DETERMINING THE FLUID DENSITY IN AN ELECTRICAL DEVICE

Abstract

In a method for determining a fluid density of a fluid in an encapsulated electrical device a sensor unit is used to acquire measurement data. The fluid density is derived from the measurement values. Weather data relating to weather conditions in an environment of the electrical device are collected. Via machine learning, a digital model is generated for the influence of the weather conditions on a measurement deviation of a measurement value from the true fluid density. Using the digital model, a correction value is calculated for measurement values according to the weather data and a measurement value is corrected using the correction value.

Claims

1-15. (canceled)

16. A method for determining a fluid density of a fluid in an encapsulated electrical device, the method comprising: acquiring measurement data by a sensor unit and deriving from the measurement data measurement values for the fluid density; collecting weather data relating to weather conditions in an environment of the electrical device; using machine learning to generate a digital model for an influence of the weather conditions on a measurement deviation of a measurement value from a true fluid density; using the digital model to calculate a correction value for the measurement values as a function of the weather data; and correcting a measurement value with the correction value.

17. The method according to claim 16, wherein the digital model comprises an artificial neural network having a plurality of layers of networked artificial neurons.

18. The method according to claim 17, wherein the artificial neural network is a recurrent artificial neural network.

19. The method according to claim 17, wherein the artificial neural network comprises at least one memory-enabled cell.

20. The method according to claim 16, which comprises transferring at least one of the measurement data or the measurement values to a data cloud, and/or calculating the correction value with the digital model in a data cloud.

21. The method according to claim 16, which comprises training the digital model by generating further training values for measurement values and/or weather data from measurement values and/or weather data.

22. The method according to claim 21, which comprises generating the training values by at least one of: temporally shifting weather data relative to measurement values, scaling measurement values and/or weather data, or shifting a value range of the measurement values.

23. The method according to claim 16, which comprises training the digital model by generating training values for simulated fluid losses.

24. The method according to claim 16, which comprises specifying a calculation period and calculating with digital model the correction value for the measurement values that are acquired within the calculation period is calculated.

25. The method according to claim 24, which comprises specifying a period of 24 hours for the calculation period.

26. The method according to claim 16, wherein the weather data are selected from the group consisting of a temperature, a wind speed, precipitation, an air humidity, and an air pressure in the environment of the electrical device.

27. The method according to claim 16, which comprises generating the digital model specifically for a given electrical device.

28. The method according to claim 16, which comprises generating the digital model for mutually different electrical devices.

29. The method according to claim 16, which comprises feeding only measurement values and weather data as input variables to the digital model.

30. The method according to claim 16, which comprises feeding measurement values, weather data, and additional data, generated from the measurement values and the weather data, as input variables to the digital model.

31. A computer program, comprising computer-executable commands which, when the commands are executed by a control unit or in a data cloud, implement the digital model of the method according to claim 16.

32. An electrical device with encapsulated fluid, the electrical device comprising: a sensor unit for acquiring measurement data relating to a fluid density of the fluid; a control unit or a connection to a data cloud; a computer program residing in the control unit or in the data cloud, the computer program being configured to: derive measurement values from the measurement data acquired by the sensor unit; collect weather data relating to weather conditions in an environment of the electrical device; use machine learning to generate a digital model for an influence of the weather conditions on a measurement deviation of the measurement values from a true fluid density of the fluid; use the digital model to calculate correction values for the measurement values as a function of the weather data; and correct the measurement values with the correction values.

Description

[0019] The only FIGURE shows a structural diagram of an exemplary embodiment of the method according to the invention for determining a fluid density of a fluid in an encapsulated electrical device 1.

[0020] In the exemplary embodiment shown, measurement data 5 is acquired with a sensor unit 3 and from the measurement data 5, measurement values 9 for the fluid density are derived using a processing unit 7. In addition, weather data 13 provided by a weather data source 11 relating to weather conditions in an environment of the electrical device 1 is collected. Machine learning is used to generate a digital model 15 for the influence of the weather conditions on the measurement deviation of a measurement value 9 from the true fluid density. Using the digital model 15, a correction value 17 for measurement values 9 is calculated as a function of the weather data 13 and a measurement value 9 is corrected with the correction value 17.

[0021] For example, the electrical device 1 is a gas-insulated switchgear system and the fluid is a pressurized insulating gas in the gas-insulated switchgear system. Alternatively, the electrical device 1, for example, is an oil-filled transformer and the fluid is a transformer oil in the transformer. The invention is not restricted to a nature or type of the electrical device.

