Estimation device

Abstract

To provide an estimation device capable of precisely estimating the maximum oil temperature of a transformer.SOLUTION: An estimation device includes: an acquisition section for acquiring a measurement value of the maximum oil temperature of a transformer; a generation section for generating a relational expression showing a relationship between the current maximum oil temperature and the past maximum oil temperature; and a temperature estimation section for estimating the maximum oil temperature in time series by applying a Kalman filter to the relational expression. The temperature estimation section calculates a prediction value of the maximum oil temperature on the basis of the last estimation value of the maximum oil temperature, and calculates a current estimation value of the maximum oil temperature by updating the prediction value by using the measurement value as an observation value.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] The present disclosure relates to an estimation device. [Background technology]

[0002] Transformers installed in power company substations are important devices that supply power, so their condition needs to be monitored by measuring the voltage, current, oil temperature, etc. of the transformer.

[0003] For example, Japanese Patent Laid-Open Publication No. 2005-302794 (Patent Document 1) discloses an oil-filled transformer monitoring system. The oil-filled transformer monitoring system includes an oil temperature conversion means for converting the oil-filled transformer current measurement means and the outside air temperature measurement means into the oil-filled transformer oil temperature, and an oil temperature measurement means. The oil-filled transformer monitoring system compares the output value of the oil temperature calculation means with the output value of the oil temperature measurement means to detect an abnormality. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2005-302794 A Summary of the Invention [Problem to be solved by the invention]

[0005] The IEC 60076-7 standard specifies an equation that indicates the relationship between the ambient temperature and the maximum oil temperature of an oil-filled transformer. In this equation, the oil time constant of the oil-filled transformer, parameters according to the cooling method, etc. are given as constants, and a method is known in which the maximum oil temperature is calculated periodically using this equation, and an abnormality is detected when the calculated value reaches a threshold value. However, the constant used in this method is set based on the results of factory tests of the oil-filled transformer, etc. Therefore, it is unclear whether the constant is a valid value even in the operating environment after the oil-filled transformer is installed. If the constant is not a valid value, the maximum oil temperature may not be estimated accurately.

[0006] An object of one aspect of the present disclosure is to provide an estimation device capable of accurately estimating a maximum oil temperature of a transformer. [Means for solving the problem]

[0007] An estimation device according to an embodiment includes an acquisition unit that acquires a measured value of a maximum oil temperature of a transformer, a generation unit that generates a relational equation indicating a relationship between a current maximum oil temperature and a past maximum oil temperature, and a temperature estimation unit that estimates the maximum oil temperature in a time series by applying a Kalman filter to the relational equation. The temperature estimation unit calculates a predicted value of the maximum oil temperature based on an estimate of the previous maximum oil temperature, and calculates an estimate of the current maximum oil temperature by updating the predicted value using the measured value as an observation value. Effect of the Invention

[0008] According to the present disclosure, the maximum oil temperature of a transformer can be estimated with high accuracy. [Brief description of the drawings]

[0009] [Figure 1] FIG. 1 illustrates an example of an overall configuration of an estimation system according to a first embodiment. [Diagram 2] FIG. 2 illustrates an example of a hardware configuration of the estimation device according to the first embodiment. [Diagram 3] 1 is a block diagram showing an example of a functional configuration of an estimation device according to a first embodiment. FIG. [Figure 4] 5 is a flowchart showing an example of a processing procedure of the estimation device according to the first embodiment. [Diagram 5] FIG. 11 is a configuration diagram of a learning device according to a second embodiment. [Figure 6] 11 is a flowchart illustrating a learning process of the learning device according to the second embodiment. [Figure 7] FIG. 11 is a configuration diagram of a remaining life estimation unit according to the second embodiment. [Figure 8] 13 is a flowchart showing a process for estimating remaining life using a trained model. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] Hereinafter, the present embodiment will be described with reference to the drawings. In the following description, the same components are denoted by the same reference numerals. Their names and functions are also the same. Therefore, detailed description thereof will not be repeated.

[0011] Embodiment 1 <System configuration> The lifespan of an oil-immersed transformer is shortened by the heat generated by the windings inside the transformer due to temperature rise caused by overload, which causes deterioration. This shortening of lifespan is described, for example, in the IEC 60076-7 standard. According to this standard, the remaining lifespan of a transformer is calculated from the maximum winding temperature, which is calculated from the maximum oil temperature. Here, the maximum oil temperature is the temperature at the top of the insulating oil in which the windings of an oil-immersed transformer are immersed.

[0012] Therefore, by accurately estimating the maximum oil temperature, it is possible to accurately estimate the maximum winding temperature and the remaining life of the transformer. In the first embodiment, a method for estimating the maximum oil temperature and a method for calculating the maximum winding temperature and the remaining life of the transformer using the estimated maximum oil temperature will be specifically described.

[0013] Fig. 1 is a diagram showing an example of an overall configuration of an estimation system 100 according to the first embodiment. With reference to Fig. 1, the estimation system 100 is a system for estimating a maximum oil temperature of a transformer 20. Specifically, the estimation system 100 includes an estimation device 10, a transformer 20, and a monitoring device 40.

