Power system state estimation device, computer program for power system state estimation device, and power system state estimation method

Abstract

To provide a power system state estimation device capable of accurately grasping a state of a power system.SOLUTION: A power system state estimation device 1 includes: an observation object determination unit 201 which determines whether a power system 90 includes a measurement point where a measurement point is lost, on the basis of measurement values regarding electric quantities measured at a plurality of measurement points of the power system 90 which supplies power; an estimation AI model creation unit 205 which, when the observation object determination unit 201 has determined that the power system 90 has a measurement point where the measurement value has lost, creates an arithmetic model which estimates a missing value being a lost measurement value, on the basis of a measurement value measured at another measurement point related to a measurement point where a measurement value; and a missing value estimation unit 206 which calculates the missing value with the arithmetic model created by the estimation AI model creation unit 205.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present embodiment relates to a power system state estimation device, a computer program for the power system state estimation device, and a power system state estimation method that estimate the state of a power system for which no measurement value is available. [Background technology]

[0002] There are known systems that estimate the state of a power system by calculating estimated values ​​of measurement errors in the power system and correction amounts based on power flow calculations.There are also known systems that estimate the state of a power system based on estimated state values ​​of a power system for which measured values ​​are available and state constraint values ​​of a power system for which measured values ​​are not available. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2008-154418 [Patent Document 2] Japanese Patent Application Laid-Open No. 2015-541311 Summary of the Invention [Problem to be solved by the invention]

[0004] In order to control the power system appropriately, it is necessary to grasp the state of the power system with high accuracy. Because the power flow in the power system changes due to load fluctuations and the like, it is preferable to grasp the state of the entire power system successively. Furthermore, with the recent liberalization of the power industry, the use of renewable energy such as solar and wind power has been promoted, and the state of the power system is changing even more significantly. For this reason, it is necessary to grasp the state of the power system with high accuracy.

[0005] The state of a power grid is estimated by power flow calculations based on measured values ​​such as voltage and current measured by sensors installed in the power transmission grid and system configuration data. However, there are cases where the measured values ​​of the power grid cannot be obtained due to malfunctions of sensors installed in the power grid. When the measured values ​​of the power grid cannot be obtained, it is inconvenient that the state of the power grid cannot be accurately grasped.

[0006] In conventional technology, when data on measured values ​​such as active power and reactive power in loads, generators, or transmission lines is missing due to a malfunction of a sensor or the like, state estimation is performed by retaining the previous value, which complements the data from the immediately previous measured value (sometimes called the "previous value").

[0007] However, when state estimation is performed by holding previous values, the previous values ​​and the actual measured values ​​diverge over time, which poses a problem in that the state of the power system cannot be grasped accurately.

[0008] In view of the above-mentioned problems, the present embodiment aims to provide a power system state estimation device, a computer program for the power system state estimation device, and a power system state estimation method that can accurately grasp the state of a power system. [Means for solving the problem]

[0009] The power system state estimating device of this embodiment has the following features. (1) The power system has an observation target determination unit that determines whether the power system includes a measurement point where measurements are missing, based on measurements of electrical quantities measured at multiple measurement points in the power system that supplies power. (2) When the observation target determination unit determines that the power system includes a measurement point with missing measurement values, the power system has an estimation AI model creation unit that creates a computational model that estimates the missing measurement values ​​based on measurement values ​​measured at other measurement points related to the measurement point with missing measurement values. (3) A missing value estimation unit that calculates the missing values ​​using the calculation model created by the estimation AI model creation unit. [Brief explanation of the drawings]

[0010] [Figure 1] FIG. 1 is a diagram showing a configuration of a power system state estimating device 1 according to a first embodiment. [Figure 2] FIG. 1 is a diagram showing a program flow of a power system state estimating device 1 according to a first embodiment. [Figure 3] FIG. 1 is a diagram showing measurement value data D301 of the power system state estimating device 1 according to the first embodiment. [Figure 4] FIG. 10 is a diagram showing model data D304 of the power system state estimating device 1 according to the first embodiment. [Figure 5] FIG. 1 is a diagram showing input and output data of an estimation AI model creation unit 205 of a power system state estimation device 1 according to a first embodiment. [Figure 6] FIG. 10 is a diagram showing estimation AI model data D306 of the power system state estimation device 1 according to the first embodiment. [Figure 7] FIG. 10 is a diagram showing estimation data D305 of the power system state estimating device 1 according to the first embodiment. [Figure 8] FIG. 1 is a diagram showing input and output data of a missing value estimation unit 206 of a power system state estimation device 1 according to a first embodiment. [Figure 9] FIG. 10 is a diagram showing state estimation value data D308 of the power system state estimating device 1 according to the first embodiment. [Figure 10] FIG. 10 is a diagram illustrating fluctuations in measured values ​​at a node with respect to time according to the second embodiment. [Figure 11] FIG. 10 is a diagram showing input and output data of the estimation AI model creation unit 205 of the power system state estimation device 1 according to the second embodiment. [Figure 12] FIG. 10 is a diagram showing estimation AI model data D306 of the power system state estimation device 1 according to the second embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0011] [First embodiment] [1-1.Configuration] As an example of this embodiment, the configuration of a power system state estimating device 1 will be described with reference to Fig. 1. In this embodiment, when there are multiple devices or components with the same configuration, they will be described with the same number, and when describing each individual device or component with the same configuration, they will be distinguished by adding an alphabetic suffix to the common number.

[0012] The power system state estimation device 1 is configured by a computer. The power system state estimation device 1 estimates the state of a power system including measurement points with missing measurement values. The power system state estimation device 1 is used to appropriately control or manage the power system.

[0013] The power system state estimating device 1 includes an input unit 10, a calculation unit 20, a storage unit 30, and an output unit 40.

[0014] The input unit 10 is configured with a transmission / reception circuit. The input unit 10 may be configured with an external memory connection circuit, or an operation device such as a touch switch, keyboard, or mouse arranged on a display. The input unit 10 is connected to the calculation unit 20 and the storage unit 30.

[0015] The input unit 10 receives data transmitted from outside the power system state estimating device 1 or data input by an operator.

[0016] As an example, the input unit 10 is connected to a sensor 80 (described later) via a communication line 70. The sensor 80 is disposed in a power system 90. The sensor 80 measures the active power, reactive power, etc. in the load, generator, or transmission line of the power system 90 and transmits the measured value data D301.

[0017] The input unit 10 receives measurement value data D301 from the sensor 80 via the communication line 70. The input unit 10 stores the received measurement value data D301 in a measurement value database 301 (sometimes referred to as "measurement value DB301") of the storage unit 30, which will be described later.

[0018] The input unit 10 may be configured so that the measurement value data D301 is inputted by an external memory connection circuit or an operating device such as a touch switch, keyboard, or mouse arranged on a display.

[0019] The input unit 10 also receives commands input by an operator. For example, the input unit 10 receives a command related to estimating the state of a power system. The input unit 10 transmits the received command to the calculation unit 20.

[0020] The calculation unit 20 is configured by a CPU (Central Processing Unit) of a computer, etc. The calculation unit 20 may be configured by a software module. The calculation unit 20 is connected to the input unit 10, the memory unit 30, and the output unit 40. The calculation unit 20 has an observation target determination unit 201, a measured system state calculation unit 202, a previous value retention data selection unit 203, a model data selection unit 204, an estimation AI model creation unit 205, a missing value estimation unit 206, a supplementary data selection unit 207, and a state estimation unit 208. The calculation unit 20 performs calculation processing using each of the above units.

