) Determine whether one or more phases of the A and C currents are close to zero; if so, compare the power theft behavior feature library under the high-supply high-metering metering mode and output two types of power theft causes and the power theft phases with power close to zero; if not, proceed to the next step;

[0183] 3206) Call for phase power factor measurement;

[0184] 3207) Determine whether the power factors of phases A and C are all normal; if not, proceed to step 2210; if so, proceed to the next step;

[0185] 3208) Determine whether the voltages of phases A and C are both greater than or close to 220V; if not, compare the power theft behavior feature library under the high-supply high-metering metering mode, output the three types of power theft causes and the power theft phase with too low voltage, and end; if yes, proceed to the next step;

[0186] 3209) Determine whether the currents of phases A and C are equal; if so, compare the electricity theft behavior feature library under the high-supply, high-meter metering mode, output the four types of electricity theft causes and the electricity theft phases A and C, and end; if not, compare the electricity theft behavior feature library under the high-supply, high-meter metering mode, output the four types of electricity theft causes and the electricity theft phase with smaller current, and end;

[0187] 3210) Determine whether the voltage of phases A and C is less than 0; if so, compare the voltage of phases A and C with the voltage of phases C and C under the high-voltage power supply and high-voltage power metering mode, output the 7 types of power theft reasons and the power theft phases with voltage less than 0, and end; if not, proceed to the next step;

[0188] 3211) Determine whether the current of phases A and C is less than 0; if so, compare the power theft behavior feature library under the high-supply high-metering metering mode, output the five types of power theft causes and the power theft phases with current less than 0, and end; if not, proceed to the next step;

[0189] 3212) Determine whether the user's daily electricity consumption is 0; if so, compare the electricity theft behavior feature library under the high-supply and high-metering metering mode, output 6 or 8 types of electricity theft causes and the electricity theft phases AC or AB, and end; if not, compare the electricity theft behavior feature library under the high-supply and high-metering metering mode, output 8 types of electricity theft causes and the electricity theft phases BC, and end;

[0190] The identification of electricity theft under the low-supply and low-metering method includes the following steps:

[0191] 3301) Input the electricity usage data of the suspected electricity thief

[0192] 3302) Determine whether the user's daily electricity consumption has a value; if not, consider it impossible to determine and end; if so, proceed to the next step;

[0193] 3303) Determine whether the user's daily electricity consumption is close to 0; if so, compare the electricity theft behavior feature library under the low supply and low metering mode, output Class 1 electricity theft reason, and end; if not, proceed to the next step;

[0194] 3304) Determine whether the user's daily electricity consumption is negative; if so, compare the electricity theft behavior feature library under the low supply and low metering method, output two types of electricity theft causes, and end; if not, compare the electricity theft behavior feature library under the low supply and low metering method, output three types of electricity theft causes, and end.

[0195] The database server stores electricity theft behaviors and correction coefficients for each electricity theft behavior. The correction coefficients include the correction coefficients for high-voltage power supply and high-metering users of dedicated transformers, the correction coefficients for high-voltage power supply and low-metering users of dedicated transformers, and the correction coefficients for low-voltage users.

[0196] The correction coefficients for high-power and high-meter users of dedicated transformers are shown in Table 2 below:

[0197] Table 2 Correction coefficients for high-voltage power supply and high-voltage meter users

[0198]

[0199]

[0200]

[0201] The correction coefficients for high-supply and low-cost users of dedicated transformers are shown in Table 3:

[0202] Table 3 Correction coefficients for high-supply and low-cost users of special transformers

[0203]

[0204]

[0205]

[0206] The correction coefficients for low-voltage users are shown in Table 4:

[0207] Table 4 Correction coefficient table for low voltage users

[0208]

[0209] The following is a further explanation of this technical solution using specific examples:

[0210] 1. Calculation process of electricity consumption compensation for a low-voltage user based on clustering algorithm to determine the time of electricity theft

[0211] like Figure 2 As shown, obtain the electricity consumption data of a low-voltage user and generate Figure 2 The line graph of daily electricity consumption shown in Figure 3 The schematic diagram of the clustering results is shown.