[0014] The transformer 20 is an oil-filled transformer having an iron core, a winding, and a tank. The tank contains the iron core and the winding together with insulating oil.

[0015] The monitoring device 40 is a device that measures and monitors the maximum oil temperature of the insulating oil in the transformer 20 and the ambient temperature of the transformer 20. Specifically, the monitoring device 40 receives input of the maximum oil temperature measured by a resistance temperature detector (e.g., a platinum sound detector) provided in the transformer 20 and the ambient temperature measured by a temperature sensor provided around the transformer 20. The monitoring device 40 is configured to be able to communicate with the estimation device 10. Note that the monitoring device 40 may be communicatively connected to the estimation device 10 by a cloud that provides information via an Internet network, or the like.

[0016] The estimation device 10 is a device for estimating the maximum oil temperature of the transformer 20. The estimation device 10 collects the current measured values ​​of the maximum oil temperature and the ambient temperature of the transformer 20 from the monitoring device 40 at regular intervals (e.g., every minute). Therefore, the internal memory of the estimation device 10 accumulates the measured values ​​of the maximum oil temperature and the ambient temperature at regular intervals. In addition, the internal memory stores various parameters required for estimating the maximum oil temperature in advance.

[0017] For example, the various parameters include an oil temperature rise value indicating the difference between the maximum oil temperature in the transformer tank and the ambient temperature at rated loss, an oil time constant, a ratio of load loss to no-load loss, a load factor, a constant determined by the cooling method, a constant determined by the transformer 20, and the like.

[0018] The estimation device 10 uses the above-mentioned measured values, the above-mentioned parameters, and a Kalman filter to calculate an estimate of the maximum oil temperature of the transformer 20, and outputs the estimate. For example, the estimate of the maximum oil temperature may be displayed on a display of the estimation device 10, or may be transmitted to an external device.

[0019] Furthermore, the estimation device 10 estimates the remaining life of the transformer 20 based on the estimated value of the maximum oil temperature, and outputs the estimated value of the remaining life. The estimated value of the remaining life of the transformer 20 may be displayed on a display or may be transmitted to an external device. Note that the estimation device 10 is, for example, a data collection server provided on a cloud, but may be another device.

[0020] <Hardware configuration> Fig. 2 is a diagram illustrating an example of a hardware configuration of the estimation device 10 according to the first embodiment. With reference to Fig. 2, the estimation device 10 includes a processor 101, a memory 103, a display 105, an input device 107, an input / output interface (I / F) 109, and a communication interface (I / F) 111. These are connected to each other so as to be able to communicate data with each other.

[0021] The processor 101 is typically an arithmetic processing unit such as a CPU (Central Processing Unit) or an MPU (Multi Processing Unit). The processor 101 reads out and executes a program stored in the memory 103 to control the operation of each part of the estimation device 10. More specifically, the processor 101 realizes each function of the estimation device 10 by executing the program.

[0022] The memory 103 is realized by a RAM (Random Access Memory), a ROM (Read-Only Memory), a flash memory, a hard disk, or the like. The memory 103 stores a program executed by the processor 101, the maximum oil temperature and the ambient temperature of the transformer 20 received from the monitoring device 40, and the like. Note that the estimation device 10 may have a storage device accessible from the estimation device 10 in addition to the memory 103. In this case, the storage device may store the above data stored in the memory 103 or other data.

[0023] The display 105 is, for example, a liquid crystal display, an organic EL (Electro Luminescence) display, etc. The display 105 may be configured integrally with the estimation device 10, or may be configured separately from the estimation device 10.

[0024] The input device 107 accepts operation inputs for the estimating device 10. The input device 107 is realized by, for example, a keyboard, a button, a mouse, etc. The input / output interface 109 mediates data transmission between the processor 101 and an external device.

[0025] The communication interface 111 mediates data transmission from the monitoring device 40 to the estimation device 10. Communication via the communication interface 111 is performed using the Internet, a wide area network (WAN), a local area network (LAN), an Ethernet network, a wired network, a wireless network, or the like, or a combination of these.

[0026] <Functional configuration> Fig. 3 is a block diagram showing an example of a functional configuration of estimation device 10 according to the first embodiment. With reference to Fig. 3, estimation device 10 includes an acquisition unit 220, a generation unit 240, a temperature estimation unit 260, and a remaining life estimation unit 280. These functions are realized, for example, by processor 101 of estimation device 10 executing a program stored in memory 103. Note that some or all of these functions may be configured to be realized by hardware.

[0027] Hereinafter, the above-mentioned functions of the estimation device 10, as well as a method for estimating the maximum oil temperature, a method for calculating the maximum winding temperature, and a method for calculating the remaining life of the transformer 20 will be described.

[0028] The differential equation expressing the temperature rise of the oil temperature of the transformer 20 is described in the standard IEC 60076-7, and is expressed by the following formula (1). θo is the maximum oil temperature, θa is the ambient temperature, K is the load factor, R is the ratio of the load loss to the no-load loss, Δθor is the difference between the maximum oil temperature in the transformer tank and the ambient temperature at the time of rated loss (hereinafter also referred to as the "oil temperature rise value"), k11 is a constant determined by the transformer 20, and τo is the oil time constant. Note that in the following formula (1), Δθ is used to simplify the notation. or For example, "Δθ in a formula" or" is the same as "Δθor" in the text. This also applies to equations (2) and (10) to (18) described later.

