Connection phase determination device, connection phase determination system

Abstract

The present invention provides a connection phase determination device that can accurately determine the connection phase of a transformer connected to a power distribution system. [Solution] A connection phase determination device applied to a connection phase determination system comprising a plurality of transformers connected to each phase of a power distribution system, sensors for measuring the current of each phase, and a plurality of energy meters for measuring the amount of electricity supplied from the plurality of transformers to consumers, the device determines the connection phase of the plurality of transformers connected to the power distribution system, the device identifies a predetermined number of connection states that minimize an evaluation value corresponding to the difference between a first current of each phase, obtained based on the measured values ​​measured by the plurality of energy meters and assuming a connection state for each of the plurality of transformers, and a second current of each phase obtained from the measurement results of the sensors, using an optimization method with a predetermined number of different conditions, and determines the connection phase of each of the plurality of transformers connected to the power distribution system based on the identified predetermined number of connection states.

Description

Technical Field

[0001] The present invention relates to a connection phase discrimination device and a connection phase discrimination system.

Background Art

[0002] Techniques for discriminating the connection phase of a transformer connected to a power distribution system are known (see, for example, Patent Document 1).

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] By the way, when discriminating the connection phase, generally, the measurement results of sensors that measure the current flowing through each phase of the power distribution system and the measurement results of smart meters that measure the power supplied to consumers are used. However, since the measurement periods of the sensors and the smart meters are different, it is difficult to discriminate the connection phase with high accuracy.

[0005] The present invention has been made in view of the above conventional problems, and an object thereof is to provide a connection phase discrimination device and a connection phase discrimination system that can accurately discriminate the connection phase of a transformer connected to a power distribution system.

Means for Solving the Problems

[0006] The first aspect of the present invention, which is the main aspect of solving the aforementioned problems, is applied to a connection phase determination system comprising a plurality of transformers connected to each phase of a power distribution system, sensors for measuring the current of each phase, and a plurality of energy meters for measuring the amount of electricity from the plurality of transformers to consumers, and is a connection phase determination device for determining the connection phase of the plurality of transformers connected to the power distribution system, which determines the connection phase of each of the plurality of transformers connected to the power distribution system, by using an optimization method with a predetermined number of different conditions to identify a predetermined number of connection states that minimize an evaluation value corresponding to the difference between a first current of each phase when assuming the connection state of each of the plurality of transformers, obtained based on the measured values ​​measured by the plurality of energy meters, and a second current of each phase obtained from the measurement results of the sensors, and determines the connection phase of each of the plurality of transformers connected to the power distribution system based on the predetermined number of identified connection states.

[0007] A connection phase determination system comprising: a plurality of transformers connected to each phase of a power distribution system; sensors for measuring the current of each phase; a plurality of energy meters for measuring the amount of electricity from the plurality of transformers to a consumer; and a connection phase determination device for determining the connection phase of the plurality of transformers connected to the power distribution system, wherein the connection phase determination device identifies a predetermined number of connection states that minimize an evaluation value corresponding to the difference between the first current of each phase, obtained based on the measured values ​​measured by the plurality of energy meters and assuming the connection state of each of the plurality of transformers, and the second current of each phase obtained from the measurement results of the sensors, using an optimization method with a predetermined number of different conditions, and determines the connection phase of each of the plurality of transformers connected to the power distribution system based on the identified predetermined number of connection states. [Effects of the Invention]

[0008] According to the present invention, it is possible to provide a connection phase discrimination device and a connection phase discrimination system that can accurately determine the connection phase of a transformer connected to a power distribution system. [Brief explanation of the drawing]

[0009] [Figure 1] This figure shows an example of a connection phase discrimination system 1. [Figure 2] This figure shows an example of measurement information stored in the sensor measurement value DB11. [Figure 3] This figure shows an example of the information stored in the power distribution system configuration DB12. [Figure 4] This figure shows an example of the information stored in the smart meter DB13. [Figure 5] This diagram illustrates the currents in the power distribution system 20 and the pole-mounted transformers TR1 to TR3. [Figure 6] This flowchart shows an example of the process performed by the connection phase determination device 10. [Figure 7] This is a flowchart showing an example of step S70. [Figure 8] This is a diagram to explain the concept of ensemble learning. [Figure 9] This diagram illustrates the application of the ensemble learning concept to optimization. [Figure 10] This diagram illustrates the measured values ​​from the switch and the average current from the smart meter. [Figure 11] This is a diagram used to explain probability matrices. [Figure 12] This diagram illustrates the evaluation values ​​in connection state A. [Figure 13] This diagram illustrates the evaluation values ​​in connection state B. [Figure 14] This is a diagram illustrating the updating of the probability matrix. [Figure 15] This diagram illustrates the result of the majority vote on the connection phase. [Figure 16] This figure shows an example of simulation conditions and results. [Modes for carrying out the invention]

[0010] This specification and the accompanying drawings make at least the following matters clear. Furthermore, identical or equivalent components, members, etc., shown in each drawing are denoted by the same reference numerals, and redundant explanations are omitted where appropriate.