[0212] For the lowest class number: 2 3 4 5 7 8 9 22 25 26 35 36 37 38 39 40 41 42 43 4445 46 ....... 674 675 676 677 678 679 680 681 704 808 810 811 813 (1 corresponds to 2017.01.01; 820 corresponds to 2018.12.31)

[0213] Analysis shows that the first low electricity consumption for more than 15 days started after 2017.02.04 and lasted for more than 60 days. Figure 2 As shown in the box, the time of electricity theft can be directly determined.

[0214] Result: 2017.02.04 is the date when the electricity theft started.

[0215] Calculation of retroactive electricity consumption: The user is assumed to have committed electricity theft on February 4, 2017. Since the user's electricity consumption was close to zero during the theft period, the theft method is likely one of the following: bypassing the energy meter (Q3), loosening the voltage circuit terminal (Q9), loosening the voltage connector (Q10) on the energy meter, short-circuiting the current terminal (Q18) on the energy meter, or short-circuiting with a copper wire hook (Q37). None of these methods can determine the correction factor. Therefore, in the retroactive electricity consumption process, the capacity indicated by the calibrated current value of the billing energy meter (rated voltage * calibrated current) should be used instead of the actual load (xKW). The number of days of electricity theft was 717 (2017.02.04-2019.01.23). The residential user's electricity consumption is calculated as 6 hours, and the retroactive electricity consumption is xKW*6*717=4296xKWh.

[0216] Actual replenishment power:

[0217] Daily power consumption: 3.7*6=22.2kwh

[0218] Total replenishment power: 22.2*180=3996kwh

[0219] Electricity fee: 3996*0.538=2149.85 yuan

[0220] Electricity fee for breach of contract: (3996*0.538)*3=6449.54 yuan

[0221] Total electricity bill: 2149.85 + 6449.54 = 8599.39 yuan

[0222] 2. Calculation process of electricity consumption compensation for a dedicated transformer user based on the PQUI algorithm to determine the time of electricity theft

[0223] Obtain the electricity consumption data of a dedicated transformer user and generate Figure 4 The line graph of daily electricity consumption of a dedicated transformer user is shown in Figure 5 As shown in FIG, the result diagram of the user PQUI discrimination method.

[0224] Determination of electricity theft time: Figure 4 It can be seen that the power consumption decreased from June 17 to August 3, 2016 (e.g. Figure 4 As shown in the box), the power consumption during this period is abnormal. Figure 5 It can be seen that, at the same time, the ratio increases (such as Figure 5 The ratio was significantly higher than normal for 50 consecutive days, indicating abnormal electricity usage during this period. Both factors indicate abnormal electricity usage during this period. Therefore, the electricity theft was determined to have occurred between June 17 and August 3, 2016.

[0225] Identification of electricity theft methods: Figure 4 It can be seen that during the electricity theft period, electricity consumption decreased by 2 / 3. Therefore, it is determined that the method used by the user is to reduce or cut off the current of two phases, which corresponds to Q14-Q20 in Table 1 and Category 1 in Table 3.

[0226] Electricity replenishment: When reducing the two-phase current, the correction factor is 3, so the replenishment electricity is:

[0227] W 追 =W 窃 *3-W 窃 =178212*3-178212=296416(W)

[0228] Actual catch-up results:

[0229] Accrued electricity = 3UI Cos∮*T

[0230] =3*225.5*612.9*0.98*1128 / 1000

[0231] =458345 degrees

[0232] Understated electricity consumption: 458345-161929 kWh.

[0233] above Figure 1 The method for replenishing power based on clustering and PQUI recognition algorithm shown is a specific embodiment of the present invention, which has reflected the essential characteristics and progress of the present invention. According to actual usage needs and under the guidance of the present invention, equivalent modifications in shape, structure, etc. can be made to it, which are all within the scope of protection of this solution.