[0029]

number

[0030] The differential equation shown in Equation (1) is a differential equation for the maximum oil temperature θo using a number of predetermined parameters and a measured value of the ambient temperature (i.e., θa). The parameters include the constants load factor K, ratio R, oil temperature rise value Δθor, constant k11, and oil time constant τo.

[0031] The generating unit 240 converts the differential equation shown in formula (1) and generates a difference equation obtained by the conversion. The difference equation is a relational equation that represents the relationship between the current maximum oil temperature and the past maximum oil temperature. Specifically, the generating unit 240 generates a linear difference equation shown in formula (2) based on formula (1), which is a differential equation.

[0032]

number

[0033] In equation (2), the current (e.g., this time) maximum oil temperature is θo(t+1) on the left side, and the past (e.g., last time) maximum oil temperature is θo(t) on the right side. The maximum oil temperature θo(t+1) at time (t+1) represents the maximum oil temperature one period ahead of the maximum oil temperature θo(t) at time t.

[0034] The acquisition unit 220 acquires (receives) the measurement value of the maximum oil temperature of the transformer 20 and the measurement value of the ambient temperature of the transformer 20 from the monitoring device 40 at regular intervals (e.g., every minute). As a result, the measurement values ​​of the maximum oil temperature and the ambient temperature are stored in the memory 103, for example, every minute.

[0035] The temperature estimation unit 260 estimates the maximum oil temperature in a time series manner by applying a Kalman filter to the difference equation shown in the above formula (2). Hereinafter, a method for estimating the maximum oil temperature using a Kalman filter will be described.

[0036] The Kalman filter predicts the state of the current term based on the state equation and the observation equation, and then corrects the predicted value using the observed value of the current term to obtain an accurate estimate of the state of the current term. In addition, by repeatedly obtaining a predicted value for the next term using the estimated value of the current term, it is possible to obtain an estimated value in a time series with high accuracy.

[0037] Specifically, a Kalman filter is applied to a state space model based on the difference equation (i.e., state equation) of equation (2). The state space model at a discrete time t (hereinafter also simply referred to as "time t") is expressed in the form of equations (3) and (4).

[0038]

number

[0039]

number

[0040] Equation (3) is x t are state variables, and u t This is a state equation with the state variable x as an input variable. t corresponds to the maximum oil temperature θo(t) in equation (2), and the state variable x t+1 corresponds to the maximum oil temperature θo(t+1) in equation (2). t corresponds to the ambient temperature θa(t) in equation (2). t The observation equation is with the observed variable y t corresponds to the measured value of the maximum oil temperature θo(t) at time t.

[0041] The coefficients A and B in equation (3) and the coefficient C in equation (4) are constants that do not depend on time. For example, the coefficient A corresponds to the coefficient of the maximum oil temperature θo(t) in equation (2), and the coefficient B corresponds to the coefficient of the ambient temperature θa(t) in equation (2). t is the system noise that follows a Gaussian distribution (N(0,Q)) with mean 0 and covariance matrix Q. t is the observation noise that follows a Gaussian distribution (N(0,S)) with mean 0 and covariance matrix S.

[0042] There are two steps in the process of applying the Kalman filter based on the state space model. The first step is the one-period-ahead prediction process, which calculates the predicted value of the current state variable based on the estimated value of the state variable one period ago.

[0043] The temperature estimation unit 260 calculates a predicted value of the current maximum oil temperature based on the estimated value of the previous maximum oil temperature. Specifically, the temperature estimation unit 260 calculates the predicted value using the following formula (5).

[0044]

number

[0045] In equation (5), the subscript on the left side of the subscript of the state variable x indicates the time of the state variable, and the subscript on the right side indicates the time of the observation variable. Therefore, x on the left side t|t-1 is the observed variable y at time t t The state variables at time t when the state variable x t|t-1 indicates the predicted value of the state variable at time t at time (t-1). Also, x on the right hand side t-1|t-1 is the observed variable y at time (t-1) t-1 The state variables at time (t-1) when the state is obtained are shown. That is, the state variables x t-1|t-1 denotes the final estimate of the state variables at time (t-1).

[0046] From the above, the state variable x t-1|t-1corresponds to the estimated maximum oil temperature at the previous time (e.g., time (t-1)), and the state variable x t|t-1 corresponds to the predicted maximum oil temperature at this time (for example, at time t).

[0047] Furthermore, the temperature estimation unit 260 calculates the covariance matrix of the predicted value of the current maximum oil temperature based on the covariance matrix of the estimated value of the previous maximum oil temperature. Specifically, the temperature estimation unit 260 calculates the covariance matrix of the predicted value using the following equation (6).

[0048]

number

[0049] In equation (6), V on the left side t|t-1 is the state variable x t|t-1 Denotes the covariance matrix of V t|t-1 indicates the covariance matrix of the predicted value of the state variable at time t at time (t-1). Also, V on the right-hand side t-1|t-1 is the state variable x t-1|t-1 Denotes the covariance matrix of V t-1|t-1 denotes the covariance matrix of the final estimate of the state variables at time (t-1). T is the transpose of A. From the above, the covariance matrix V t-1|t-1 corresponds to the covariance matrix of the estimated maximum oil temperature at the previous time (e.g., time (t-1)), and the covariance matrix V t|t-1 corresponds to the covariance matrix of the predicted value of the maximum oil temperature at this time (e.g., time t).