[0021] The observation target determination unit 201 determines whether the power system 90 includes any measurement points with missing measurement values, based on the measurement value data D301 stored in the measurement value DB 301. The measurement points are located at nodes in the generators, loads, or transmission lines of the power system 90. An observation target is a node from which measurement values ​​used in the power flow calculation can be obtained. A non-observation target is a node from which measurement values ​​used in the power flow calculation cannot be obtained.

[0022] Furthermore, the observation target determination unit 201 extracts data used to estimate missing values ​​from the measurement value data D301 stored in the measurement value DB301, and stores the data as estimation data D305 in an estimation database 305 (sometimes referred to as "estimation DB305") in the storage unit 30. Missing values ​​are measurement values ​​that are missing due to, for example, a failure of a sensor placed at a measurement point.

[0023] When the observation target determination unit 201 determines that the power system 90 does not include a measurement point with missing measurement values, the measured system state calculation unit 202 estimates the state of the power system 90. The measured system state calculation unit 202 estimates the state of the power system 90 based on the measurement value data D301 stored in the measurement value DB301 of the storage unit 30 and the equipment data D302 stored in the equipment database 302 (sometimes referred to as "equipment DB302").

[0024] The equipment data D302 is information relating to equipment such as loads, generators, and transmission lines arranged in the power system 90. The equipment data D302 is set in advance and stored in the equipment DB 302. The measured system state calculation unit 202 estimates the state of the power system 90 and transmits the estimation result to the output unit 40 as state estimation value data D308a.

[0025] When the observation target determination unit 201 determines that the power system 90 includes a measurement point with a missing measurement value, the previous value retention data selection unit 203 extracts a previous value from the measurement value data D301 stored in the measurement value DB 301. The previous value is a measurement value immediately before the measurement value becomes missing at the measurement point with a missing measurement value in the power system 90. The previous value retention data selection unit 203 extracts the previous value at the measurement point with a missing measurement value, and stores it as previous value retention data D303 in the previous value retention database 303 (sometimes referred to as "previous value retention DB 303") of the storage unit 30.

[0026] When the observation target determination unit 201 determines that the power system 90 includes a measurement point with missing measurement values, the model data selection unit 204 extracts data to be used for creating an AI model from the measurement value data D301 stored in the measurement value DB301. The model data selection unit 204 extracts data from the measurement value data D301 and creates model data D304. The model data selection unit 204 stores the created model data D304 in a model database 304 (sometimes referred to as "model DB304") in the storage unit 30. The model data D304 may be so-called training data used for calculations using machine learning.

[0027] The estimation AI model creation unit 205 creates a computation model using an AI model. The computation model is used to estimate measurement values ​​at measurement points where measurement values ​​are missing in the power system 90. The estimation AI model creation unit 205 creates a computation model for estimating measurement values ​​at measurement points where measurement values ​​are missing by machine learning, based on the model data D304 stored in the model DB 304 of the storage unit 30.

[0028] The estimation AI model creation unit 205 stores the created computation model as estimation AI model data D306 in an estimation AI model database 306 (sometimes referred to as "estimation AI model DB306") in the storage unit 30. The model data selection unit 204, model DB304, and estimation AI model creation unit 205 constitute the AI ​​model creation unit 21.

[0029] The missing value estimation unit 206 estimates measurement values ​​at measurement points where measurement values ​​are missing. The missing value estimation unit 206 estimates missing values, which are missing measurement values, based on estimation data D305 stored in an estimation DB 305 in the storage unit 30 and estimation AI model data D306 stored in an estimation AI model DB 306.

[0030] Missing value estimation unit 206 stores the estimated missing values ​​as estimated missing value data D307 in estimated missing value database 307 (sometimes referred to as "estimated missing value DB307") in storage unit 30. Missing value estimation unit 206, estimation DB305, and estimated AI model DB306 constitute AI estimated value creation unit 22.

[0031] The interpolation data selection unit 207 selects data related to the state estimation of the power system 90. If estimated missing value data D307 is stored in the estimated missing value DB 307 of the storage unit 30, the interpolation data selection unit 207 selects the estimated missing value data D307. If estimated missing value data D307 is not stored in the estimated missing value DB 307 of the storage unit 30, the interpolation data selection unit 207 selects the previous value storage data D303 stored in the previous value storage DB 303.

[0032] The state estimation unit 208 estimates the state of the power system 90. The state estimation unit 208 estimates the state based on the estimated missing value data D307 or the previous value retention data D303 selected by the interpolation data selection unit 207, and the equipment data D302 stored in the equipment DB302 of the storage unit 30.

[0033] The state estimation unit 208 estimates the state of the power system 90 and transmits the estimation result to the output unit 40 as state estimation value data D308.

[0034] The storage unit 30 is configured with a storage medium such as a hard disk or semiconductor memory. The storage unit 30 is connected to the calculation unit 20 and the input unit 10. The storage unit 30 has data writing and reading controlled by the calculation unit 20 or the input unit 10.

[0035] The memory unit 30 has a measurement value database 301 (sometimes referred to as "measurement value DB301"), an equipment database 302 (sometimes referred to as "equipment DB302"), a previous value holding database 303 (sometimes referred to as "previous value holding DB303"), a model database 304 (sometimes referred to as "model DB304"), an estimation database 305 (sometimes referred to as "estimation DB305"), an estimated AI model database 306 (sometimes referred to as "estimated AI model DB306"), and an estimated missing value database 307 (sometimes referred to as "estimated missing value DB307").

[0036] The measurement value DB301 stores measurement value data D301. The measurement value data D301 is data related to measurement values ​​such as active power and reactive power in loads, generators, or transmission lines of the power system 90. The measurement values ​​related to the measurement value data D301 are measured by a sensor 80 arranged in the power system 90. The measurement value data D301 is transmitted from the sensor 80 to the input unit 10 and stored in the measurement value DB301. The measurement value DB301 accumulates and stores the measurement value data D301.

[0037] The equipment DB 302 stores equipment data D302. The equipment data D302 is information relating to equipment such as loads, generators, and transmission lines arranged in the power system 90. The equipment DB 302 stores preset equipment data D302.

[0038] The previous value retention DB303 stores the previous value retention data D303. The previous value retention data D303 is data in which previous values ​​at measurement points where measurement values ​​are missing in the power system 90 are extracted from the measurement value data D301. The previous value retention data D303 is created by the previous value retention data selection unit 203. The previous value retention DB303 accumulates and stores the previous value retention data D303.

[0039] The model DB304 stores model data D304. The model data D304 is data used to create an AI model by the estimation AI model creation unit 205. The model data D304 may be so-called teacher data used for calculations by machine learning. The model data D304 is created by the model data selection unit 204. The model DB304 accumulates and stores the model data D304.

[0040] The estimation DB 305 stores estimation data D305. The estimation data D305 is data used by the missing value estimation unit 206 to estimate missing values. A missing value is a measurement value that is missing due to a malfunction of a sensor placed at a measurement point, for example. The estimation data D305 is created by the observation target determination unit 201. The estimation DB 305 accumulates and stores the estimation data D305.

[0041] The estimated AI model DB306 stores the estimated AI model data D306. The estimated AI model data D306 is data related to a computation model that estimates missing values. The estimated AI model data D306 is created by the estimated AI model creation unit 205. The estimated AI model DB306 accumulates and stores the estimated AI model data D306.

[0042] The estimated missing value DB307 stores estimated missing value data D307. The estimated missing value data D307 is data related to estimated missing values ​​at measurement points where measurement values ​​are missing. The estimated missing value data D307 is created by the missing value estimation unit 206. The estimated missing value DB307 accumulates and stores the estimated missing value data D307.