Claims

1. A method for replenishing power based on clustering and PQUI recognition algorithm, characterized by: The system is implemented by using a remote monitoring and diagnosis system for electric energy metering devices, which includes an electric energy information collection system, a data collection and analysis server, and a database server; the electric energy information collection system collects electric energy user's electric energy data and events in real time; The data collection and analysis server analyzes and calculates the electricity theft behavior and the corresponding amount of electricity to be compensated based on the acquired data; the database server stores electricity usage data, events, analysis and judgment thresholds, a library of electricity theft behavior characteristics, and correction coefficients for each electricity theft behavior. The data collection and analysis server includes the following steps when analyzing and calculating the amount of electricity to be compensated: 1) Determine whether the amount of electricity theft depends on the duration of the electricity consumption; if not, calculate directly based on the difference between the electricity consumption recorded on site and the electricity consumption finally settled by the power supply company; if it does, proceed to the next step; 2) Calculate the number of days of electricity theft and determine whether the number of days of electricity theft can be determined; if so, directly calculate the amount of electricity theft based on the determined number of days; if not, calculate the amount of electricity theft based on a preset number of days. The calculation of the number of days of electricity theft uses a clustering algorithm and / or a PQUI identification algorithm. The PQUI identification algorithm requires electricity, voltage, current, and power to calculate, so it is not suitable for low-voltage users and can only be calculated for dedicated transformer users. The PQUI identification algorithm uses K value fluctuations to determine whether electricity usage is abnormal. 3) Identify electricity theft behavior; obtain electricity usage data of suspected electricity theft users; compare the electricity theft behavior feature library under the corresponding metering method to match the most likely electricity theft behavior; 4) Query the database server and check whether there is a correction coefficient for the electricity theft behavior. If so, determine the actual daily electricity consumption based on the corresponding correction coefficient, and then add the daily electricity consumption during the electricity theft time to obtain the actual electricity consumption; if not, proceed to the next step; 5) Determine whether the daily electricity consumption forecast requirements are met; if not, replace the actual load with the capacity indicated by the billing electricity meter. Then, determine the daily electricity consumption time based on the production and operation electricity consumption and the daily electricity consumption of different electricity users, and multiply it by the electricity theft time to obtain the stolen electricity amount. If so, use the algorithm to predict the daily electricity consumption, and calculate the stolen electricity amount as of the day by multiplying the daily electricity consumption by the number of days of electricity theft. 6) Storing the calculated amount of stolen electricity and the basis for the calculation, when the difference between the calculated amount of stolen electricity and the actual amount of stolen electricity exceeds a set threshold, modifying the correction coefficient of the database server.

2. The method for replenishing power based on clustering and PQUI recognition algorithm according to claim 1, characterized in that: When calculating the number of days of electricity theft in step 2), the specific steps of using the clustering algorithm and / or the PQUI identification algorithm are as follows: 201) Clustering algorithm: Electricity consumption is divided into three categories. Based on the cluster center, electricity consumption is divided into label 1, label 2, and label 3 from high to low. Label 3 is abnormal electricity. For label 3, if the time corresponding to the electricity consumption of label 3, that is, the abnormal time, occurs for more than 15 consecutive days, it is considered that the continuous abnormal time is electricity theft time. 202) PQUI Identification Algorithm: Perform relevant calculations based on the user's load; the three-phase four-wire calculation formula is as follows: S1=U a AND a +U b AND b +U c AND c ……(1) K=(S1-S2) / W……(3) Where U a ,U b ,U c ,I a ,I b ,I c are three-phase voltage and three-phase current respectively; W is power consumption; P, Q are active power and reactive power respectively; S1, S2 are apparent power; For normal users, their electricity consumption will not change much, and the voltage and current values will not fluctuate too much, so the K value will not fluctuate too much and will be in a stable range. When a user steals electricity, the change in their electricity usage directly changes the relationship between the loads. S1 and S2 decrease, the difference between the two increases, and the power consumption W decreases, so the ratio K increases. Therefore, when the K value is much larger than the K value during normal electricity consumption, it is considered that the electricity consumption on that day is abnormal. If abnormal K values occur for 15 consecutive days, the user is considered to have stolen electricity. The PQUI identification algorithm requires electricity consumption, voltage, current, and power for calculation, so it is not suitable for low-voltage users and can only be calculated for dedicated transformer users. Clustering only requires electricity consumption data, so it can be used by both low-voltage and dedicated transformer users. Therefore, when the user is a low-voltage user, the clustering algorithm is used to determine the time of electricity theft. When the user is a dedicated transformer user, additional conditions are required. The PQUI identification algorithm uses the K value fluctuation to determine whether electricity consumption is abnormal. Therefore, when the fluctuation is small, it cannot make an accurate judgment. The clustering algorithm directly judges based on the continuous time of the lowest class, avoiding the disadvantage of too small fluctuations, but due to the need for multiple iterations, a lot of time is required for calculation. In order to avoid the disadvantages of the two algorithms, the parameter D is added to calculate the fluctuation through distance, and then the appropriate algorithm is selected based on D. The specific calculation is as follows: Normalize the data first: Where W i (i=1,2,3,…,n) is a set of power data of the user, W min is the minimum value in the data, W max is the maximum value in the data; Then calculate the parameter D: D=|W i * -mean(W * )| / n……(5) Where W * W i * The sum of all data in, mean(W * ) is the average value of the normalized data, and n is the number of data; The threshold is set to 0.