[0050] Next, we will explain the second step, the update process. The update process is a process in which the predicted value of the maximum oil temperature for the current period, which was predicted in the first step, is corrected and updated using the observed value of the maximum oil temperature for the current period.

[0051] The temperature estimation unit 260 estimates the covariance matrix V t|t-1and the covariance matrix S of the observation noise. Specifically, the temperature estimation unit 260 calculates the Kalman gain using the following equation (7).

[0052]

number

[0053] In equation (7), J on the left side t is the Kalman gain at time t. The Kalman gain is a parameter that determines the extent to which the correction is reflected when the state variables are corrected based on the observation variables. As described above, the covariance matrix V calculated by equation (6) t|t-1 By applying this to equation (7), the Kalman gain J t is calculated.

[0054] Next, the temperature estimation unit 260 calculates an estimate of the current maximum oil temperature by updating the predicted value of the current maximum oil temperature based on the observed value (i.e., the measured value) of the current maximum oil temperature. Specifically, the temperature estimation unit 260 calculates the estimate of the current maximum oil temperature using equation (8).

[0055]

number

[0056] In equation (8), x on the left side t|t indicates the final estimated value of the state variable at time t (i.e., the maximum oil temperature this time). From equation (8), the observed variable y t and the predicted value (i.e., the state variable x t|t-1 ) is the Kalman gain J t It can be seen that the predicted value is updated using the value multiplied by . In this manner, the current measured value of the maximum oil temperature is used as the observed value to update the predicted value of the current maximum oil temperature, thereby calculating an estimated value of the current maximum oil temperature.

[0057] The temperature estimation unit 260 calculates the covariance matrix of the estimated value of the current maximum oil temperature by updating the covariance matrix of the predicted value of the current maximum oil temperature based on the Kalman gain. Specifically, the temperature estimation unit 260 calculates the covariance matrix of the estimated value of the current maximum oil temperature using equation (9).

[0058]

number

[0059] In equation (9), V on the left side t|t indicates the covariance matrix of the final estimated value of the state variables at time t. From equation (9), the covariance matrix V t|t-1 Kalman gain J t Using the value multiplied by, the covariance matrix V t|t-1 It can be seen that the covariance matrix of the current predicted value of the maximum oil temperature is updated based on the Kalman gain in this manner, whereby the covariance matrix of the current estimated value of the maximum oil temperature is calculated.

[0060] The temperature estimation unit 260 repeatedly executes the above-mentioned one-period-ahead prediction process and update process at each time t to estimate the state variable x t|t (i.e., the estimated maximum oil temperature) and its covariance matrix V t|t (i.e., the covariance matrix of the estimated value of the maximum oil temperature) is calculated in sequence. This provides time series data of the estimated value of the maximum oil temperature (for example, θ(0) to θ(n)) and time series data of its covariance matrix.

[0061] Next, a method for calculating the winding maximum point temperature and a method for calculating the remaining life of the transformer 20 will be described. The remaining life estimation unit 280 estimates the remaining life of the transformer 20 based on the estimated value of the maximum oil temperature estimated by the temperature estimation unit 260. Specifically, the remaining life estimation unit 280 calculates the estimated value of the winding maximum point temperature using the estimated value of the maximum oil temperature and equations (10), (11), and (12).

[0062]

number

[0063]

number

[0064]

number

[0065] In equations (10) and (11), k21 and k22 are constants determined by the transformer, e is the winding index, τw is the time constant of the winding temperature, and Δθhr is the difference between the maximum winding temperature in the transformer tank and the maximum oil temperature at the rated current. Δθh1 in equation (10) indicates the basic temperature rise component (hereinafter also referred to as the "rise component") excluding the effect of the change in oil flow rate due to natural convection passing through the winding, and Δθh2 in equation (11) indicates the temperature rise suppression (mitigation) component (hereinafter also referred to as the "suppression component") corresponding to the change in oil flow due to natural convection. θh in equation (12) indicates the maximum winding temperature. Note that the suppression component Δθh2 has a negative sign in equation (12) because it has the effect of suppressing heating by the winding.

[0066] The remaining life estimation unit 280 calculates the increase component Δθh1 and suppression component Δθh2 at time t using equations (10) and (11). The remaining life estimation unit 280 calculates the estimate of the winding maximum temperature at time t (i.e., θh(t)) by applying the increase component Δθh1, the suppression component Δθh2, and the estimate of the maximum oil temperature at time t (i.e., θo(t)) to equation (12).

[0067] Next, the remaining life estimation unit 280 calculates the loss of life of the transformer 20 based on the calculated estimated value of the maximum winding temperature, and estimates the remaining life of the transformer 20 based on the loss of life. Specifically, the remaining life estimation unit 280 calculates the loss of life LOL in the period from time t1 to t2 using equation (13).