[0022] The sensor unit 3 is configured to acquire measurement data 5 from which measurement values 9 of the fluid density can be derived. For example, the sensor unit 3 has sensors that are configured to acquire a fluid pressure of the fluid and a fluid temperature of the fluid as measurement data 5. Alternatively, the sensor unit 3, for example, has two quartz oscillators, one quartz oscillator being operated in a controlled reference environment and the other quartz oscillator being operated in the fluid, and the sensor unit 3 detects resonance frequencies of the two quartz oscillators as measurement data 5. The invention is not restricted to a nature or type of the sensor unit 3.

[0023] The processing unit 7 determines a measurement value 9 for the fluid density from the measurement data 5. For example, if the measurement data 5 comprises a fluid pressure of the fluid and a fluid temperature of the fluid, the processing unit 7 calculates a measurement value 9 for the fluid density from the fluid pressure and the fluid temperature. For example, if the measurement data 5 comprises resonance frequencies of two quartz oscillators of a sensor unit 3 described above, the processing unit 7 calculates a measurement value 9 for the fluid density from the difference between the resonance frequency in the fluid and the resonance frequency in the reference environment. The invention is not restricted to a nature or type of the processing unit 7.

[0024] For example, the weather data source 11 is a weather station that collects the weather data 13. Alternatively, the weather data source 11 is a weather database, for example in a data cloud, that provides the weather data 13. The weather data source 11 can also comprise a weather station and such a weather database. The weather data 13 comprises, for example, a temperature, a wind speed, precipitation, an air humidity and/or an air pressure in the environment of the electrical device 1. The invention is not restricted to a nature or type of the weather data source 11.

[0025] The digital model 15 has an artificial neural network 19 having a plurality of layers 21, 22, 23 of networked artificial neurons 25 and memory-enabled cells 27 (LSTM cells). The neural network 19 is designed as a recurrent neural LSTM network. An arrow from a neuron 25 to another neuron 25 or to a memory-enabled cell 27 symbolizes that an output value of the neuron 25 is transferred to the other neuron 25 or to the memory-enabled cell 27 as an input value. Accordingly, an arrow from a memory-enabled cell 27 to a neuron 25 symbolizes that an output value of the memory-enabled cell 27 is transferred to a neuron 25 as an input value. The neural network 19 here is only shown schematically with an input layer 21, an intermediate layer 22, an output layer 23, and a memory-enabled cell 27. In an actual embodiment, the neuronal network 19 has considerably more intermediate layers 22, neurons 25 and memory-enabled cells 27 than are shown in the FIGURE.

[0026] As an option, in addition to the measurement values 9 and the weather data 13 supplied as input variables the digital model 15 is also supplied with additional data 29, which is generated by the processing unit 7 from the measurement values 9 and the weather data 13. The additional data 29 includes, for example, derivatives of measurement values according to weather data 13, which describe, for example, changes in measurement values 9 as a function of the temperature or the air pressure in the environment of an electrical device 1. Such additional data 29 is fed to the digital model 15 as input variables, in particular when the digital model 15 is not only generated specifically for a particular electrical device 1 but for separate (but similar to each other) electrical devices 1.

[0027] For example, the measurement data 5 and/or the measurement values 9 are transferred to a data cloud and/or the correction value 17 is calculated with the digital model 15 in a data cloud.

[0028] For training the digital model 15, further training values for measurement values 9 and/or weather data 13 can be generated from measurement values 9 and/or weather data 13 by means of so-called data augmentation, in particular if there is initially an insufficient amount of suitable measurement values 9 and/or weather data 13 available for training the digital model 15. For example, such training values are generated by temporally shifting weather data 13 relative to measurement values 9, scaling measurement values 9 and/or weather data 13, and/or shifts in the value range of the measurement values 9. In addition, training values for simulated fluid losses are generated to train the digital model 15.

[0029] For example, the digital model is used to calculate a correction value 17 for measurement values 9 which are acquired within a specified calculation period, for example a period of 24 hours.

[0030] Although the invention has been illustrated and described in greater detail by means of preferred exemplary embodiment, the invention is not restricted by the examples disclosed and other variations can be derived therefrom by the person skilled in the art without departing from the scope of protection of the invention.