[0043] The output unit 40 is composed of a display device such as a liquid crystal display, a printing device such as a printer, a transmission circuit, an external memory connection circuit, etc. The output unit 40 is connected to the calculation unit 20. The output unit 40 outputs data created by the calculation unit 20. The output unit 40 outputs the data created by the calculation unit 20 as a display, printout, or electronic data.

[0044] The output unit 40 may transmit the data created by the calculation unit 20 to an external device. The external device connected to the output unit 40 may be an I / O device such as a printer or a monitor, an external storage device, or an external control device or computer device.

[0045] The output unit 40 outputs, as a result of the state estimation, state estimation value data D308a created by the measured system state calculation unit 202 of the calculation unit 20, or state estimation value data D308b created by the state estimation unit 208, or state estimation value data D308c.

[0046] The power system 90 is a power supply network that supplies generated power to loads. The power system 90 is made up of loads, generators, transmission lines, and the like.

[0047] The sensor 80 is configured by a detection device that detects state quantities such as active power, reactive power, and voltage. The sensor 80 is placed at measurement points such as loads, generators, and transmission lines in the power system 90. A plurality of sensors 80 are placed in the power system 90. The sensor 80 transmits the detected state quantities such as active power, reactive power, and voltage to the power system state estimation device 1 via the communication line 70.

[0048] The communication line 70 is a communication network formed by dedicated lines. The communication line 70 may be formed by a communication line such as the Internet or a telephone line. Communication is carried out between the power system state estimation device 1 and the sensors 80 arranged in the power system 90 via the communication line 70.

[0049] The above is the configuration of the power system state estimating device 1 according to this embodiment.

[0050] [1-2. Effect] Next, an outline of the operation of the power system state estimating device 1 of this embodiment will be described with reference to FIGS.

[0051] The power system state estimation device 1 estimates the state of a power system including measurement points with missing measurement values. The power system state estimation device 1 is used to appropriately control or manage the power system.

[0052] The observation target determination unit 201 of the calculation unit 20 determines whether the power system 90 includes any measurement points with missing measurements, based on measurements of electrical quantities measured at multiple measurement points in the power system 90 that supplies power.

[0053] When the observation target determination unit 201 determines that the power system 90 includes a measurement point with missing measurement values, the estimation AI model creation unit 205 of the calculation unit 20 creates a calculation model that estimates the missing measurement values, which are the missing measurement values, based on measurement values ​​measured at other measurement points related to the measurement point with missing measurement values.

[0054] The estimation AI model creation unit 205 uses machine learning to learn the relationship between measurement values ​​at a measurement point where measurement values ​​are missing that were measured at the same time in the past and measurement values ​​at other measurement points related to the measurement point where measurement values ​​are missing, and creates a computational model that estimates the missing values.

[0055] The missing value estimation unit 206 of the calculation unit 20 calculates the missing values ​​using the calculation model created by the estimation AI model creation unit 205.

[0056] When the missing value estimation unit 206 does not calculate a missing value, the previous value retention data selection unit 203 of the calculation unit 20 extracts, from the measurement values ​​related to the electrical quantity measured at multiple measurement points in the power system 90, the previous value, which is the measurement value immediately before the measurement value at the measurement point where the measurement value is missing,

[0057] The interpolation data selection unit 207 of the calculation unit 20 selects the missing value calculated by the missing value estimation unit 206 or the previous value extracted by the previous value retention data selection unit 203 as the missing value.

[0058] The state estimation unit 208 of the calculation unit 20 estimates the state of the power system 90 based on the missing values ​​selected by the interpolation data selection unit 207, and causes the output unit 40 to output the estimation result.

[0059] The electrical quantity in the above means a quantity related to active power, reactive power, and voltage. The electrical quantity may also be a quantity related to current, frequency, power factor, and phase angle.

[0060] In the above, "measured values ​​measured at the same time" may be measured values ​​measured in a time period before or after a predetermined measurement time. For example, if measurements are taken every 30 minutes, they may be measured values ​​measured within one minute before or after the measurement time. "The same time" does not necessarily mean the same time.

[0061] The other related measurement points in the above are, for example, measurement points adjacent to the measurement point with a missing measurement value, measurement points located in the vicinity of the measurement point with a missing measurement value, or measurement points that affect the measurement value of the measurement point with a missing measurement value.

[0062] The details of the operation of the power system state estimating device 1 according to the first embodiment are as follows.

[0063] The input unit 10 receives data transmitted from outside the power system state estimation device 1 or data input by an operator. The input unit 10 receives measurement value data D301 from a sensor 80 via a communication line 70. The sensor 80 is disposed in the power system 90. The sensor 80 measures active power, reactive power, etc. in a load, a generator, or a transmission line of the power system 90.

[0064] The input unit 10 accumulates and stores the received measurement value data D301 in the measurement value DB 301 of the storage unit 30.

[0065] An example of the measurement value data D301 is shown in Fig. 3. As shown in Fig. 3(a), the measurement value data D301 indicates the active power P, reactive power Q, and voltage V in the load of the power system 90, the active power P, reactive power Q, and voltage V in the generator, and the active power P and reactive power Q in the transmission line.

[0066] As shown in Fig. 3(b), the measurement value data D301 indicates the measurement value of active power P, reactive power Q, or voltage V at the measurement point of the node for each time. In Fig. 3(b), time t0 is the current time, time t-1 is the previous measurement time, and time t-2 is the measurement time before the previous one. The measurement point for the generator of power system 90 is located at node a, the measurement point for the load of power system 90 is located at node g, and the measurement points on the transmission line are located at nodes b to f.

[0067] For example, at the current time t0, the active power at node a is Pa0, the reactive power is Qa0, and the voltage is Va0. At the current time t-1, the active power at node d is Pd-1 and the reactive power is Qd-1.

[0068] The observation target determination unit 201 of the calculation unit 20 determines whether the power system 90 includes any measurement points with missing measurement values, based on the measurement value data D301 stored in the measurement value DB 301. The measurement points are located at nodes in the generators, loads, or transmission lines of the power system 90. An observation target is a node from which measurement values ​​used in the power flow calculation can be obtained. A non-observation target is a node from which measurement values ​​used in the power flow calculation cannot be obtained.

[0069] When the observation target determination unit 201 determines that the power system 90 does not include a measurement point with missing measurement values, the measured power system state calculation unit 202 estimates the state of each node of the power system 90.

[0070] The measured system state calculation unit 202 estimates the state of the power system 90 based on the measurement value data D301 stored in the measurement value DB 301 of the storage unit 30 and the equipment data D302 stored in the equipment DB 302.

[0071] The equipment data D302 is information relating to equipment such as loads, generators, and transmission lines arranged in the power system 90. The equipment data D302 is set in advance and stored in the equipment DB 302. The measured system state calculation unit 202 estimates the state of the power system 90 and transmits the estimation result to the output unit 40 as state estimation value data D308a.

[0072] When the observation target determination unit 201 determines that the power system 90 includes a measurement point with missing measurement values, the previous value retention data selection unit 203 extracts the previous value from the measurement value data D301 stored in the measurement value DB301.

[0073] The previous value is the measurement value immediately before the measurement value is missing at the measurement point where the measurement value is missing in the power system 90. The immediately previous measurement value is the measurement value from the previous measurement. For example, if the active power Pd0 and reactive power Qd0 at node d at current time t0 shown in FIG. 3(b) are missing, the active power Pd-1 and reactive power Qd-1 at node d at time t-1 are extracted as the previous value.

[0074] The previous value retention data selection unit 203 extracts previous values ​​at measurement points where measurement values ​​are missing, creates previous value retention data D303, and stores it in the previous value retention DB 303 of the storage unit 30.