[0068]

number

[0069] Specifically, the remaining life estimation unit 280 calculates the loss of life LOL by substituting the estimated value of the winding maximum temperature into equation (13) and integrating from time t1 (e.g., time t) to time t2 (e.g., design life t0).The remaining life estimation unit 280 then calculates the difference between the period length (t0-t) and the LOL found by equation (13) as the estimated value of the remaining life at time t.

[0070] <Processing Procedure> Fig. 4 is a flowchart showing an example of a processing procedure of the estimation device 10 according to the embodiment 1. Referring to Fig. 4, the processor 101 acquires the measurement values ​​of the maximum oil temperature and the ambient temperature of the transformer 20 from the monitoring device 40 via the communication interface 111 (step S10).

[0071] The processor 101 applies a Kalman filter to the above-mentioned formula (2) to calculate a predicted value of the maximum oil temperature, and then updates the predicted value to calculate an estimated value of the maximum oil temperature (step S12). Specifically, the processor 101 calculates the estimated value by executing the process by the above-mentioned temperature estimation unit 260.

[0072] The processor 101 estimates the winding maximum temperature based on the estimated value of the maximum oil temperature (step S14). Specifically, the processor 101 calculates the estimated value of the winding maximum temperature using the estimated value of the maximum oil temperature and the above-mentioned equations (10) to (12).

[0073] The processor 101 estimates the remaining life of the transformer 20 based on the estimated value of the winding maximum point temperature (step S16). Specifically, the processor 101 calculates the estimated value of the remaining life of the transformer 20 by executing the process by the remaining life estimation unit 280 described above.

[0074] According to the above-mentioned embodiment 1, even if the constants in the formula (1) set in advance are not valid (for example, they change after being set), the maximum oil temperature can be accurately estimated using the Kalman filter. As a result, the maximum winding temperature and the remaining life of the transformer can also be accurately estimated.

[0075] <Variation 1> In the above, a configuration has been described in which the generating unit 240 generates a differential equation to be applied as a state equation in the Kalman filter by converting the differential equation of formula (1). In Modification 1, a configuration will be described in which the generating unit 240 generates a differential equation by using time-series data of the measured value of the maximum oil temperature.

[0076] The generation unit 240 according to the first modification obtains, for each of a plurality of timings, a data point including the measured value of the maximum oil temperature measured at that timing and the measured value of the maximum oil temperature measured at the timing next to that timing, and generates a differential equation by linear regression of the series of data points. First, a differential equation showing the relationship between the previous maximum oil temperature and the current maximum oil temperature is defined as shown in Equation (14) using coefficients α and β.

[0077]

number

[0078] The maximum oil temperature measured at a certain time (e.g., time t) is θo(t) on the right-hand side. The maximum oil temperature measured at the next time (e.g., time (t+1)) is θo(t+1) on the left-hand side.

[0079] The generation unit 240 obtains the coefficients α and β in formula (14) using linear regression. Specifically, the generation unit 240 obtains (N-1) data points, (θo(0), θo(1)), (θo(1), θo(2)), ..., (θo(N-2), θo(N-1)), using the measured values ​​of the maximum oil temperature at N points, θo(0), θo(1), ..., θo(N-1), from the past to the present.

[0080] The generation unit 240 plots the above (N-2) data points on a graph with θ(t) as the first axis and θ(t+1) as the second axis, and calculates the coefficients α and β of equation (14) by linear regression of the series of plotted data points. Specifically, the generation unit 240 calculates a regression line that linearly regresses the series of data points, and sets the coefficients of the regression line as coefficients α and β. The regression line is calculated by the least squares method or the like. The temperature estimation unit 260 estimates the maximum oil temperature in time series by applying a Kalman filter to equation (14) using the calculated coefficients α and β.

[0081] According to the first modification, the coefficients α and β in the formula (14) are appropriately calculated. Therefore, by applying a Kalman filter to the formula (14), the maximum oil temperature can be estimated with high accuracy.

[0082] <Variation 2> The above describes a configuration in which remaining life estimation unit 280 calculates an estimated value of the winding maximum point temperature using an estimated value of the maximum oil temperature and equations (10) to (12), and estimates the remaining life using the estimated value of the winding maximum point temperature and equation (13).

[0083] In the second modification, a configuration will be described in which an estimated value of the winding maximum temperature is calculated using a Kalman filter. However, in the second modification, it is assumed that the winding maximum temperature can be measured, and the measured value of the winding maximum temperature is used as an observation value in the Kalman filter.

[0084] The generation unit 240 according to the second modification generates the differential equation shown in equation (15) by transforming the differential equation (10), and generates the differential equation shown in equation (16) by transforming the differential equation (11).

[0085]

number

[0086]

number

[0087] In equation (15), the current (i.e., this time) rising component is Δθh1(t+1) on the left side, and the past (i.e., previous) rising component is Δθh1(t) on the right side. The rising component Δθh1(t+1) at time (t+1) represents the rising component one period ahead of the rising component Δθh1(t) at time t. In equation (16), the suppression component Δθh2(t+1) at time (t+1) represents the suppression component one period ahead of the suppression component Δθh2(t) at time t.

[0088] The generation unit 240 generates the determinant shown in equation (17) using equations (15) and (16).