[0011] =====Connection Phase Discrimination System 1===== FIG. 1 is a diagram showing an example of a connection phase discrimination system 1 which is an embodiment of the present embodiment. The connection phase discrimination system 1 is a system for discriminating the connection phases of pole-mounted transformers (described later) connected to the three-phase distribution lines 20u, 20v, 20w (20u to 20w) of the power distribution system 20. The connection phase discrimination system 1 includes a switch SE1, pole-mounted transformers TR1 to TR3, smart meters SM1 to SM4, and a connection phase discrimination device 10.

[0012] The switch SE1 is a switch with sensors including sensors for measuring the current, voltage, power factor, etc. of each of the three-phase distribution lines 20u to 20w, for example, every minute. Further, the switch SE1 transmits the measured values measured by the sensors to a connection phase discrimination device 10 described later via a communication line (not shown). Hereinafter, each of the distribution lines 20u to 20w may be simply referred to as "each phase".

[0013] The pole-mounted transformers TR1 to TR3 are distribution transformers located on the downstream side (consumer side) of the switch SE1 and connected to each phase. The primary sides of the pole-mounted transformers TR1, TR2, and TR3 are connected to each phase of the power distribution system 20. Here, although the pole-mounted transformers TR1 and TR2 are illustrated as single-phase transformers and the pole-mounted transformer TR3 is illustrated as a three-phase transformer, the number of pole-mounted transformers and the individual phase types are not limited to the illustrated examples at all. Also, in the actual power distribution system 20, a large number of pole-mounted transformers (not shown) other than the pole-mounted transformers TR1 to TR3 are connected, but are omitted for the sake of convenience.

[0014] The smart meters SM1 to SM4 are watt-hour meters with communication functions connected to the secondary sides of the pole-mounted transformers TR1, TR2, and TR3, and measure the power consumption, supply voltage, and current of consumers (not shown). The smart meters SM1 to SM4 measure the above-described power consumption, etc. every 30 minutes, for example, and transmit the measured values to the connection phase discrimination device 10. In the present embodiment, the smart meter SM1 is connected to the pole-mounted transformer TR1, the smart meters SM2 and SM3 are connected to the pole-mounted transformer TR2, and the smart meter SM4 is connected to the pole-mounted transformer TR3.

[0015] Also, as will be described later in detail, each of the smart meters SM1 to SM4 transmits various information such as the model type and rated voltage of the smart meters SM1 to SM4 to the connection phase discrimination device 10 in addition to the measurement values measured by the smart meters SM1 to SM4. Further, the pole-mounted transformers TR1 to TR3 correspond to the "plurality of transformers", and the smart meters SM1 to SM4 correspond to the "plurality of watt-hour meters".

[0016] ==Connection Phase Discrimination Device 10== The connection phase discrimination device 10 is a device that discriminates the connection phase of the pole-mounted transformers connected to each phase based on the current measured by the switch SE1, the measurement values from the smart meters SM1 to SM4, and various information. The connection phase discrimination device 10 includes a sensor measurement value database (hereinafter, appropriately referred to as "DB") 11, a distribution system configuration DB12, a smart meter measurement value DB13, a transformer current calculation unit 14, a data selection unit 15, a voltage and current analysis unit 16, a connection phase discrimination unit 17, and a result output unit 18. Note that the functional blocks such as the transformer current calculation unit 14 are realized in the connection phase discrimination device 10 when a CPU (not shown) executes a predetermined program.

[0017] As shown in FIG. 2, the sensor measurement value DB11 stores information including the measurement date and time of the measurement value of the switch SE1, the line-to-line voltages, the currents of each phase, the power factor, etc. as the measurement value of the switch SE1.

[0018] As shown in FIG. 3, the distribution system configuration DB12 stores the pole number, the connection destination pole number, the impedance (R, X) of the distribution line, the attached equipment of each pole, and the identification information (ID) of the equipment. Hereinafter, the information stored in the distribution system configuration DB12 is referred to as distribution system configuration information I1.

[0019] As shown in Figure 4, the smart meter DB13 stores smart meter master data such as the identification information (ID) of smart meters SM1 to SM4, information on the connected pole-mounted transformer, model (including single-phase / three-phase identification information), and rated voltage, as well as the smart meter's measured values. The smart meter's measured values ​​are those measured every 30 minutes by smart meters SM1 to SM4.

[0020] ==Transformer Current Calculation Unit 14== The transformer current calculation unit 14 in Figure 1 calculates the average current flowing into the pole-mounted transformers TR1 to TR3 over a predetermined time (e.g., 30 minutes) from the power consumption of the consumer at a predetermined time cross-section, based on the measured values ​​of a smart meter, etc.

