Monitoring of electrical devices using physical information neural machine learning

Abstract

The present disclosure provides a method for modifying a digital model, the method comprising: providing a digital model of an electrical device based on set boundary conditions of the electrical device; providing measurement data from a sensor detecting a physical aspect of the electrical device; and modifying the digital model by processing the measurement data by a physics informed neural network, PINN.

Description

Technical Field

[0001] This disclosure relates to a method for changing a digital model and a system for changing a digital model. Background Technology

[0002] The design, operation analysis, and diagnostics of complex electrical devices have always been nontrivial problems, and mathematical modeling has proven to be a useful tool in such research. However, it is not always possible to model the dynamics of complex devices at the sub-component level. Summary of the Invention

[0003] This disclosure includes an efficient surrogate model for capturing latent dynamics, which can facilitate diagnostics, design improvements, and feature recognition, as well as monitoring the operation of electrical installations, improving existing models, understanding uncertainties, and benchmarking for monitoring. Physical Information Machine Learning (PIML) uses specially equipped neural network models (also known as Physical Information Neural Networks (PINNs)) that can estimate dynamics governed by physical equations, such as partial differential equations (PDEs). Knowing the underlying physics, PINNs can present estimates with high accuracy but using a smaller amount of high-cost input data. Compared to classical solvers, trained PINNs can provide high-quality estimates much faster during inference.

[0004] PINN's ability to satisfactorily estimate complex dynamics using limited data can be leveraged by using it as a proxy model (digital twin / digital model) of a range of electrical installations during factory testing and operation. This disclosure describes possible embodiments of this emerging technology and how the tool can be used in various application scenarios.

[0005] This disclosure relates to a method for modifying a digital model, the method comprising: providing a digital model of an electrical device based on (initial) boundary conditions of the device’s setup; providing (first) measurement data from a (first) sensor that detects physical aspects of the electrical device; and modifying the digital model (to provide a modified digital model) by processing the measurement data by a (first) physical information neural network PINN.

[0006] Among other advantages, this method allows for the modification and / or improvement of digital models in a shorter time with less data (which may be of poor quality). Furthermore, it requires fewer computational resources.

[0007] A digital model can be a digital twin of an electrical installation. An initial digital model can be created from boundary conditions using software such as, for example, CAD software, PINN, or the like. The digital model can be created on the same computer system (server, processor, etc.) on which other (mentioned above) steps are processed. Alternatively, a digital model can be created on a different computer system and then downloaded to another computer system to process the steps mentioned above. The digital model can represent the entire electrical installation or a portion of it at the sub-component level.

[0008] Measurement data can be provided by sensors directly connected to a computer system that performs the steps mentioned above (which can be done in real time). Alternatively, measurement data can be generated and saved in advance (before performing other steps). The computer system performing the steps mentioned above does not need to be directly connected to the sensors. This can be advantageous because a computer system might require a significant amount of space, and integrating it into the electrical system would be difficult.

[0009] The physical aspect can be the physical properties of electrical devices. Examples of physical measurements include the dimensions (length) of a wall or wire, the width of a wall, the thickness of a wall, the current flowing through a conductor, the strength of a magnetic field, the location of partial discharge, and so on.

[0010] Physical quantities can be things that use the same unit (e.g., meter or ampere; meters and amperes are considered different units and therefore different physical quantities, while meters and millimeters are considered the same unit and therefore represent the same physical quantity). Different physical aspects can involve the same physical quantity. Examples of this could be the current flowing through different conductors, or the dimensions of different parts, or the length and width of an object.

[0011] A set of physical equations (PDEs) can be based on physical equations (such as the Navier-Stokes equations, Maxwell's equations, Ampere's law, the heat equation, etc.). Such physical equations often involve a set of physical quantities but not other physical quantities.

[0012] Various embodiments may preferably implement the following features.

[0013] Preferably, boundary conditions are geometric conditions and constraints. Boundary conditions may additionally or alternatively include parameters.

[0014] Preferably, the steps of modifying the digital model include: creating additional boundary conditions and / or adjusting the boundary conditions. Additionally or alternatively, additional parameters may be created and / or adjusted.

[0015] When a previous (or initial) digital model is based on an incomplete set of boundary conditions, it is possible to add boundary conditions and / or parameters. This can be called the "inverse problem." In this case, the (initial) boundary conditions may only include information about the various parts of the electrical installation. For example, the (initial) boundary conditions may only include information about the outer wall and that the transformer is located inside the outer wall. The exact wiring, dimensions, and locations may not have been provided in the initial boundary conditions. In this (inverse problem) scenario, the modified digital model may also include information about the exact wiring, dimensions, and locations inside the outer wall. In the inverse problem, PINN may also modify the PDE while modifying the digital model.