[0089]

number

[0090] In equation (17), constant H1 corresponds to the coefficient of Δθh1(t) in equation (15), and constant I1 corresponds to the second term on the right side of equation (15). Constant H2 corresponds to the coefficient of Δθh2(t) in equation (16), and constant I2 corresponds to the second term on the right side of equation (16). Thus, equation (17) shows the relationship between the current increase component Δθh1 and the previous increase component Δθh1, and the relationship between the current suppression component Δθh2 and the previous suppression component Δθh2.

[0091] Moreover, the above equation (12) can be rewritten as the following equation (18).

[0092]

number

[0093] State variables (Δθh1, Δθh2) T By defining the vector of and the observation variable as θh, equation (17) is substituted into equation (3) and equation (18) is substituted into equation (4). T represents the transpose of a vector.

[0094] The temperature estimator 260 according to the second modification estimates the increasing component Δθh1 and the suppressing component Δθh2 in time series by applying a Kalman filter to the equation (17). The estimation method is similar to the flow of the equations (5) to (9).

[0095] Specifically, the temperature estimation unit 260 calculates a predicted value of the current increase component based on the previous estimated value of the increase component, and calculates a predicted value of the current suppression component based on the previous estimated value of the suppression component. The temperature estimation unit 260 calculates the estimated values ​​of the current increase component and the current suppression component by updating the predicted values ​​of the current increase component and the current suppression component using the measured value of the winding maximum point temperature as the observed value. The temperature estimation unit 260 calculates an estimated value of the current winding maximum point temperature based on each of the calculated estimated values ​​and the estimated value of the maximum oil temperature.

[0096] Similarly, the temperature estimation unit 260 calculates the covariance matrix of the predicted value of the current increase component based on the covariance matrix of the estimated value of the previous increase component, and calculates the covariance matrix of the predicted value of the current suppression component based on the covariance matrix of the estimated value of the previous suppression component. The temperature estimation unit 260 calculates a Kalman gain based on the covariance matrix of the predicted values ​​of each of the current increase component and suppression component and the covariance matrix of the observation noise. The temperature estimation unit 260 calculates the covariance matrix of the estimated values ​​of each of the current increase component and suppression component by updating the covariance matrix of the predicted values ​​of each of the current increase component and suppression component based on the Kalman gain.

[0097] According to the second modification, even if the constants in equations (10) and (11) used to calculate the winding maximum temperature are not valid, the winding maximum temperature can be estimated with higher accuracy by applying a Kalman filter to equation (17).

[0098] Embodiment 2 In the second embodiment, a configuration will be described in which machine learning is used to estimate the remaining life of the transformer 20 from an estimated value of the maximum oil temperature obtained using a Kalman filter. For example, the estimation device 10 according to the second embodiment is configured to perform supervised learning.

[0099] <Learning Phase> Fig. 5 is a configuration diagram of a learning device 400 according to the embodiment 2. With reference to Fig. 5, learning device 400 includes a data acquiring unit 420, a model generating unit 440, and a trained model storage unit 460. In this embodiment, a configuration of learning device 400 realized by (for example, built into) estimation device 10 will be described.

[0100] The data acquisition unit 420 acquires time series data of the estimated value of the maximum oil temperature estimated by the temperature estimation unit 260 using a Kalman filter and the remaining life of the transformer 20 as learning data. The time series data of the estimated value of the maximum oil temperature is N estimated values ​​of the maximum oil temperature θo(0) to θo(N-1). The remaining life of the transformer 20 is correct data of the remaining life obtained by the following known method. This method utilizes the fact that the life of the transformer 20 is determined by the deterioration of the insulation due to thermal history. In order to indirectly estimate the deterioration state of the insulation, the concentration of a characteristic component (e.g., furfural, "CO+CO2", etc.) dissolved in the insulating oil is measured. The characteristics of the time series change of this concentration are applied to a specified characteristic and extrapolated, and the time until a threshold value is reached is calculated as the life.

[0101] The model generation unit 440 learns the remaining life of the transformer 20 based on learning data created based on a combination of the time series data of the estimated value of the maximum oil temperature and the correct answer data of the remaining life. That is, the model generation unit 440 generates a trained model that estimates an optimal remaining life based on the time series data of the estimated value of the maximum oil temperature and the correct answer data of the remaining life. The learning data is data in which the time series data of the estimated value of the maximum oil temperature and the correct answer data of the remaining life are associated with each other.

[0102] The learning device 400 is used to learn the remaining life of the transformer 20, and may be, for example, a device connected to the estimation device 10 via a network and separate from the estimation device 10. Furthermore, the learning device 400 may exist on a cloud server.

[0103] The learning algorithm used by the model generation unit 440 may be a known algorithm such as supervised learning, unsupervised learning, reinforcement learning, etc. As an example, a case where a neural network is applied will be described.

[0104] The model generation unit 440 learns the remaining life of the transformer 20 by supervised learning, for example, according to a neural network model. Supervised learning is a method in which a set of input and result data is provided to the learning device 400, and the learning device 400 learns the features of the learning data and estimates the result from the input.