[0021] For example, as shown in Figure 5, suppose that in a smart meter SM1 connected to a pole-mounted transformer TR1, the power consumption over 30 minutes is 750Wh, and the average voltage over 30 minutes is 102V. In such a case, the average current flowing from the distribution system 20 to the pole-mounted transformer TR1 over 30 minutes can be calculated by converting the average current flowing into the smart meter SM1 over 30 minutes to the high-voltage side.

[0022] Specifically, the average current flowing into the smart meter SM1 over a 30-minute period can be calculated using the following formula (1). Power consumption over 30 minutes / (0.5 × average voltage over 30 minutes) ... (1) Note that 0.5 in equation (1) is a coefficient used to convert the power consumption per hour to a 30-minute equivalent.

[0023] As a result, the average current flowing into the smart meter SM1 over 30 minutes is 750[Wh] / 0.5×10²[V]=14.7[A]. In this embodiment, the 30-minute average voltage is calculated based on the voltage measured by the smart meter SM1, but for example, a value estimated based on the measured voltage of the switch SE1 and the system impedance (for example, a predetermined impedance) may also be used. Also, here, the 30-minute average power factor is assumed to be 1.0.

[0024] Furthermore, the average current flowing into the pole-mounted transformer TR1 can be calculated by multiplying the average current flowing into the smart meter SM1 by the transformation ratio. Here, the transformation ratio is 105 / 6600, so the average current flowing into the pole-mounted transformer TR1 over 30 minutes is 14.7[A] × (105 / 6600) = 0.23[A].

[0025] Similarly, calculating for pole-mounted transformer TR2, the 30-minute average current flowing into smart meters SM2 and SM3 is 15.8[A] and 11.9[A], respectively. Converting the total value of 27.7 (=15.8 + 11.9)[A] to the high-voltage side gives 0.44 (=0.25 + 0.19)[A].

[0026] Since pole-mounted transformer TR3 is a three-phase transformer, the 30-minute average current flowing into smart meter SM4 is 1143[Wh] / (0.5×√3×201[V])=6.6[A]. Converting this to the high-voltage side, the current flowing into pole-mounted transformer TR3 is 6.6[A] × transformation ratio = 6.6[A] × (210 / 6600)=0.21[A].

[0027] Thus, based on information from the smart meter, the transformer current calculation unit 14 calculates the average current (0.23A, 0.44A, 0.21A) for a predetermined time cross-section (for example, 30 minutes) flowing into the pole-mounted transformers TR1 to TR3, as shown in Figure 5.

[0028] The 30-minute average voltage can be any of the following: the average voltage measured by a smart meter, the average voltage estimated based on the measured voltage of a sensor-equipped switch and wiring impedance, or the average voltage estimated by power flow calculation.

[0029] == Data Selection Section 15 == The data selection unit 15 in Figure 1 selects the current and power factor of each phase for a predetermined period (for example, one year) from the data stored in the sensor measurement value DB 11, and outputs them to the voltage-current analysis unit 16 as sensor measurement data I0.

[0030] ==Voltage and Current Analysis Unit 16== The voltage-current analysis unit 16 calculates the average active current at a predetermined time cross-section (for example, 30 minutes) in the sensor. Specifically, the voltage-current analysis unit 16 uses the sensor measurement data I0 to calculate the 30-minute average active currents Iure, Ivre, and Iwre for each phase of the power distribution system 20, as shown in Figure 5.

[0031] Here, the average active currents Iure, Ivre, and Iwre are obtained using equation (2), which multiplies the 30-minute average current measured by the sensor-equipped switch SE1 by the 30-minute average power factor (which is assumed to be 1.0 for convenience). Average current / active current [A] = Average current [A] × Average power factor ... (2) As a result, the average active currents Iure, Ivre, and Iwre are calculated to be 0.763A, 0.420A, and 0.620A, respectively.

[0032] ==Connection Phase Determination Unit 17== The connection phase determination unit 17 in Figure 1 determines the connection phase of pole-mounted transformers TR1 to TR4 based on the average active currents Iure, Ivre, Iwre, distribution system configuration information I1, smart meter information I2, and the 30-minute average current calculated by the transformer current calculation unit 14. Details of the processing performed by the connection phase determination unit 17 will be described later.

[0033] ==Result Output Section 18== The result output unit 18 displays the result of the connection phase determination by the connection phase determination unit 17 on the user interface (e.g., a display) and stores it in an external storage device (not shown).

[0034] <<<Processing at the connection phase determination device 10>>> Referring to Figure 6, an example of the processing performed by the connection phase determination device 10 will be explained. First, the data selection unit 15 selects and acquires data for the time cross-section tk~tn that will be used for connection phase determination from all the data in the sensor measurement value database 11 (S10). By setting the time cross-section tk~tn, for example, sensor measurement data I0 from the present to the past year is acquired from all the data.

[0035] Based on the sensor measurement data I0, the voltage and current analysis unit 16 calculates the 30-minute average active currents Iure, Ivre, and Iwre from the active currents of each phase of the power distribution system 20 measured by the sensor-equipped switch SE1 (S20).