[0016] In different cases that can be termed "positive problems," the initial boundary conditions may include a complete set of information about the electrical installations, since all information regarding the building plans for the electrical installations is provided as initial boundary conditions. The boundary conditions can be changed by PINN if a deviation from the building plans or damage (etc.) is detected. This may be critical (and then the electrical installation needs repair), or it may not be critical (e.g., because it does not pose a danger). Preferably, in positive problems, the PDE is not changed after the (initial) digital model is provided.

[0017] Preferably, the step of modifying the digital model by processing the measurement data by PINN further includes modifying the digital model by processing the initial conditions of the digital model by PINN. Alternatively or additionally, PINN may also consider other data to modify the digital model.

[0018] Preferably, the method further includes: providing the types of one or more output variables; and providing information about the modified digital model based on the types of the one or more output variables. The step of “providing the types of one or more output variables” can be input by a user or through an API (Application Programming Interface) to another software.

[0019] Preferably, the method further includes: adjusting the electrical device according to the modified digital model.

[0020] This can further include inspecting electrical installations for faults and / or damage based on the revised digital model. Such inspections can be performed more efficiently because more information about the damage is now known.

[0021] Alternatively, the method may further include: examining the modified digital model (e.g., by comparing it with (other) measurement data) and correcting PINN. This trains and improves PINN.

[0022] Preferably, the electrical device is a transformer, shunt reactor, distribution transformer, circuit breaker, tap changer, bushing, or gas-insulated switchgear (GIS). The electrical device may also be another suitable electromagnetic / electromechanical device.

[0023] Preferably, the digital model is an acoustic model, and the measurement data is vibration data or acoustic data. Preferably, the digital model is an electromagnetic model, and the measurement data is electrical data, magnetic data, or electromagnetic data, with the sensor being a transient voltage-to-ground sensor, an electrical sensor, a magnetic sensor, or an electromagnetic sensor. Preferably, the digital model is a stray magnetic flux model, and the measurement data is stray magnetic flux data. Preferably, the digital model is a thermal model, the measurement data is thermal data, and the sensor is a thermal imager or a thermocouple. Preferably, the digital model is a fluid dynamics model. Other types of digital models, measurement data, and / or sensors, and combinations thereof, are also possible.

[0024] Preferably, the sensor is a fiber optic sensor, a static sensor, a motion sensor, a camera (visible light, ultraviolet light and / or infrared light), or a drone.

[0025] Preferably, the measurement data is one or more images. The measurement data may be in the form of video (visible light, ultraviolet light, and / or infrared light).

[0026] Preferably, the method further includes: providing second measurement data from a second sensor of a second physical aspect of the detection electrical device; wherein the step of changing the digital model by processing the (first) measurement data by the (first) PINN further includes: changing the digital model by processing the second measurement data by the second PINN.

[0027] Preferably, the second physical aspect and the (first) physical aspect involve the same physical measure, and wherein the second PINN is the (first) PINN.

[0028] Preferably, the method further includes (as a supplement or alternative to the above two paragraphs): providing third measurement data from a third sensor of a third physical aspect of the detection electrical device, wherein the third physical aspect and the (first) physical aspect involve different physical measures; and further modifying the modified digital model by processing the third measurement data with a third PINN using a partial differential equation (PDE) involving the third physical aspect, wherein the (first) PINN uses a different partial differential equation (PDE) involving the (first) physical aspect.

[0029] The first PINN and the second PINN can also be a PINN that considers the PDE with respect to two (first and third) physical measures.

[0030] Preferably, the boundary conditions on which the (initial) digital model is based are not a complete representation of the setup of the electrical apparatus.

[0031] This disclosure also relates to a system for modifying a digital model, the system comprising: a processor configured to provide a digital model of an electrical device based on boundary conditions set by the electrical device; and a sensor configured to detect physical aspects of the electrical device to provide measurement data; wherein the processor is further configured to modify the digital model by processing the measurement data by a Physical Information Neural Network (PINN).

[0032] The advantages described in each aspect are neither limited to nor exclusive to that aspect. One aspect may have additional advantages that are not explicitly mentioned.

[0033] The exemplary embodiments disclosed herein are intended to provide features that will become readily apparent from the following description when understood in conjunction with the accompanying drawings. Exemplary systems, methods, apparatuses, and computer program products are disclosed herein according to various embodiments. However, it should be understood that these embodiments are presented by way of example and not limitation, and that various modifications can be made to the disclosed embodiments by those skilled in the art who read this disclosure, while remaining within the scope of this disclosure. Attached Figure Description

[0034] The above and other aspects and their embodiments are described in more detail in the accompanying drawings, description and claims.