[0105] A neural network is composed of an input layer consisting of a number of neurons, an intermediate layer (e.g., a hidden layer) consisting of a number of neurons, and an output layer consisting of a number of neurons. The intermediate layer may be one layer, or two or more layers.

[0106] For example, in the case of a three-layered neural network, when multiple inputs are input to the input layer, the values ​​are multiplied by a weight W1 and input to the middle layer, and the result is further multiplied by a weight W2 and output from the output layer. This output result varies depending on the values ​​of the weights W1 and W2. The neural network learns the estimated value of the remaining life of the transformer 20 by supervised learning according to learning data created based on a combination of time-series data of the estimated value of the maximum oil temperature and correct data of the remaining life. That is, the neural network learns by inputting time-series data of the estimated value of the maximum oil temperature to the input layer and adjusting the weights W1 and W2 so that the result output from the output layer approaches the correct data of the remaining life of the transformer 20.

[0107] The model generation unit 440 generates and outputs a trained model by executing the above-mentioned learning. The trained model storage unit 460 stores the trained model output from the model generation unit 440.

[0108] Fig. 6 is a flowchart relating to the learning process of learning device 400 according to the second embodiment. The steps of Fig. 6 are typically executed by processor 101. Referring to Fig. 6, learning device 400 acquires time series data of the estimated value of the maximum oil temperature and the remaining life of transformer 20 (i.e., correct answer data of the remaining life) (step S30). Note that it is only necessary that the time series data of the estimated value of the maximum oil temperature and the correct answer data of the remaining life are input in association with each other, and these may be acquired at different times.

[0109] The learning device 400 learns the remaining life estimate by supervised learning according to learning data created based on a combination of the acquired time-series data of the estimated value of the maximum oil temperature and the ground truth data of the remaining life, and generates a trained model (step S32). The learning device 400 stores the generated trained model (step S34).

[0110] <Utilization phase> 7 is a configuration diagram of remaining life estimation unit 280 according to the embodiment 2. With reference to FIG. 7, remaining life estimation unit 280 includes a data acquisition unit 620 and an estimate value calculation unit 640.

[0111] The data acquisition unit 620 acquires time-series data of the estimated value of the maximum oil temperature estimated by the temperature estimation unit 260 using a Kalman filter.

[0112] The estimated value calculation unit 640 calculates an estimated value of the remaining life obtained by using the trained model. Specifically, the estimated value calculation unit 640 estimates the remaining life by inputting the time series data of the estimated value of the maximum oil temperature acquired by the data acquisition unit 620 to the trained model.

[0113] In the present embodiment, the estimation device 10 has been described as outputting an estimated value of the remaining life using a trained model trained by the model generation unit 440, but the present invention is not limited to this configuration. The estimate calculation unit 640 may obtain a trained model from outside and output an estimated value of the remaining life based on the trained model.

[0114] Fig. 8 is a flowchart showing the remaining life estimation process using the trained model. Referring to Fig. 8, the data acquisition unit 620 acquires time series data of the estimated value of the maximum oil temperature (step S50). The estimated value calculation unit 640 inputs the time series data of the estimated value of the maximum oil temperature estimated by the temperature estimation unit 260 to the trained model stored in the trained model storage unit 460, thereby calculating an estimated value of the remaining life of the transformer 20 (step S52).

[0115] The estimated value calculation unit 640 displays the estimated value of the remaining life obtained by the trained model on the display 105 (step S54). This allows the user to grasp the remaining life of the transformer 20.

[0116] In this embodiment, the case where supervised learning is applied to the learning algorithm used by the model generation unit 440 has been described, but the present invention is not limited to this. As for the learning algorithm, in addition to supervised learning, reinforcement learning, unsupervised learning, semi-supervised learning, etc. can also be applied. In addition, as the learning algorithm used by the model generation unit 440, deep learning that learns to extract the feature amount itself can also be used, and machine learning may be performed according to other known methods, such as genetic programming, functional logic programming, and support vector machines.

[0117] <Advantages> According to the second embodiment, the remaining life of the transformer 20 can be estimated with high accuracy using machine learning.

[0118] Other embodiments. (1) In the above-mentioned first modification of the first embodiment, the coefficients α and β of the formula (14) are calculated by linear regression of a series of plotted data points, but the present invention is not limited to this configuration. For example, N maximum oil temperature measurements from the past to the present, θo(0), θo(1), ..., θo(N-1), are prepared, and the coefficient (e.g., corresponding to the coefficient α) and intercept (e.g., corresponding to the coefficient β) of θo(t) obtained by the formula (2) are calculated using the measured maximum point temperatures at the N points, and these are used as the correct data for the coefficients α and β. In this case, the learning data is data in which the measured maximum oil temperatures and the correct data for the coefficients α and β are associated with each other.

[0119] The learning device 400 (for example, the model generation unit 440) learns the estimated values ​​of the coefficients α and β by supervised learning according to the learning data, and generates a trained model. This trained model is a trained model for estimating optimal coefficients α and β from N measured values ​​of the maximum oil temperature. The generation unit 240 calculates the estimated values ​​of the coefficients α and β in equation (14) by inputting the N measured values ​​of the maximum oil temperature to the trained model. The generation unit 240 generates a differential equation using the calculated coefficients α and β.