[0075] When the observation target determination unit 201 determines that the power system 90 includes a measurement point with missing measurement values, the model data selection unit 204 extracts data to be used for creating an AI model from the measurement value data D301 stored in the measurement value DB 301. The model data selection unit 204 extracts data from the measurement value data D301 and creates model data D304. The model data selection unit 204 stores the created model data D304 in the model DB 304 of the storage unit 30. The model data D304 may be so-called training data used for calculations by machine learning.

[0076] An example of the model data D304 is shown in Fig. 4. When the active power Pd0 and reactive power Qd0 at node d at the current time t0 shown in Fig. 3(b) are missing, for example, the active powers Pd-1, Pd-2, Pd-3 and reactive powers Qd-1, Qd-2, Qd-3 at node d, the active powers Pc-1, Pc-2, Pc-3 and reactive powers Qc-1, Qc-2, Qc-3 at node c adjacent to node d, and the active powers Pe-1, Pe-2, Pe-3 and reactive powers Qe-1, Qe-2, Qe-3 at node e are extracted from the measurement value data D301 and set as the model data D304.

[0077] Active powers Pd-1, Pc-1, Pe-1 and reactive powers Qd-1, Qc-1, Qe-1 are data relating to measurements taken at time t-1. Active powers Pd-2, Pc-2, Pe-2 and reactive powers Qd-2, Qc-2, Qe-2 are data relating to measurements taken at time t-2. Active powers Pd-3, Pc-3, Pe-3 and reactive powers Qd-3, Qc-3, Qe-3 are data relating to measurements taken at time t-3.

[0078] The model data D304 shown in Figure 4 is assumed to be data related to measurements taken at node d and adjacent nodes c and e, but the nodes related to the measurements are not limited to this. The model data D304 shown in Figure 4 is not limited to measurements at adjacent nodes, and may include data related to measurements at measurement points located in the vicinity of a measurement point with a missing measurement value, and measurement points that affect the measurement value of the measurement point with a missing measurement value. For example, if the active power and reactive power at node d are affected by nodes a and b, the model data D304 may include data related to the active power and reactive power at nodes a and b.

[0079] 4 is data relating to measurement values ​​measured at times t-1, t-2, and t-3, the times at which the measurements were taken are not limited to these. The model data D304 may include data relating to measurement values ​​measured before time t-4 or at the current time t0.

[0080] The estimation AI model creation unit 205 creates a computation model using an AI model. The created computation model is used to estimate missing values ​​at measurement points where measurement values ​​are missing in the power system 90. The estimation AI model creation unit 205 creates a computation model for estimating missing values ​​at measurement points where measurement values ​​are missing by machine learning, based on the model data D304 stored in the model DB 304 of the storage unit 30.

[0081] 5 shows an example of input / output data for the estimation AI model creation unit 205. The estimation AI model creation unit 205 stores the created computation model as estimation AI model data D306 in the estimation AI model DB 306 of the storage unit 30. The model data selection unit 204, model DB 304, and estimation AI model creation unit 205 constitute the AI ​​model creation unit 21.

[0082] The estimation AI model creation unit 205 creates an AI model through machine learning using an LSTM (Long Short Term Memory) neural network. Based on the past measurement data of the unobserved object and the past measurement data of the observed object adjacent to the unobserved object, it creates a calculation model as a missing value estimation AI model.

[0083] The calculation model may be a calculation block, a calculation parameter, or a calculation formula. An example of the estimation AI model data D306 is shown in Fig. 6. Fig. 6 shows a case where the calculation model related to the estimation AI model data D306 is a calculation block based on an AI model.

[0084] The estimation AI model creation unit 205 creates a calculation block F1 through machine learning using the model data D304 as training data. As an example, if the active power Pd0 and reactive power Qd0 at node d at the current time t0 are missing, the calculation block F1 is created as follows:

[0085] The estimation AI model creation unit 205 learns the relationship between the active power Pc-1 and reactive power Qc-1 of node c adjacent to node d, the active power Pe-1 and reactive power Qc-1 of node e, and the active power Pd-1 and reactive power Qd-1 of node d at time t-1. The estimation AI model creation unit 205 also learns the relationship between the active power Pc-2 and reactive power Qc-2 of node c, the active power Pe-2 and reactive power Qc-2 of node e, and the active power Pd-2 and reactive power Qd-2 of node d at time t-2. The estimation AI model creation unit 205 also learns the relationship between the active power Pc-3 and reactive power Qc-3 of node c, the active power Pe-3 and reactive power Qc-3 of node e, and the active power Pd-3 and reactive power Qd-3 of node d at time t-3.

[0086] The estimation AI model creation unit 205 creates a calculation block F1 using an AI model based on the learned relationships between the active power and reactive power of node c, the active power and reactive power of node e, and the active power and reactive power of node d. The calculation block F1 is a functional block that receives the active power Pc0 and reactive power Qc0 of node c and the active power Pe0 and reactive power Qe0 of node e at the current time t0 as inputs, and outputs the active power Pd0X and reactive power Qd0X of node d.

[0087] The estimation AI model creation unit 205 stores the created operation block F1 in the estimation AI model DB 306 of the storage unit 30 as estimation AI model data D306.

[0088] The estimation AI model creation unit 205 may use the model data D304 as training data to create calculation parameters or calculation formulas for the AI ​​model as a calculation model through machine learning. As an example, if the active power Pd0 and reactive power Qd0 at node d at current time t0 are missing, the calculation parameters for the AI ​​model are created as follows:

[0089] The estimation AI model creation unit 205 creates the following (Equation 1), (Equation 2), and (Equation 3) based on the model data D304.

number

[0090] (Equation 1) represents the relationship between the active power Pc-1 and reactive power Qc-1 of node c, the active power Pe-1 and reactive power Qe-1 of node e, and the active power Pd-1 and reactive power Qd-1 of node d at time t-1.

[0091] (Equation 2) represents the relationship between the active power Pc-2 and reactive power Qc-2 of node c, the active power Pe-2 and reactive power Qe-2 of node e, and the active power Pd-2 and reactive power Qd-2 of node d at time t-2.

[0092] Equation 3 represents the relationship between the active power Pc-3 and reactive power Qc-3 of node c, the active power Pe-3 and reactive power Qe-3 of node e, and the active power Pd-3 and reactive power Qd-3 of node d at time t-3.

[0093] In (Equation 1), (Equation 2), and (Equation 3), K1 and K2 are matrices composed of coefficients or functions. The estimation AI model creation unit 205 calculates K1 and K2, which are common to (Equation 1), (Equation 2), and (Equation 3), and creates the following (Equation 4).

number

[0094] Equation 4 is a mathematical formula for calculating the active power Pd0X and reactive power Qd0X of node d based on the active power Pc0 and reactive power Qc0 of node c and the active power Pe0 and reactive power Qe0 of node e at current time t0. The estimation AI model creation unit 205 stores K1 and K2 as estimation AI model data D306 in the estimation AI model DB 306 of the storage unit 30. K1 and K2 are calculation parameters of the AI ​​model.

[0095] The estimation AI model creation unit 205 may store the created (Equation 4) in the estimation AI model DB 306 of the storage unit 30 as an arithmetic equation related to the estimation AI model data D306.

[0096] When the observation target determination unit 201 determines that the power system 90 includes a measurement point with missing measurement values, it extracts data to be used for estimating the missing values ​​from the measurement value data D301 stored in the measurement value DB 301. The observation target determination unit 201 extracts data from the measurement value data D301 and creates estimation data D305. The observation target determination unit 201 stores the created estimation data D305 in the estimation DB 305 of the storage unit 30.