[0035] Figure 1 This is a schematic diagram illustrating a system for changing a digital model.

[0036] Figure 2 This is a schematic diagram of the PINN architecture.

[0037] Figure 3 This is a schematic illustration of an embodiment of a system for modifying a digital model.

[0038] Figure 4 This is a schematic illustration of an embodiment of a system for modifying a digital model.

[0039] Figure 5 It is a schematic diagram of a multi-step process used to modify a digital model.

[0040] Figure 6 This indicates the method used to modify the digital model. Detailed Implementation

[0041] Figure 1 This is a schematic diagram illustrating a system 10 for modifying a digital model. System 10 includes a computer system 12 (including a processor and other components necessary to run the computer system 12, the processor and other components not explicitly shown), a sensor 14, and an electrical device 16.

[0042] Computer system 12 may have been provided with boundary conditions for electrical device 16, and a digital model may have been generated based on these boundary conditions. Alternatively, computer system 12 may have been provided with a digital model. The computer system is connected (wired or wirelessly) to sensor 14. Computer system 12 may be a computer (fixed or mobile), a server, a computing center (or part thereof), etc.

[0043] Sensor 14 can be attached to electrical device 16 or positioned at a distance from electrical device 16. The sensor can detect a physical quantity. Sensor 14 is connected to computer system 12. Sensor 14 can output digital or analog data. Sensor 14 can be any kind of sensor used for the desired physical quantity. Some possible examples of sensor 14 are: transient voltage to ground (TEV) sensor, microphone, transducer, thermometer, electrical sensor, ammeter, voltmeter, magnetic sensor, electromagnetic sensor, thermal imager, thermocouple, barometer.

[0044] Electrical device 16 includes an outer wall 18 and an inner portion 20. The inner portion 20 represents electrical components. The outer wall 18 is optional and represents a kind of enclosure. Inside the outer wall 18, there exists a source 22 (a source of a certain physical quantity; for example, partial discharge) and an effect 24 (the same physical quantity; for example, in the form of electromagnetic waves). If sensor 14 is configured to sense the corresponding physical quantity, effect 24 can be detected by the sensor. Sensor 14 creates measurement data from this. This measurement data is then sent to computer system 12 (and processor). There, the measurement data is processed by a Physical Information Neural Network (PINN) that operates on computer system 12 to modify the digital model.

[0045] Source 22 can be any kind of physical quantity that can be detected by sensor 14. Source 22 can be a dimension of the outer wall 18, which can be measured as a physical quantity of length (e.g., in meters). Alternatively, source 22 can be a partial discharge that generates electromagnetic waves, can be sound waves, and can be heat. Electromagnetic waves can be detected by transient voltage to ground (TEV) sensor 14, etc. Sound waves can be detected by a microphone, transducer, or some other acoustic sensor 14. Heat can be detected by a thermometer or some other thermal sensor 14. However, this disclosure is not limited to these examples.

[0046] In one example, electrical device 16 is a transformer box. Source 22 is a partial discharge in electrical device 16 with electromagnetic radiation that can induce a voltage in the outer wall 18. Sensor 14 is a TEV sensor (which can be mounted on the outer wall 18). TEV sensor 14 can measure the induced voltage in the outer wall 18. The measurement data from TEV sensor 14 can then be used by PINN using corresponding partial differential equations (PDEs) (e.g., Maxwell's equations). In this way, the modified digital model can show the location of the partial discharge in electrical device 16. Electrical device 16 can then be repaired or modified.

[0047] In another example, the acoustic signal from partial discharge can be detected by a corresponding sensor 14. PINN can then use the acoustic equations and measurement data to create a modified digital model. The same applies to heat generated by partial discharge. However, this disclosure is not limited to these examples.

[0048] In another example, electrical device 16 includes a rotating electric motor. Stray flux can be measured on the housing of the rotating electric motor using stray flux (such as source 22). Using a modified digital model (based on these measured stray fluxes), faults such as inter-turn short circuits and eccentricity in the rotor shaft can be detected.

[0049] In another example, vibration measurements can be used to detect problems in rotating electrical machines. In this way, the location, type, and severity of the damage can be determined using a modified digital model.

[0050] In this manner, it is also possible to detect faults within the outer wall 18 without opening the electrical installation 16. This facilitates detection. Without this disclosure, it is extremely difficult to detect internal damage (such as mechanical deformation, movement, tilting, misalignment, minor inter-turn short circuits) in electrical installations in service (e.g., power transformers). Such damage alters the distribution of stray magnetic flux, which in turn affects the loss distribution on the tank walls.