[0036] The transformer current calculation unit 14 obtains smart meter measurement values, including the 30-minute power consumption of each customer, from the smart meter information DB 13 (S30).

[0037] Furthermore, the transformer current calculation unit 14 calculates the 30-minute average current for each pole-mounted transformer, as illustrated in the table at the bottom of Figure 5, based on the smart meter measurement values ​​(S40). Although details are omitted here, the transformer current calculation unit 14 in this embodiment calculates the average current over a period of time cross-section tk to tn (for example, one year) based on the information in the smart meter information DB 13.

[0038] The connection phase determination unit 17 acquires the average active current obtained in step S20, the system configuration information I1 from the distribution system configuration DB12, the average current for each pole-mounted transformer obtained in step S40, and the smart meter information I2 (S50). The connection phase determination unit 17 also identifies the single-phase / three-phase type of the pole-mounted transformer to be connected to from the smart meter information I2 shown in Figure 4 (S60).

[0039] As will be described in detail later, when step S60 is executed, the connection phase determination unit 17 determines the connection phase (S70), and the result output unit 18 outputs the determination result of the connection phase (S80).

[0040] <<<Ensemble Learning and Optimization Methods>>> Here, ensemble learning applied to this embodiment will be explained with reference to Figures 8 and 9. Figure 8 is a diagram illustrating the concept of ensemble learning. For example, one learner (here, a trained model using supervised data) can be used to predict the solutions for 10 data points X1 to X10. If the accuracy of one learner is 70%, then, for example, data points X1 to X7 will be correct, and data points X8 to X10 will be incorrect. Note that the number of data points X1 to X10 here is an example and is not limited to 10. Also, data point X1 can be, for example, the connection phase of a pole-mounted transformer.

[0041] On the other hand, if we use multiple learners (three in this case) with a lower prediction accuracy of 60% than described above, each of the weak learners No. 1 to No. 3 will only obtain 60% of the correct answers. However, if we consider the majority vote of the results of each of the three weak learners No. 1 to No. 3, eight of the ten data points X1 to X10 will be accurate. In other words, calculating the result of the majority vote of the results of each of the weak learners No. 1 to No. 3 improves accuracy compared to using a single learner with 70% accuracy.

[0042] In this embodiment, the concept of ensemble learning is applied to the optimization method. Specifically, as shown in Figure 9, instead of using a single optimization method to find the solution for the unknown data X1 to X10, multiple (for example, three) weak optimization methods with lower accuracy than the single optimization method are used, and then the result of a majority vote is calculated to find the solution for the unknown data X1 to X10. This improves the prediction accuracy, similar to the case in Figure 8. As will be described in detail later, in this embodiment, the concept of ensemble learning shown in Figure 9 is applied to the optimization method to determine the connected phase of the pole-mounted transformer.

[0043] The connection phase determination unit 17 of this embodiment uses the state of the connection phase of the pole-mounted transformer (hereinafter referred to as "connection state θ") as a determination variable and optimizes the evaluation value of the evaluation function, which is determined by the difference between the average active currents Iure, Ivre, and Iwre and the sum of the average currents for each pole-mounted transformer, to minimize the evaluation value.

[0044] Hereafter, the number of evaluation functions to be optimized (hereinafter also referred to as the number of weak optimization groups) will be denoted as "g". For example, if the number of weak optimization groups g is 3, then three different evaluation functions will be used.

[0045] Furthermore, in this embodiment, Population-Based Incremental Learning (PBIL) is used as the optimization method. PBIL is an optimization method that enhances the efficiency of genetic algorithms, and generally, three parameters can be set: the data used, the number of iterations (iter), and the number of individuals (m). As will be described in detail later, "data used" is the data used to calculate the evaluation function, "number of iterations (iter)" is the number of times the probability matrix in PBIL is updated, and "number of individuals (m)" is a candidate solution (i.e., the connection state θ of the pole-mounted transformer) generated based on the probability matrix.

[0046] <<<Details of Step S70>>> Here, with reference to Figure 7, an example of steps S71a to S79a included in step S70 in which the connection phase determination unit 17 is executed will be explained. First, the connection phase determination unit 17 obtains the number g of weak optimization groups, the data used, the number of iterations iter, and the number of individuals m, which are stored in advance in the memory unit (not shown) (S71a). Note that the four parameters in step S71a are assumed to be set in advance by the user, for example.

[0047] The connection phase discrimination unit 17 generates g weak optimization groups based on the number g of weak optimization groups (S72a). Each of the g weak optimization groups generated in step S72a is based on the PBIL optimization method, and the data used, the number of iterations (iter), and the number of individuals (m) are set accordingly.