[0097] An example of the estimation data D305 is shown in Fig. 7. When the active power Pd0 and reactive power Qd0 at node d at the current time t0 are missing, for example, the active power Pc0 and reactive power Qc0 at node c adjacent to node d, and the active power Pe0 and reactive power Qe0 at node e are extracted from the measurement value data D301 and used as the estimation data D305.

[0098] The missing value estimation unit 206 estimates missing values ​​at measurement points where measurement values ​​are missing. The missing value estimation unit 206 calculates missing values, which are missing measurement values, based on estimation data D305 stored in an estimation DB 305 in the storage unit 30 and estimation AI model data D306 stored in an estimation AI model DB 306.

[0099] 8 shows an example of input / output data of the missing value estimation unit 206. The missing value estimation unit 206 estimates the missing values ​​of active power Pd0 and reactive power Qd0 using the above-mentioned calculation block F1, calculation parameters K1 and K2, or calculation formula (Formula 4), and calculates them as active power Pd0X and reactive power Qd0X, respectively.

[0100] The missing value estimation unit 206 creates estimated missing value data D307 based on the calculated active power Pd0X and reactive power Qd0X. The missing value estimation unit 206 stores the created estimated missing value data D307 in an estimated missing value DB 307 in the storage unit 30. The missing value estimation unit 206, estimation DB 305, and estimated AI model DB 306 constitute the AI ​​estimated value creation unit 22.

[0101] The missing value estimation unit 206 estimates missing values, which are missing measurement values ​​for the active power Pd0 and reactive power Qd0 at node d at the current time t0, and sets the estimated missing value data D307.

[0102] The interpolation data selection unit 207 selects data related to the state estimation of the power system 90. When estimated missing value data D307 is stored in the estimated missing value DB 307 of the storage unit 30, the interpolation data selection unit 207 selects the estimated missing value data D307 as the missing value. When estimated missing value data D307 is not stored in the estimated missing value DB 307 of the storage unit 30, the interpolation data selection unit 207 selects the previous value retention data D303 stored in the previous value retention DB 303 as the missing value.

[0103] The state estimation unit 208 estimates the state of the power system 90. The state estimation unit 208 estimates the state based on the equipment data D302 stored in the equipment DB 302 of the storage unit 30, and the estimated missing value data D307 or the previous value retention data D303 selected as a missing value by the interpolation data selection unit 207.

[0104] The state estimation unit 208 performs state estimation when, for example, the measurement accuracy of the sensors 80 arranged in the power system 90 is poor. The state estimation unit 208 calculates the amount of correction to the system state by power flow calculation based on the measurement values ​​related to the measurement value data D301, the estimated missing value data D307 selected by the interpolation data selection unit 207, or the previous value retention data D303, and estimates the true value of the measurement value at each node of the power system 90.

[0105] The state estimation unit 208 creates state estimated value data D308c based on the estimated missing value data D307 and the equipment data D302. The state estimation unit 208 creates state estimated value data D308b based on the previous value retention data D303 and the equipment data D302. The state estimation unit 208 transmits the state estimated value data D308c or the state estimated value data D308b to the output unit 40.

[0106] State estimation value data D308c and D308b indicate the state of each node at the current time t0. The active power Pd0X and reactive power Qd0X in the state estimation value data D308c are estimated measurement values. The active power Pd-1 and reactive power Qd-1 in the state estimation value data D308b are measurement values ​​estimated by pre-emption.

[0107] When the input unit 10 receives a command instructing output of state estimation value data, the calculation unit 20 causes the output unit 40 to output state estimation value data D308a, state estimation value data D308b, or state estimation value data D308c. The command instructing output of state estimation value data is input to the input unit 10 by an operator.

[0108] The output unit 40 outputs, as a result of the state estimation, state estimate value data D308a created by the measured system state calculation unit 202 of the calculation unit 20, or state estimate value data D308b or state estimate value data D308c created by the state estimation unit 208. The output unit 40 outputs the state estimate value data D308a, state estimate value data D308b or state estimate value data D308c by display, printout or electronic data. Fig. 9 shows examples of the state estimate value data D308a, state estimate value data D308b and state estimate value data D308c.

[0109] FIG. 9 shows examples of state estimated value data D308a, state estimated value data D308b, and state estimated value data D308c in a case where the measurement values ​​related to the measurement value data D301, the estimated missing value data D307 selected as missing values, and the previous value retention data D303 match the true values ​​of the measurement values ​​at each node of the power system 90 calculated by the power flow calculation.

[0110] The output unit 40 may transmit the state estimation value data D308a, the state estimation value data D308b, or the state estimation value data D308c to an external device. The external device connected to the output unit 40 may be an I / O device such as a printer or a monitor, an external storage device, or an external control device or computer device.

[0111] The operation of the power system state estimating apparatus 1 may be realized by a computer program shown in Fig. 2. The computer program shown in Fig. 2 is stored in the calculation unit 20 of the power system state estimating apparatus 1.

[0112] (Step S01: Receiving measurement data) The calculation unit 20 receives measurement value data D301 via the input unit 10. The measurement value data D301 is transmitted from the sensor 80 via the communication line 70. The input unit 10 accumulates and stores the received measurement value data D301 in the measurement value DB 301 of the storage unit 30.

[0113] (Step S02: Determine whether measurement points with missing measurements are included) The calculation unit 20 determines whether the power system 90 includes a measurement point with missing measurement values. The operation in step S02 is executed by the observation target determination unit 201 of the calculation unit 20.

[0114] If it is determined that the power system 90 does not include a measurement point with missing measurements (NO in step S02), the program proceeds to step S03. If it is determined that the power system 90 includes a measurement point with missing measurements (YES in step S02), the program proceeds to step S04.

[0115] (Step S03: Estimation of the state of the power system) If it is not determined in step S02 that power system 90 includes a measurement point with missing measurement values, calculation unit 20 estimates the state of power system 90. The operation in step S03 is executed by measured system state calculation unit 202 of calculation unit 20. Based on measurement value data D301 stored in measurement value DB301 of storage unit 30 and equipment data D302 stored in equipment DB302, measured system state calculation unit 202 estimates the state of power system 90 and creates state estimation value data D308a.

[0116] (Step S04: Extraction of previous value) If it is determined in step S02 that the power system 90 includes a measurement point where a measurement value is missing, the calculation unit 20 extracts a previous value at the measurement point where a measurement value is missing. The operation in step S04 is executed by the previous value retention data selection unit 203 of the calculation unit 20. The previous value retention data selection unit 203 extracts a previous value from the measurement value data D301 stored in the measurement value DB 301. The previous value is a measurement value at a node where a measurement value is missing in the power system 90, immediately before the measurement value is missing.

[0117] The previous value retention data selection unit 203 extracts previous values ​​at measurement points where measurement values ​​are missing, creates previous value retention data D303, and stores it in the previous value retention DB 303 of the storage unit 30.

[0118] (Step S05: Extraction of data used to create an AI model) The calculation unit 20 extracts data used to create an AI model. The operation in step S05 is executed by the model data selection unit 204 of the calculation unit 20. The model data selection unit 204 extracts data used to create an AI model from the measurement value data D301 stored in the measurement value DB301.

[0119] The model data selection unit 204 extracts data from the measurement value data D301 and creates model data D304. The model data selection unit 204 stores the created model data D304 in the model DB 304 of the storage unit 30.

[0120] (Step S06: Creating an estimation AI model) The calculation unit 20 creates an estimation AI model. The operation of step S06 is executed by the estimation AI model creation unit 205 of the calculation unit 20. The AI ​​model creation unit 205 creates a calculation block, calculation parameters, or calculation formula as a calculation model of the AI ​​model. The estimation AI model creation unit 205 creates the calculation parameters or calculation formula for the calculation block F1 or (Equation 4) through machine learning, based on the model data D304 stored in the model DB 304 of the storage unit 30.