2. When D≧0.2, the PQUI recognition algorithm is used; when D<0.2, the clustering algorithm is used.

3. The method for replenishing power based on clustering and PQUI recognition algorithm according to claim 2, characterized in that: The remote monitoring and diagnosis system for electric energy metering devices also includes a module for establishing a feature library of electric energy theft behavior. When establishing the feature library of electric energy theft behavior, the module includes the following steps: A1) Obtain historical electricity theft data; A2) Processing historical electricity theft data; obtaining electricity theft methods, assigning numbers to each theft method, and classifying them; the theft methods include: electricity theft without a meter, voltage circuit disconnection, poor voltage circuit contact, voltage circuit division, current circuit open circuit, current circuit short circuit, current circuit shunting, phase shifting, altering the internal structure of the meter, damaging the meter with high current or mechanical force, and external interference; A3) Establish a corresponding electricity theft behavior feature database for the three metering methods: high supply high metering, high supply low metering, and low supply low metering; among which: The data in the electricity theft behavior feature database under the high-power supply and high-metering metering method includes the category, the output theft method corresponding to the category, phase A voltage, phase C voltage, phase A current, phase C current, and active power. When determining the electricity theft behavior, the fault phase and the theft method are determined based on the phase voltage, phase current, and active power. The data in the electricity theft behavior feature database under the high-supply, low-metering metering method includes the category, the output electricity theft method corresponding to the category, phase A voltage, phase B voltage, phase C voltage, phase A current, phase B current, phase C current, and active power. When judging electricity theft behavior, the fault phase and the electricity theft method are determined based on the phase voltage, phase current, and active power. The data of the electricity theft behavior feature library under the low supply and low metering mode includes categories, output electricity theft methods and power consumption corresponding to the categories; when judging electricity theft behavior, the fault phase and electricity theft method are determined based on the power consumption.

4. The method for replenishing power based on clustering and PQUI recognition algorithm according to claim 3, characterized in that: The electricity theft methods classified as non-metered electricity theft include: disconnecting the voltage transformer Q1, bypassing the transformer and bridging Q2, and adding a bypass to bypass the electricity meter Q3; Theft methods classified as voltage circuit disconnection include: loosening the TV fuse Q4, breaking the fuse in the fuse tube Q5, loosening the voltage circuit terminal Q6, breaking the core of the voltage circuit wire Q7, and loosening the voltage connector Q8 of the energy meter; Theft methods classified as poor voltage circuit contact include: loosening the voltage connector Q9 of the energy meter, loosening the voltage circuit terminal Q10, and loosening the low-voltage fuse Q11 of the TV; The electricity theft methods classified as voltage circuit division include: inserting a resistor Q12 in series in the secondary circuit of the TV, disconnecting the neutral wire on the incoming line of a single-phase meter and inserting a voltage-reducing resistor Q13 in series between the outgoing line and the ground; Theft methods classified as open current loop include: loosening the TA secondary output terminal (Q14), artificially creating a poor contact fault in the TA secondary circuit terminal (Q15), breaking the core of the current loop conductor (Q16), and breaking the neutral wire (Q17). The electricity theft methods classified as current loop short-circuit electricity theft methods include: short-circuiting the current terminal Q18 of the energy meter, short-circuiting the terminal block Q19 in the current loop, and short-circuiting the primary or secondary side Q20 of the TA; The electricity theft methods classified as current loop shunting include: replacing TAs with different transformation ratios Q21, changing the secondary tap of tapped TAs Q22, and changing the number of primary turns of through-core TAs Q23; The electricity theft methods classified as phase-shifting theft include: swapping the phase and neutral wires of a single-phase meter, using the ground wire as the neutral wire (Q24); swapping the primary side input and output wires of a TA (Q25); swapping the same-name terminals of a TA (Q26); swapping the current terminals of a TA (Q27); swapping the phases of the TA-to-meter wiring (Q28); swapping the primary or secondary polarity of a TV (Q29); swapping the phases of the TV-to-meter wiring (Q30); using special inductors or capacitors for phase shifting (Q31); Theft methods classified as altering the internal structure of the meter include: reducing the number of turns of the current coil (Q32); spot welding the manganese copper resistor and cutting the manganese copper signal line (Q33); connecting resistors in parallel or series with the current sampling circuit (Q34); replacing the graded sampling resistors in the voltage sampling circuit (Q35); connecting the voltage coil with resistors and other electronic components for voltage division (Q36); short-circuiting with copper wire hooks (Q37); and implanting remote control for current diversion (Q38). Theft methods are classified as using large current or mechanical force to damage the meter. These methods include: burning the current coil Q39 with overload current, shocking the meter Q40 with the electric force of short-circuit current, and damaging the meter Q41 with mechanical external force. The electricity theft methods classified as external interference include: strong magnetic interference theft Q42, high-frequency interference theft Q43, high-voltage pulse theft Q44, and short-circuiting the inlet and outlet lines of the meter box Q45.