[0048] The connection phase discrimination unit 17 sets the data to be input to the g weak optimization groups by random sampling in order to change the characteristics of the solution search (modification of the probability matrix) performed by the g weak optimization groups (S73a). Figure 10 shows the 30-minute average active currents Iure, Ivre, and Iwre of the switch SE1 over a period of time cross-section tk to tn (for example, one year), and the 30-minute average current for each smart meter obtained from smart meter measurements.

[0049] The connection phase determination unit 17 generates g data groups randomly, for example, using a random sampling method. In this embodiment, the number of data points in each g data group is assumed to be a predetermined proportion of the total number of data points (for example, 4 / 5 or 1 / 12). Therefore, if g is 12 and the predetermined proportion is 1 / 12, then data for each month will be assigned from the total data for one year (12 months). The number of data points in each of the g data groups may be different. Furthermore, while the connection phase determination unit 17 uses a random sampling method to randomly extract a predetermined proportion of data from the total data, it is not limited to this.

[0050] The connection phase discrimination unit 17 generates a population of connection state θ that indicates the state of the connection phase of the pole-mounted transformer based on a probability matrix (S74a). Here, the population of connection state θ includes m states (i.e., connection phase patterns). On the left side of Figure 11, an example of a probability matrix corresponding to each of the g weak optimization groups is shown. In step S74a, the g corresponding probability matrices of the weak optimization groups are applied, and m connection state θs are generated based on these probability matrices.

[0051] For example, in the evaluation function for the first weak optimization (hereinafter referred to as "Weak Optimization No. 1"), among the m individuals, individual 1 is assumed to have the connection phases of pole-mounted transformers TR1 and TR2 as VW phase and UV phase, respectively. Hereafter, the connection phases of individual 1 may be written as (TR1,TR2)=(VW phase,UV phase).

[0052] Furthermore, for convenience, only the connection phases of pole-mounted transformers TR1 and TR2 have been explained here. In reality, connection phases are also set for pole-mounted transformers other than TR1 and TR2 (not shown) installed in the distribution system 20.

[0053] Here, we have explained the probability matrix for weak optimization No. 1 and the connection states θ for m individuals. The same applies to the probability matrices for weak optimizations No. 2 to No. g and the connection states θ for m individuals, so a detailed explanation will be omitted.

[0054] The connection phase discrimination unit 17 calculates the evaluation value f(θ)kt of individual k in the assumed connection state θ at the time cross-section t (S75a). Note that the time cross-section t refers to one time, for example, "2024 / 4 / 1 / 9:00", among multiple target times in the evaluation function for performing weak optimization No. 1 in Figure 10. Here, the state of individual 1 in weak optimization No. 1 in Figure 11 (hereinafter referred to as connection state A) will be explained with reference to Figure 12.

[0055] The connection phase determination unit 17 calculates the total current value of each phase from the average active current of each phase based on the following formula in connection state A. Iu = Iuv - Iwu(-1 / 2 + j(√3 / 2)) Iv = Ivw(-1 / 2 - j(√3 / 2)) - Iuv Iu=Iwu(-1 / 2+j(√3 / 2))-Ivw(-1 / 2-j(√3 / 2))

[0056] Here, if the current Ivw = 0.23A for the single-phase pole-mounted transformer TR1 and the current Iuv = 0.44A for the single-phase pole-mounted transformer TR2, then the currents for the U, V, and W phases from the single-phase pole-mounted transformers TR1 and TR2 will be 0.44A, 0.59A, and 0.23A, respectively, based on the above formula. Furthermore, if the current to each phase of the three-phase pole-mounted transformer TR3 is 0.21A, then the currents for the U, V, and W phases will be 0.65A (=0.44 + 0.21), 0.80A (=0.59 + 0.21), and 0.44A (=0.23 + 0.21).

[0057] The connection phase determination unit 17 then calculates the evaluation value f(θ)kt in connection state A based on the following equation (3). f(θ)kt= |Iure-Iu|+||Ivre-Iv|+||Iwre-Iw|···(3) Here, the average active currents Iu, Iv, and Iw for each phase correspond to the "first current obtained based on the measured values ​​measured by the energy meter," while the average active currents Iure, Ivre, and Iwre for each phase correspond to the "second current for each phase obtained from the measurement results of the sensor."

[0058] In this embodiment, the evaluation value f(θ)kt of weak optimization No. 1 in connection state A (i.e., state of individual 1) at a certain time cross-section t (for example, 2024 / 4 / 1 / 9:00) is 0.673, as shown in Figure 12.

[0059] In step S75a of Figure 7, the connection phase determination unit 17 calculates an evaluation value f(θ)kt of a certain time cross-section t based on equation (3), and updates the time cross-section to the next randomly sampled time cross-section (step S75b). Then, the connection phase determination unit 17 repeatedly calculates equation (3) in the updated time cross-section t (step S75a).