[0121] The estimation AI model creation unit 205 stores the created computation model, computation parameters, or computation formula as estimation AI model data D306 in the estimation AI model DB 306 of the storage unit 30.

[0122] (Step S07: Extraction of data used to estimate missing values) The calculation unit 20 extracts data used to estimate the missing values. The operation of step S07 is executed by the observation target determination unit 201 of the calculation unit 20. When the observation target determination unit 201 determines that the power system 90 includes a measurement point with missing measurement values, it extracts data used to estimate the missing values ​​from the measurement value data D301 stored in the measurement value DB301, and creates estimation data D305.

[0123] The observation target determination unit 201 stores the created estimation data D305 in the estimation DB 305 of the storage unit 30.

[0124] (Step S08: Estimation of missing values ​​at measurement points with missing measurements) The calculation unit 20 estimates missing values ​​at measurement points where measurement values ​​are missing. The operation of step S08 is executed by the missing value estimation unit 206 of the calculation unit 20. The missing value estimation unit 206 estimates missing values, which are missing measurement values, based on estimation data D305 stored in the estimation DB 305 of the storage unit 30 and estimation AI model data D306 stored in the estimation AI model DB 306.

[0125] The missing value estimation unit 206 estimates missing values, which are missing measurement values, to generate estimated missing value data D307, and stores the data in the estimated missing value DB 307 in the storage unit 30.

[0126] (Step S09: Selection of data for state estimation) The calculation unit 20 selects data related to state estimation. The operation of step S09 is executed by the interpolation data selection unit 207 of the calculation unit 20. If estimated missing value data D307 is stored in the estimated missing value DB 307 of the storage unit 30, the interpolation data selection unit 207 selects the estimated missing value data D307. If estimated missing value data D307 is not stored in the estimated missing value DB 307 of the storage unit 30, the interpolation data selection unit 207 selects the previous value storage data D303 stored in the previous value storage DB 303.

[0127] (Step S10: Estimation of the state of the power system) The calculation unit 20 estimates the state of the power system 90. The operation of step S10 is executed by a state estimation unit 208 of the calculation unit 20. The state estimation unit 208 estimates the state based on the estimated missing value data D307 or the previous value retention data D303 selected by the interpolation data selection unit 207, and the equipment data D302 stored in the equipment DB302 of the storage unit 30.

[0128] The state estimation unit 208 estimates the state of the power system 90 and transmits state estimation value data D308c or state estimation value data D308b to the output unit 40. The state estimation value data D308c or state estimation value data D308b indicates the state of each node at the current time t0. The state estimation value data D308c and state estimation value data D308b include the estimated measurement value at node d.

[0129] (Step S11: Output of state estimate data) The calculation unit 20 instructs the output unit 40 to output state estimation value data. When the input unit 10 receives a command instructing the output of state estimation value data, the calculation unit 20 causes the output unit 40 to output state estimation value data D308a, state estimation value data D308b, or state estimation value data D308c. The command instructing the output of state estimation value data is input to the input unit 10 by an operator.

[0130] The output unit 40 outputs the state estimation data D308a, D308b, or D308c as a result of the state estimation. The output unit 40 outputs the state estimation data D308a, D308b, or D308c as a display, printout, or electronic data.

[0131] The above is the operation of the power system state estimating device 1 according to this embodiment.

[0132] [1-3.Effects] (1) According to this embodiment, the power system state estimation device 1 includes an observation target determination unit 201 that determines whether the power system 90 includes a measurement point where a measurement value is missing, based on measurements of electrical quantities measured at a plurality of measurement points in the power system 90 that supplies power; an estimation AI model creation unit 205 that, when the observation target determination unit 201 determines that the power system 90 includes a measurement point where a measurement value is missing, creates a computational model that estimates missing values, which are missing measurement values, based on measurements taken at other measurement points related to the measurement point where the measurement value is missing; and a missing value estimation unit 206 that calculates the missing values ​​using the computational model created by the estimation AI model creation unit 205. This makes it possible to provide a power system state estimation device that can accurately grasp the state of the power system.

[0133] The estimation AI model creation unit 205 creates a computational model that estimates the missing measurement value, which is the missing measurement value, based on measurement values ​​measured at other measurement points related to the measurement point where the measurement value is missing, and the missing value estimation unit 206 calculates the missing value using the computational model created by the estimation AI model creation unit 205.Therefore, it is possible to estimate the state of the power system 90 where measurement values ​​are missing due to a failure of the sensor 80, etc., with higher accuracy than state estimation by retaining the previous value.

[0134] (2) According to this embodiment, the estimation AI model creation unit 205 of the power system state estimation device 1 uses machine learning to learn the relationship between the measurement values ​​at a measurement point where measurement values ​​are missing that were measured at the same time in the past and the measurement values ​​at other measurement points related to the measurement point where measurement values ​​are missing, and creates a calculation model that estimates the missing values.Therefore, the state of the power system 90 where measurement values ​​are missing can be estimated with higher accuracy than state estimation by retaining previous values.

[0135] The computational model is created using machine learning after the data is lost, so it can be flexibly created for measurement points with missing measurements, regardless of the specific power system. The computational model is created using the same process regardless of the measurement point with missing measurements.

[0136] Furthermore, since the computational model is created after the data is lost, there is no need to create the computational model in advance or store the data related to the computational model, which saves storage capacity in the storage unit and makes data management easier.

[0137] The missing value estimation unit 206 calculates the missing value using a computation model with input data of current measurements taken at other measurement points related to the measurement point with missing measurements, so that the missing value can be calculated easily and quickly. The calculated missing value is used to complement the missing value at the measurement point with missing measurements.

[0138] (3) According to this embodiment, the power system state estimation device 1 includes a previous value retention data selection unit 203 that, when a missing value is not calculated by the missing value estimation unit 206, extracts a previous value, which is a measurement value at a measurement point where a measurement value is missing, from measurements of electrical quantities measured at multiple measurement points in the power system. The previous value extracted by the previous value retention data selection unit 203 is set as the missing value, so that the estimation AI model creation unit 205 can complement the previous value with the previous value during the time until creating a computation model. This makes it possible to provide a power system state estimation device that can grasp the state of the power system 90 where no measurement values ​​are missing.

[0139] [2. Second Embodiment] [2-1. Composition and Function] The power system state estimation device 1 according to the second embodiment will be described. The configuration of the power system state estimation device 1 according to the second embodiment is the same as the configuration of the power system state estimation device 1 according to the first embodiment. The power system state estimation device 1 according to the second embodiment differs from the power system state estimation device 1 according to the first embodiment in the processing in the estimation AI model creation unit 205. Duplicate explanations of the processing that is the same as that of the power system state estimation device 1 according to the first embodiment will be omitted.

[0140] When a measurement value is missing, the estimation AI model creation unit 205 of the power system state estimation device 1 according to the first embodiment estimates the missing measurement value based on the relationship between the past measurement value at the measurement point where the measurement value is missing and the measurement value at another measurement point measured at the same time.

[0141] When a measurement value is missing, the power system state estimation device 1 according to the second embodiment estimates the missing value based on the relationship between the fluctuation of the past measurement value at the measurement point where the measurement value is missing and the fluctuation of the measurement value at other measurement points. The fluctuation of the measurement value at other measurement points may be a fluctuation related to the past measurement value or a fluctuation based on the past and present measurement values.

[0142] The estimation AI model creation unit 205 uses machine learning to learn the relationship between fluctuations in measurement values ​​measured in the past at a measurement point where measurement values ​​are missing and fluctuations in measurement values ​​measured in the past and present at other measurement points related to the measurement point where measurement values ​​are missing, and creates a computational model that estimates the missing values.