5. The method for replenishing power based on clustering and PQUI recognition algorithm according to claim 4, characterized in that: The characteristic database of electricity theft behavior under the high supply and high metering mode is: The characteristic database of electricity theft behavior under the high supply and low metering mode is: The characteristic database of electricity theft behavior under the low supply and low metering mode is:

6. The method for replenishing power based on clustering and PQUI recognition algorithm according to claim 5, characterized in that: The judgment of electricity theft includes the identification of electricity theft under high supply and low metering mode, the identification of electricity theft under high supply and high metering mode, and the identification of electricity theft under low supply and low metering mode; The identification of electricity theft under the high-supply and low-metering method includes the following steps: 3101) Inputting electricity usage data of a user suspected of electricity theft; 3102) Determine whether the current and voltage of phases A, B, and C have values; if not, consider it impossible to determine and end; if so, proceed to the next step; 3103) Determine whether the three-phase voltage and current are close to 0; if so, compare the electricity theft behavior feature library under the high-supply, low-metering metering mode, output 9 types of electricity theft reasons, and end; if not, proceed to the next step; 3104) Determine whether one or more phases of the three-phase current are close to zero; if so, compare the power theft behavior feature library under the high-supply, low-metering metering mode, output Class 1 power theft cause and the power theft phase with output current close to zero, and end; if not, proceed to the next step; 3105) Determine whether one or more phases of the three-phase voltage are close to zero; if so, compare the voltage with the electricity theft behavior feature library under the high-supply, low-metering metering mode, output two types of electricity theft reasons and the phases with the electricity theft voltage close to zero, and end; if not, proceed to the next step; 3106) Call for phase power factor measurement; 3107) Determine whether all three-phase power factors are normal; if not, proceed to step 2110; if so, proceed to the next step; 3108) Determine whether all three-phase voltages are greater than or close to 220V; if not, compare the three-phase voltages with the electricity theft behavior feature library under the high-supply, low-metering metering mode, output the four types of electricity theft causes and the phase with the theft voltage that is too low, and end; if yes, proceed to the next step; 3109) Determine whether the three-phase currents are equal. If not, compare the three-phase currents with the electricity theft behavior feature library under the high-supply, low-metering metering mode, output the three types of electricity theft causes and the phase with the smaller output current, and end; if so, output the three types of electricity theft causes and the phase with the smaller output current, and end; 3110) Determine whether the user's daily electricity consumption is 0; if so, compare it with the electricity theft behavior feature library under the high-supply and low-metering metering method, output 6 or 8 types of electricity theft reasons and the electricity theft phase with abnormal power factor, and end; if not, proceed to the next step; 3111) Determine whether one or more phases of the three-phase current are less than 0; if so, compare the three-phase current with the electricity theft behavior feature library under the high-supply, low-metering mode, output five types of electricity theft causes and the phases with negative current, and terminate; if not, output seven types of electricity theft causes and the phases with negative voltage, and terminate; Identifying electricity theft under high-supply and high-metering mode includes the following steps: 3201) Input the electricity usage data of the suspected electricity thief 3202) Determine whether the current and voltage of phases A and C have values. If not, consider that the determination cannot be made and terminate. 