[0060] Specifically, the connection phase discrimination unit 17 calculates the evaluation value f(θ)kt of weak optimization No. 1 for connection state A (i.e., the state of individual 1) at the time cross-section t of 2024 / 4 / 1 / 9:30, following 2024 / 4 / 1 / 9:00. In this way, the connection phase discrimination unit 17 of this embodiment repeatedly calculates the evaluation value f(θ)kt for the time cross-section t of all data included in the data group that is the target of the weak optimization evaluation function (steps S75a and S75b in Figure 7).

[0061] Then, steps S75a and S75b in Figure 7 are repeated until the evaluation value f(θ)kt is calculated for all time sections tk~tn of the data included in the data set, and the sum of all multiple time sections of individual k is calculated based on equation (4) (step S76a). f(θ)k=Σf(θ)kt ···(4) Note that, for convenience, the variable t is omitted in the Σ symbol here, but the variable time section t changes by 1 from tk to tn. Also, the evaluation value f(θ)k on the left side of equation (4) corresponds to "an evaluation value obtained by summing the differences between the first current and the second current for multiple time sections."

[0062] When step S76a is executed, the connection phase determination unit 17 increments the number of k by one and updates individual k (step S76b). As a result, for example, the individual in weak optimization No. 1 in Figure 11 changes from individual 1 to individual 2. Consequently, the connection state θ becomes the state of individual 2 in weak optimization No. 1 in Figure 11 (hereinafter referred to as connection state B).

[0063] Figure 13 shows an example of the calculated evaluation value f(θ)kt for weak optimization No. 1 of connection state B (i.e., the state of individual 2). The calculation for Figure 13 is the same as that for Figure 12 described above, so a detailed explanation is omitted here.

[0064] When the connection state changes to B, the connection phase determination unit 17 repeats steps 75a and 75b for all time sections tk to tn, and then calculates equation (4) (step S76a). As a result, the number of k in individual k is incremented by one and updated (step S76b). When this process is repeated, the evaluation value f(θ)k of equation (4) is calculated for each of individuals from 1 to m.

[0065] Subsequently, the connection phase discrimination unit 17 updates the probability matrix (step S77a). First, as shown in Figure 14, the connection phase discrimination unit 17 selects, for example, N individuals with small evaluation values ​​f(θ)k in ascending order. In the example in Figure 14, for example, individual i has the smallest evaluation value at 0.01, and individual j has the next smallest evaluation value at 0.015. Here, N is set to 1, so the components of the probability matrix, which will be described later, are updated to approach the state of individual i.

[0066] Here, since the connection state of individual i with the smallest evaluation value is the most likely, the connection phase discrimination unit 17 increases the probability that individual i is in that connection state. Specifically, in Figure 14, for example, since there is a high probability that the connection phase of pole-mounted transformer TR1 is the UV phase, the probability that the connection phase of pole-mounted transformer TR1 is the UV phase is updated based on the following formula. Probability after update = Probability before update + Learning rate ε ···(5) Here, the learning rate ε is a value set or calculated appropriately based on the PBIL algorithm. In this embodiment, since the learning rate ε is set to 1 / 10, the probability after updating becomes 13 / 30 (= 1 / 3 + 1 / 10).

[0067] In this state, for example, the probabilities that the connected phases of pole-mounted transformer TR1 are UV phase, VW phase, and WU phase are 13 / 30, 1 / 3, and 1 / 3, respectively, so the sum of the three probabilities is not 1. Therefore, the connection phase discrimination unit 17 adjusts the probability matrix so that the sum of the probabilities that the connected phases of pole-mounted transformer TR1 are UV phase, VW phase, and WU phase is 1. As a result, for example, the probabilities that the connected phases of pole-mounted transformer TR1 are UV phase, VW phase, and WU phase are 130 / 330, 100 / 330, and 100 / 330, respectively. Note that although pole-mounted transformer TR1 was used as an example here, similar calculations are performed and the probability matrix is ​​adjusted for other pole-mounted transformers TR2 as well.

[0068] In step S77a, when the probability matrix is ​​updated, the connection phase determination unit 17 updates the iteration count (step S77b), and then repeats steps S74a to S77a described above. As a result, in this embodiment, the probability matrix is ​​updated by the number of iterations (iter), and a connection state closer to the correct answer is searched for for the number of iterations (iter).

[0069] When steps S74a to S77b are repeated for the number of iter iterations, the connection phase determination unit 17 selects the connection state θ with the smallest value among the evaluation values ​​f(θ)k obtained through the repetitions as θming (step S78a).

[0070] When step S78a is executed, the connection phase determination unit 17 selects the evaluation function of the next target weak optimization from the group of g weak optimizations (step S78b). For example, if the connection state θming of weak optimization No. 1 is obtained from the group of g weak optimizations, the connection phase determination unit 17 selects weak optimization No. 2, which follows weak optimization No. 1, and executes steps S74a to S78a.

[0071] By repeating these processes, the connection state θming for each of the weak optimizations No. 1 to No. g among the g weak optimizations is identified. For convenience, the upper part of Figure 15 shows only the connection states θming for each of the weak optimizations No. 1 to No. 3. Here, the connection state θming for weak optimization No. 1 is (TR1,TR2)=(UV,WU). Note that g corresponds to a "predetermined number".