[0143] Fig. 10 shows the relationship between measurement time and measurement values ​​at nodes. As shown in Fig. 10, the measurement values ​​at each node fluctuate over time. When a measurement value is missing, the estimation AI model creation unit 205 of the power system state estimation device 1 according to the second embodiment estimates the missing value, including fluctuations over time.

[0144] The details of the operation of the power system state estimating device 1 according to the second embodiment are as follows.

[0145] When the observation target determination unit 201 determines that the power system 90 includes a measurement point with missing measurement values, the model data selection unit 204 extracts data to be used for creating an AI model from the measurement value data D301 stored in the measurement value DB 301. The model data selection unit 204 extracts data from the measurement value data D301 and creates model data D304.

[0146] The model data selection unit 204 stores the created model data D304 in the model DB 304 of the storage unit 30. The model data D304 may be so-called training data used for calculations by machine learning.

[0147] The model data D304 may be the data shown in Fig. 4, similar to the power system state estimation device 1 according to the first embodiment. Furthermore, the model data D304 may include, in addition to the data shown in Fig. 4, an active power Pc0 and a reactive power Qc0 at node c, and an active power Pe0 and a reactive power Qe0 at node e at the current time t0.

[0148] The model data D304 indicates fluctuations over time at node d, which is a measurement point with missing measurements, and fluctuations over time at nodes c and e adjacent to node d. The model data D304 includes measurement values ​​at node d at multiple past times, and multiple measurement values ​​at nodes c and e adjacent to node d. The multiple measurement values ​​at nodes c and e adjacent to node d may include current measurement values ​​in addition to measurement values ​​at multiple past times.

[0149] The model data D304 is not limited to the measured values ​​of adjacent nodes, and may include data on the measured values ​​of related nodes. For example, if the active power and reactive power at node d are affected by nodes a and b, the model data D304 may include data on the active power and reactive power at nodes a and b.

[0150] Furthermore, the model data D304 is not limited to data relating to measurements taken at times t-1, t-2, and t-3, but may also include data relating to measurements taken before time t-4 or at the current time t0.

[0151] The estimation AI model creation unit 205 creates a computation model using an AI model. The created computation model is used to estimate missing values ​​at measurement points where measurement values ​​are missing in the power system 90. The estimation AI model creation unit 205 creates a computation model for estimating missing values ​​at measurement points where measurement values ​​are missing by machine learning, based on the model data D304 stored in the model DB 304 of the storage unit 30.

[0152] 11 shows an example of input / output data of the estimation AI model creation unit 205. The estimation AI model creation unit 205 stores the created computation model in the estimation AI model DB 306 of the storage unit 30 as estimation AI model data D306.

[0153] The calculation model may be a calculation block, a calculation parameter, or a calculation formula. An example of the estimation AI model data D306 is shown in Fig. 12. Fig. 12 shows a case where the calculation model related to the estimation AI model data D306 is a calculation block based on an AI model.

[0154] The estimation AI model creation unit 205 creates a calculation block F2 of the AI ​​model through machine learning using the model data D304 as training data. As an example, if the active power Pd0 and reactive power Qd0 at node d at the current time t0 are missing, the calculation block F2 is created as follows:

[0155] The estimation AI model creation unit 205 learns the relationship between the fluctuations in the active power Pc and reactive power Qc of node c adjacent to node d, the active power Pe and reactive power Qc of node e, and the fluctuations in the active power Pd and reactive power Qd of node d from time t-3 to time t-1.

[0156] The estimation AI model creation unit 205 may also learn the relationship between the active power Pc-1 and reactive power Qc-1 of node c, the active power Pe-1 and reactive power Qc-1 of node e, and the active power Pd-1 and reactive power Qd-1 of node d at time t-1. The estimation AI model creation unit 205 may also learn the relationship between the active power Pc-2 and reactive power Qc-2 of node c, the active power Pe-2 and reactive power Qc-2 of node e, and the active power Pd-2 and reactive power Qd-2 of node d at time t-2. Furthermore, the estimation AI model creation unit 205 may also learn the relationship between the active power Pc-3 and reactive power Qc-3 of node c, the active power Pe-3 and reactive power Qc-3 of node e, and the active power Pd-3 and reactive power Qd-3 of node d at time t-3.

[0157] The estimation AI model creation unit 205 creates a calculation block F2 based on the learned relationships between the active power and reactive power of node c, the active power and reactive power of node e, and the active power and reactive power of node d. The calculation block F2 is a functional block that receives the active power Pd and reactive power Qd of node d, the active power Pc and reactive power Qc of node c, and the active power Pe and reactive power Qe of node e as inputs, and outputs the active power Pd0X and reactive power Qd0X of node d.

[0158] The active power Pd and reactive power Qd of node d may be measured values ​​of node d at multiple times in the past. The active power Pc and reactive power Qc of node c and the active power Pe and reactive power Qe of node e may include current measurements in addition to measurements at multiple times in the past.

[0159] The estimation AI model creation unit 205 stores the created operation block F2 in the estimation AI model DB 306 of the storage unit 30 as estimation AI model data D306.

[0160] The estimation AI model creation unit 205 may use the model data D304 as training data to create calculation parameters or calculation formulas for the AI ​​model as a calculation model through machine learning. As an example, if the active power Pd0 and reactive power Qd0 at node d at current time t0 are missing, the calculation parameters for the AI ​​model are created as follows:

[0161] The estimation AI model creation unit 205 creates the following (Equation 5) based on the model data D304.

number

[0162] In equation 5, Pc / dt represents the fluctuation of the active power Pc at node c, and Qc / dt represents the fluctuation of the reactive power Qc at node c. Pc / dt is based on the fluctuation of the active power Pc at node c from time t-3 to time t-1, and Qc / dt is based on the fluctuation of the reactive power Qc at node c from time t-3 to time t-1.

[0163] In equation (5), Pe / dt represents the fluctuation of the active power Pe at node e, and Qe / dt represents the fluctuation of the reactive power Qe at node e. Pe / dt is based on the fluctuation of the active power Pe at node e from time t-3 to time t-1, and Qe / dt is based on the fluctuation of the reactive power Qe at node e from time t-3 to time t-1.

[0164] In equation 5, Pd / dt represents the fluctuation of active power Pd at node d, and Qd / dt represents the fluctuation of reactive power Qd at node d. Pd / dt is based on the fluctuation of active power Pd at node d from time t-3 to time t-1, and Qd / dt is based on the fluctuation of reactive power Qd at node d from time t-3 to time t-1.

[0165] In Equation 5, L1 and L2 are matrices composed of coefficients or functions. The estimation AI model creation unit 205 calculates L1 and L2 that satisfy Equation 5. The estimation AI model creation unit 205 stores the calculated calculation parameters L1 and L2 in the estimation AI model DB 306 of the storage unit 30 as estimation AI model data D306.

[0166] The estimation AI model creation unit 205 may calculate L1 and L2, and store the created (Equation 5) in the estimation AI model DB 306 of the storage unit 30 as an arithmetic equation related to the estimation AI model data D306.

[0167] When the observation target determination unit 201 determines that the power system 90 includes a measurement point with missing measurement values, it extracts data from the measurement value data D301 and creates estimation data D305. The observation target determination unit 201 stores the created estimation data D305 in the estimation DB 305 of the storage unit 30.

[0168] The missing value estimation unit 206 estimates missing values, which are missing measurement values, based on estimation data D305 stored in the estimation DB 305 of the memory unit 30 and estimation AI model data D306 stored in the estimation AI model DB 306.