3203) Determine whether the voltage and current of phases A and C are both close to 0; if so, compare the power theft behavior feature library under the high-supply high-metering metering mode, output 9 types of power theft causes, and end; If not, proceed to the next step; 3204) Determine whether one or more phases of the voltage of phases A and C are close to zero; if so, compare the voltage with the electricity theft behavior feature library under the high-supply high-metering metering mode, output Class 1 electricity theft cause and the electricity theft phase with voltage close to zero, and end; If not, proceed to the next step; 3205) Determine whether one or more phases of the A and C currents are close to zero; if so, compare the power theft behavior feature library under the high-supply high-metering metering mode and output two types of power theft causes and the power theft phases with power close to zero; if not, proceed to the next step; 3206) Call for phase power factor measurement; 3207) Determine whether the power factors of phases A and C are all normal; if not, proceed to step 2210; if so, proceed to the next step; 3208) Determine whether the voltages of phases A and C are both greater than or close to 220V; if not, compare the power theft behavior feature library under the high-supply high-metering metering mode, output the three types of power theft causes and the power theft phase with too low voltage, and end; if yes, proceed to the next step; 3209) Determine whether the currents of phases A and C are equal; if so, compare the electricity theft behavior feature library under the high-supply, high-meter metering mode, output the four types of electricity theft causes and the electricity theft phases A and C, and end; if not, compare the electricity theft behavior feature library under the high-supply, high-meter metering mode, output the four types of electricity theft causes and the electricity theft phase with smaller current, and end; 3210) Determine whether the voltage of phases A and C is less than 0; if so, compare the voltage with the electricity theft behavior feature library under the high-supply high-metering mode, output the 7 types of electricity theft reasons and the electricity theft phases with voltage less than 0, and end; If not, proceed to the next step; 3211) Determine whether the current of phases A and C is less than 0; if so, compare the power theft behavior feature library under the high-supply high-metering metering mode, output the five types of power theft causes and the power theft phases with current less than 0, and end; If not, proceed to the next step; 3212) Determine whether the user's daily electricity consumption is 0; if so, compare the electricity theft behavior feature library under the high-supply and high-metering metering mode, output 6 or 8 types of electricity theft causes and the electricity theft phases AC or AB, and end; if not, compare the electricity theft behavior feature library under the high-supply and high-metering metering mode, output 8 types of electricity theft causes and the electricity theft phases BC, and end; The identification of electricity theft under the low-supply and low-metering method includes the following steps: 3301) Input the electricity usage data of the suspected electricity thief 3302) Determine whether the user's daily electricity consumption has a value; if not, consider it impossible to determine and end; If yes, go to the next step; 3303) Determine whether the user's daily electricity consumption is close to 0; if so, compare the electricity theft behavior feature library under the low supply and low metering mode, output a Class 1 electricity theft reason, and end; If not, proceed to the next step; 3304) Determine whether the user's daily electricity consumption is negative; if so, compare the electricity theft behavior feature library under the low supply and low metering method, output two types of electricity theft causes, and end; if not, compare the electricity theft behavior feature library under the low supply and low metering method, output three types of electricity theft causes, and end.

7. The method for replenishing power based on clustering and PQUI recognition algorithm according to claim 5, characterized in that: The database server stores electricity theft behaviors and correction coefficients for each electricity theft behavior. The correction coefficients include the correction coefficients for high-voltage power supply and high-metering users of dedicated transformers, the correction coefficients for high-voltage power supply and low-metering users of dedicated transformers, and the correction coefficients for low-voltage users. The correction coefficients for high-power and high-meter users of dedicated transformers are shown in the following table: The correction coefficients for high-supply and low-cost users of dedicated transformers are shown in the following table: The correction factors for low-voltage users are shown in the following table:

Images