[0072] Furthermore, as shown in the lower part of Figure 15 and the bottom of Figure 8, the connection phase determination unit 17 determines the most likely connection state θmin from among the g connection states θming of weak optimization No. 1 to No. g by majority vote (step S79a). For example, in the upper part of Figure 15, the pole-mounted transformer TR1 is estimated to be in the UV phase in weak optimization No. 1, in the UV phase in weak optimization No. 2, and in the VW phase in weak optimization No. 3. In this case, there are two UV phases and one VW phase, so as a result of the majority vote, the connection phase determination unit 17 determines that the connection phase of the pole-mounted transformer TR1 is the UV phase.

[0073] Note that in Figure 15, for convenience, only weak optimizations No. 1 to No. 3 are shown out of the three weak optimizations No. 1 to No. g. However, among the UV, VW, and WU phases included in the g number of connected phases, the connected phase with the largest number is identified as the connected phase of pole-mounted transformer TR1 (in this case, the UV phase).

[0074] Then, when step S79a is executed, as shown in Figure 6, the result output unit 18 outputs the connection phase determination result (in this case, the connection phases of all pole-mounted transformers of the distribution system 20 obtained in step S79a) (step S80).

[0075] <<<Simulation conditions and results>>> Figure 16 shows an example of simulation conditions and results using the connection phase determination device 10 of this embodiment described above. The upper part of Figure 16 shows the simulation conditions, and the lower part shows the simulation results.

[0076] In Figure 16, it is assumed that 32 pole-mounted transformers are connected to the distribution system. Furthermore, the loads connected to the distribution system include a solar power system and actual consumers. Note that Figure 16 uses data obtained from an actual power system corresponding to power system 20 at a specified date and time.

[0077] Furthermore, in this embodiment, PBIL is used as the optimization method, with 60 individuals m, 200 iterations iter, a learning rate ε of 0.09, and 2 individuals N used for updating the random variable. Note that in Figure 14, the number of individuals N is 1, and only the individual with the smallest evaluation value is used, but it may be 2 or more. Also, here, the number of evaluation functions g used for weak optimization in the concept of ensemble learning is set to 60, and 80% of the data is randomly selected from the entire dataset as random sampling. Therefore, each of the weak optimizations No. 1 to No. g will use data randomly selected from the entire dataset (80%).

[0078] In this embodiment, simulations were performed using three different conditions. In simulation condition 1, only the "optimization method" was used from the "optimization method," "ensemble learning," and "random sampling" conditions in the upper part of Figure 16. Specifically, in simulation condition 1, PBIL was adopted as the optimization method, with the number of individuals m being 60, the number of iterations iter being 200, the learning rate ε being 0.09, and the number of individuals N used for updating the random variable being 2.

[0079] Therefore, under simulation condition 1, in step S70, only one weakly optimized evaluation function is generated in step S72a, and all data is used in step S73a. Furthermore, under simulation condition 1, the loop processing in step 78b is not included. When the simulation was run 100 times using this general method, the number of times all connection phases of the 32 pole-mounted transformers were correctly identified was 0.

[0080] In simulation condition 2, the "optimization method" and "ensemble learning" were used from the "optimization method," "ensemble learning," and "random sampling" conditions in the upper part of Figure 16. Specifically, in simulation condition 2, PBIL was adopted as the optimization method, with the number of individuals m being 60, the number of iterations iter being 200, the learning rate ε being 0.09, the number of individuals N used for updating the random variable being 2, and the number of weak optimizations g being 60. Here, each of the weak optimizations No. 1 to No. 60 uses data obtained from 80% of the total data. Furthermore, the data used by weak optimizations No. 1 to No. 60 are all different.

[0081] Therefore, under simulation condition 2, all data is used in step S70 and in step S73a. In this PBIL using the concept of ensemble learning, when the simulation was run 100 times, the number of times all connection phases of the 32 pole-mounted transformers were correctly identified was 42.

[0082] In simulation condition 3, all of the conditions in the upper part of Figure 16—"optimization method," "ensemble learning," and "random sampling"—were used. In other words, in simulation condition 2, BIL was adopted as the optimization method, with 60 individuals m, 200 iterations iter, a learning rate ε of 0.09, 2 individuals N used for updating the random variable, and 60 weak optimizations g. Furthermore, for random sampling, 80% of the total data was used for each of the g optimization evaluation functions.

[0083] Therefore, under simulation condition 3, the process shown in Figure 7 is executed as step S70. In this PBIL using the concept of ensemble learning and random sampling, when the simulation was run 100 times, the number of times all connection phases of the 32 pole-mounted transformers were correctly identified was 52.