[0169] The missing value estimation unit 206 estimates the missing values ​​of active power Pd0 and reactive power Qd0 using the above-mentioned calculation block F1, calculation parameters K1 and K2, or calculation formula (Formula 5), ​​and calculates them as active power Pd0X and reactive power Qd0X, respectively.

[0170] The missing value estimation unit 206 generates estimated missing value data D307 based on the calculated active power Pd0X and reactive power Qd0X. The missing value estimation unit 206 stores the generated estimated missing value data D307 in the estimated missing value DB 307 of the storage unit 30.

[0171] If estimated missing value data D307 is stored in the estimated missing value DB307 of the storage unit 30, the interpolation data selection unit 207 selects the estimated missing value data D307, and if estimated missing value data D307 is not stored in the estimated missing value DB307, the interpolation data selection unit 207 selects the previous value retention data D303 stored in the previous value retention DB303.

[0172] The state estimation unit 208 performs state estimation based on the estimated missing value data D307 or previous value retention data D303 selected by the interpolation data selection unit 207, and the equipment data D302 stored in the equipment DB 302 of the storage unit 30.

[0173] The state estimation unit 208 creates state estimated value data D308c based on the estimated missing value data D307 and the equipment data D302. The state estimation unit 208 creates state estimated value data D308b based on the previous value retention data D303 and the equipment data D302. The state estimation unit 208 transmits the state estimated value data D308c or the state estimated value data D308b to the output unit 40.

[0174] The output unit 40 outputs the state estimation value data D308a, the state estimation unit 208 state estimation value data D308b, or the state estimation value data D308c as a result of the state estimation.

[0175] The power system state estimating device 1 according to the second embodiment performs the following process in (Step S06: Creation of estimation AI model) of the computer program shown in FIG.

[0176] (Step S06: Creating an estimation AI model) The calculation unit 20 creates an estimation AI model. The operation of step S06 is executed by the estimation AI model creation unit 205 of the calculation unit 20. The AI ​​model creation unit 205 creates a calculation model as an AI model. The estimation AI model creation unit 205 creates calculation parameters or a calculation formula for the calculation block F2 or (Equation 5) by machine learning, based on the model data D304 stored in the model DB 304 of the storage unit 30.

[0177] The above is the operation of the power system state estimating device 1 according to this embodiment.

[0178] [2-2. Effects] (1) According to this embodiment, the estimation AI model creation unit 205 of the power system state estimation device 1 uses machine learning to learn the relationship between fluctuations in measurement values ​​measured in the past at a measurement point where measurement values ​​are missing and fluctuations in measurement values ​​measured in the past and present at other measurement points related to the measurement point where measurement values ​​are missing, and creates a calculation model that estimates the missing values, thereby making it possible to grasp the state of the power system 90 with greater accuracy.

[0179] The computational model is created using machine learning after the data is lost, so it can be flexibly created for measurement points with missing measurements, regardless of the specific power system. The computational model is created using the same process regardless of the measurement point with missing measurements.

[0180] Furthermore, since the computational model is created after the data is lost, there is no need to create the computational model in advance or store the data related to the computational model, which saves storage capacity in the storage unit and makes data management easier.

[0181] 3. Other Embodiments Although embodiments including modifications have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their modifications are included within the scope and spirit of the invention, as well as within the scope of the invention and its equivalents as set forth in the claims. The following is an example.

[0182] (1) Each unit in the above embodiment may be realized by steps relating to a computer program for a power outage damage calculation device or procedures relating to a power outage damage calculation method. The steps executed by the observation target determination unit 201 may be realized as observation target determination steps, and the procedures executed may be realized as observation target determination procedures. The steps executed by the estimation AI model creation unit 205 may be realized as estimation AI model creation steps, and the procedures executed may be realized as estimation AI model creation procedures. The steps executed by the missing value estimation unit 206 may be realized as missing value estimation steps, and the procedures executed may be realized as missing value estimation procedures.

[0183] (2) In the above embodiment, the storage unit 30 is provided inside the power system state estimation device 1. However, the storage unit 30 may be provided outside the power system state estimation device 1. For example, each database of the storage unit 30 may be provided on a server or cloud outside the power system state estimation device 1. Furthermore, each database of the storage unit 30 may be individually configured by a plurality of storage devices. Data stored in the storage unit 30 may be stored by individual devices. [Explanation of symbols]

[0184] 1. Power system state estimator 10 Input section 20... Arithmetic section 201... Observation target determination unit 202... Actual system state calculation unit 203: Previous value retention data selection unit 21. AI Model Creation Department 204...Model data selection section 205···Estimation AI Model Creation Department 22. AI Estimated Value Creation Unit 206···Missing Value Estimation Unit 207 Complementary data selection section 208...State estimation unit 30...Storage section 301 Measurement value DB 302...Equipment DB 303...Previous price retention DB 304···Model DB 305...Estimation DB 306···Estimation AI Model DB 307···Estimated Missing Values ​​DB 40 Output section 70. Communication line 80...Sensor 90...Power system

Claims

1. an observation target determination unit that determines whether the power system includes a measurement point where a measurement value is missing, based on measurement values ​​related to electrical quantities measured at a plurality of measurement points in the power system that supplies power; an estimation AI model creation unit that, when the observation target determination unit determines that the power system includes the measurement point with missing measurement values, creates a calculation model that estimates the missing measurement values ​​based on measurement values ​​measured at other measurement points related to the measurement point with missing measurement values; and a missing value estimation unit that calculates the missing value using the calculation model created by the estimation AI model creation unit, Power system state estimator.

2. the estimation AI model creation unit learns, by machine learning, the relationship between the measurement value at the measurement point where the measurement value is missing, which was measured at the same time in the past, and the measurement value at the other measurement point related to the measurement point where the measurement value is missing, and creates a calculation model that estimates the missing value. The power system state estimating device according to claim 1 .

3. the estimation AI model creation unit learns, by machine learning, the relationship between fluctuations in measurement values ​​measured in the past at the measurement point where measurement values ​​are missing and fluctuations in measurement values ​​measured in the past and present at the other measurement points related to the measurement point where measurement values ​​are missing, and creates a calculation model that estimates the missing values. The power system state estimating device according to claim 1 .

4. When the missing value is not calculated by the missing value estimation unit, a previous value retention data selection unit that extracts a previous value, which is a measurement value immediately before a measurement value at a measurement point where a measurement value is missing, from the measurement values ​​related to electrical quantities measured at a plurality of measurement points in the power system, and the previous value extracted by the previous value retention data selection unit is set as the missing value; The power system state estimating device according to claim 1 .

5. On the computer, an observation target determination step of determining whether the power system includes a measurement point where a measurement value is missing based on measurement values ​​of electrical quantities measured at a plurality of measurement points in the power system that supplies power; an estimation AI model creation step of creating a calculation model that estimates the missing measurement value, based on measurement values ​​measured at other measurement points related to the measurement point where the measurement value is missing, when the observation target determination step determines that the power system includes the measurement point where the measurement value is missing; a missing value estimation step unit that calculates the missing value using the calculation model created in the estimation AI model creation step; A computer program for a power system state estimation device that executes the above.

6. an observation target determination step of determining whether the power system includes a measurement point where a measurement value is missing based on measurements of electrical quantities measured at a plurality of measurement points in the power system that supplies power; an estimation AI model creation step of creating a computational model that estimates the missing measurement value based on measurements taken at other measurement points related to the measurement point where the measurement value is missing, when the observation target determination step determines that the power system includes the measurement point where the measurement value is missing; a missing value estimation procedure for calculating the missing values ​​using the computational model created by the estimation AI model creation procedure, Power system state estimation method.

Images