[0120] (2) The configurations exemplified as the above-mentioned embodiments are examples of the configurations of the present disclosure, and may be combined with other known technologies, or may be modified, such as by omitting some parts, without departing from the scope of the present disclosure. In addition, the above-mentioned embodiments may be implemented by appropriately adopting the processes and configurations described in other embodiments.

[0121] <Additional Notes> Various aspects of the present disclosure are summarized below as appendices.

[0122] (Appendix 1) An estimation device comprising: an acquisition unit that acquires measured values ​​of a maximum oil temperature of a transformer; a generation unit that generates a relational equation indicating a relationship between the current maximum oil temperature and the past maximum oil temperature; and a temperature estimation unit that estimates the maximum oil temperature in a time series by applying a Kalman filter to the relational equation, wherein the temperature estimation unit calculates a predicted value of the maximum oil temperature based on a previous estimate of the maximum oil temperature, and calculates a current estimate of the maximum oil temperature by updating the predicted value using the measured value as an observation value.

[0123] (Appendix 2) The estimation device described in Appendix 1, wherein the acquisition unit further acquires a measured value of an ambient temperature of the transformer, and the generation unit converts a differential equation for the maximum oil temperature using a plurality of predetermined parameters and the measured value of the ambient temperature, and generates a difference equation obtained by the conversion as the relational equation.

[0124] (Appendix 3) The estimation device according to claim 1, wherein the generation unit obtains, for each of a plurality of timings, a data point including the measurement value measured at that timing and the measurement value measured at a timing next to the corresponding timing, and generates the relational equation by linear regression of the series of data points.

[0125] (Appendix 4) The estimation device according to any one of Addendum 1 to Addendum 3, wherein the temperature estimation unit calculates a covariance matrix of the predicted value based on a covariance matrix of the previous estimated value, calculates a Kalman gain based on the covariance matrix of the predicted value and a covariance matrix of observation noise, and updates the covariance matrix of the predicted value based on the Kalman gain, thereby calculating the covariance matrix of the current estimated value.

[0126] (Appendix 5) The estimation device according to any one of Supplementary Note 1 to Supplementary Note 4, further comprising a remaining life estimation unit that estimates a remaining life of the transformer based on the estimated value estimated by the temperature estimation unit.

[0127] (Appendix 6) The estimation device described in Appendix 5, wherein the remaining life estimation unit estimates the remaining life of the transformer by inputting the time series data of the maximum oil temperature estimated by the temperature estimation unit into a trained model for estimating the remaining life of the transformer from the time series data of the estimated value of the maximum oil temperature.

[0128] The embodiments disclosed herein should be considered to be illustrative and not restrictive in all respects. The scope of the present disclosure is defined by the claims, not the above description, and is intended to include all modifications within the meaning and scope of the claims. [Explanation of symbols]

[0129] 10 Estimation device, 20 Transformer, 40 Monitoring device, 100 Estimation system, 101 Processor, 103 Memory, 105 Display, 107 Input device, 109 Input / output interface, 111 Communication interface, 220 Acquisition unit, 240 Generation unit, 260 Temperature estimation unit, 280 Remaining life estimation unit, 400 Learning device, 420, 620 Data acquisition unit, 440 Model generation unit, 460 Learned model memory unit, 640 Estimation value calculation unit.

Claims

1. an acquisition unit for acquiring a measured value of a maximum oil temperature of the transformer; A generation unit that generates a relational expression indicating a relationship between the current maximum oil temperature and the past maximum oil temperature; a temperature estimation unit that estimates the maximum oil temperature in a time series manner by applying a Kalman filter to the relational expression, The temperature estimation unit is Calculating a predicted value of the maximum oil temperature based on a previous estimated value of the maximum oil temperature; an estimation device that calculates a current estimated value of the maximum oil temperature by updating the predicted value using the measured value as an observed value.

2. The acquisition unit further acquires a measurement value of an ambient temperature of the transformer, The estimation device according to claim 1 , wherein the generation unit converts a differential equation for the maximum oil temperature using a plurality of predetermined parameters and the measured value of the ambient temperature, and generates a difference equation obtained by the conversion as the relational expression.

3. The generation unit is For each of a plurality of timings, a data point including the measurement value measured at the timing and the measurement value measured at a timing next to the timing is obtained; The estimator of claim 1 , wherein the relationship is generated by linear regression of a series of the data points.

4. The temperature estimation unit is Calculating a covariance matrix of the predicted value based on a covariance matrix of the previous estimated value; Calculating a Kalman gain based on a covariance matrix of the predicted value and a covariance matrix of the observation noise; 4. The estimation device according to claim 1, further comprising: a covariance matrix of the current estimated value calculated by updating a covariance matrix of the predicted value based on the Kalman gain.

5. The estimation device according to any one of claims 1 to 3, further comprising a remaining life estimation unit that estimates a remaining life of the transformer based on the estimated value estimated by the temperature estimation unit.

6. 6. The estimation device according to claim 5, wherein the remaining life estimation unit estimates the remaining life of the transformer by inputting the time series data of the maximum oil temperature estimated by the temperature estimation unit into a trained model for estimating the remaining life of the transformer from the time series data of the estimated value of the maximum oil temperature.

Images