[0084] Thus, by applying the concept of ensemble learning and random sampling techniques to the PBIL method, the discrimination of the connected phase can be improved. In this embodiment, PBIL was used as the optimization method, but it is not limited to this, and for example, genetic algorithms (GA), particle swarm optimization (PSO), and tabu search (TS) may also be used.

[0085] =====Summary===== The connection phase determination device 10 of this embodiment has been described above. The connection phase determination unit 17 identifies the connection state θming in each of the weak optimizations No. 1 to No. g from the g weak optimization groups by repeating steps S74a to S78b. Then, as shown in the lower part of Figure 15 and the bottom part of Figure 8, the connection phase determination unit 17 determines the most likely connection state θmin among the g connection states θming of weak optimizations No. 1 to No. g (step S79a). By performing this process, the connection phase can be determined with higher accuracy compared to, for example, the case where a single minimization objective function is used.

[0086] Furthermore, the connection phase discrimination unit 17 generates g data sets using a random sampling method (S73a) to change the characteristics of each of the g weak optimization evaluation functions, for example, from one year's worth of data (corresponding to the first data) as shown in Figure 10. Therefore, different data (corresponding to the second data) is used for each of the weak optimizations No. 1 to No. g. As a result, the accuracy of connection phase discrimination can be improved compared to the case where a single minimization objective function is used, for example, as shown in simulation condition 2.

[0087] Furthermore, the connection phase discrimination unit 17 calculates the evaluation value f(θ)kt for the time cross-sections tk~tn of the symmetric data, and then calculates the sum of all time cross-sections for individual k based on equation (4) (step S76a). Through this process, g evaluation values ​​that take into account multiple conditions (time cross-sections) can be obtained.

[0088] Furthermore, the connection phase determination unit 17 calculates, for example, in step S79a, the majority vote of the connection phase types (here, UV phase, VW phase, WU phase) for each of the multiple pole-mounted transformers with g connection states θming. Then, the connection phase determination unit 17 determines the connection phase for each of the multiple pole-mounted transformers based on the result of the majority vote calculation. By performing such processing, the accuracy of connection phase determination can be improved.

[0089] The embodiments described above are provided to facilitate understanding of the present invention and are not intended to limit its interpretation. Furthermore, the present invention may be modified or improved without departing from its spirit, and it goes without saying that equivalents thereof are included. [Explanation of Symbols]

[0090] 1. Connection Phase Identification System 10. Connection Phase Identification Device 11. Sensor Measurement Value Database 12 Distribution system configuration DB 13. Smart meter readings 14. Transformer current calculation unit 15. Data Selection Section 16 Voltage and Current Analysis Unit 17. Connection Phase Determination Unit 18 Result Output Section SE1 Switch TR1~TR3 Pole-mounted transformers SM1~SM4 Smart Meters

Claims

1. A connection phase determination device is applied to a connection phase determination system comprising a plurality of transformers connected to each phase of a power distribution system, sensors for measuring the current of each phase, and a plurality of energy meters for measuring the amount of electricity supplied from the plurality of transformers to consumers, and determines the connection phase of the plurality of transformers connected to the power distribution system, Based on the measured values ​​obtained from the multiple energy meters, the connection state that minimizes the evaluation value corresponding to the difference between the first current of each phase when assuming the connection state of each of the multiple transformers and the second current of each phase obtained from the measurement results of the sensors is identified a predetermined number of times using an optimization method with a predetermined number of different conditions. Based on the specified predetermined number of connection states, the connection phase of each of the plurality of transformers connected to the power distribution system is determined. Connection phase identification device.

2. A connection phase determination device according to claim 1, In each of the predetermined number of optimization methods, a predetermined proportion of second data is used, randomly selected from first data which includes the measured values ​​of the plurality of electricity meters measured over a predetermined period and the second current of each phase. Connection phase identification device.

3. A connection phase determination device according to claim 1, The aforementioned evaluation value is the sum of the differences between the first current and the second current for each of several time sections. Connection phase identification device.

4. A connection phase determination device according to any one of claims 1 to 3, From the predetermined number of connection states, calculate the majority vote for the type of connection phase for each of the multiple transformers. Based on the results of the majority vote, the connected phase for each of the multiple transformers is determined. Connection phase identification device.

5. Multiple transformers connected to each phase of the power distribution system, A sensor for measuring the current of each of the aforementioned phases, Multiple electricity meters for measuring the amount of electricity supplied from the multiple transformers to the consumer, A connection phase determination device for determining the connection phases of the plurality of transformers connected to the power distribution system, Equipped with, The aforementioned connection phase determination device is Based on the measured values ​​obtained from the multiple energy meters, the connection state that minimizes the evaluation value corresponding to the difference between the first current of each phase when assuming the connection state of each of the multiple transformers and the second current of each phase obtained from the measurement results of the sensors is identified a predetermined number of times using an optimization method with a predetermined number of different conditions. Based on the specified predetermined number of connection states, the connection phase of each of the plurality of transformers connected to the power distribution system is determined. Connection phase identification system.

Images