[0051] Figure 2 This is a schematic diagram of the PINN architecture. PINN can be applied using mathematical equations describing the dynamics of the underlying system (electrical device). This is preferably accomplished using PDEs. The PDEs can be completely known (forward problem), or some components of the PDEs can be unknown or uncertain (inverse problem).

[0052] exist Figure 2 The right-hand box shows a PDE. An example of Maxwell's equations (used for detecting discharges) is shown. Different equations can also be used as the basis for a PDE. Figure 2The left side shows the box representing the neural network of PINN. PINN uses sensed measurement data (voltage V and discharge current i). d To find more information about the discharge (location X, y; time t; and source current I (preferably)).

[0053] Following the PDE, additional components (some or all) can be defined: parameters, boundary conditions, initial conditions, input variables and their measurements, output variables, and training and validation. Parameters and boundary conditions have been explained above. The initial conditions for a time-dynamic electrical device provide the values ​​of the variables in the dynamic state at the initial time (and preferably the corresponding initial time). Input variables and their measurements can be defined because PINN knows what kind of measurement data the sensors supply. Additional information, such as the sampling rate, can also be provided. By providing output variables, the PINN system knows what information is requested as output (the modified digital model). Multiple output variables can be specified by software (e.g., via an API), or they can be specified by the user through a graphical user interface. It can also be defined whether the problem at hand is inverse or forward, and / or whether / should the dynamics (PDE) be modified. The PINN system can also be provided with information about whether training or validation has occurred. It is preferable to train the PINN model offline, followed by validation, before implementing it for online operation. Depending on the availability of measurement data, appropriate validation methods can be recommended. For example, in the forward problem, a direct validation method with real ground data can be applied, but for the inverse problem, due to the lack of real ground data, an indirect method is preferred to validate the model.

[0054] Figure 3 and Figure 4 This is a schematic illustration of an embodiment of a system for modifying a digital model. The PINN technique for estimating the dynamics of electrical installations (modifying the digital model) can be used for applications such as factory testing (see...). Figure 3 ) and on-site monitoring of the device (see Figure 4 Applications such as [missing information], and therefore can be used as a service platform. Both figures show sensor 14, computing unit with test chamber software 26, data preprocessing 28, computer system 12, design specification 30, target platform 32, and report 34.

[0055] Sensor 14 can detect measurement data from electrical device 16, as described above.

[0056] Test chamber software 26 can run on computer system 12 (where PINN also runs) or on a different computing unit. Test chamber software 26 can simulate electrical devices and provide measurement data without the need for an actual electrical device 16. Test chamber software 26 and sensor 14 can be alternatives.

[0057] Data preprocessing 28 is an optional stage in which measurement data are prepared and / or saved until they are used by PINN.

[0058] Computer system 12 implements PINN 13a, which interacts with processor 13b. Computer system 12 (and therefore PINN) may also be supplied with boundary data and other information (parameters, initial conditions) by design specification 30. Alternatively, design specification 30 may supply an initial digital model to computer system 12. Design specification 30 may, for example, be an electrical design system (EDS) and / or a mechanical design system (MDS).

[0059] The target platform 32 can store the modified digital model. The target platform 32 can be a memory module (hard disk drive, RAM, or the like) of the computer system 12.

[0060] Report 34 outputs information, preferably as specified above (see Output Variables above). The output may be information displayed on a monitor (to the user). The output may be provided to another software that can use it to perform additional steps.

[0061] Figure 3 The details are in the target platform 32 feed to report 34. When Figure 3 During an exemplary factory test, additional information about the electrical device 16 is extracted. This allows the electrical device 16 to be modified if a fault or a potentially fragile component is detected.

[0062] Figure 4 The details in (indicating monitoring) refer to another check performed on sensor data 36 and alarm 38. Another sensor 36 is checked to compare the modified digital model with actual, additional measurement data. If this comparison confirms damage or a critical situation, alarm 38 will sound. Additionally, a report 34 can be output.

[0063] Figure 5 This is a schematic diagram illustrating a multi-step process for modifying a digital model. In most cases, only one physical measure (a set of measurement data) and PINN are considered, only for a specific set of PDEs. It is also possible to consider different physical aspects of the electrical device 16. Preferably, multiple sets of measurement data from different sensors 14 are considered for different physical aspects (and possibly even different physical measures). Each set of measurement data may be provided by a corresponding sensor 14. Figure 5 The multi-step process demonstrates how each set of measurement data can be considered one after another in successive (sub)steps of changing the digital model.

[0064] In step 40, a first set of measurement data (e.g., transformer winding geometry) is provided from a first sensor. Arrow 42 indicates the step of using a first set of equations (e.g., based on Ampere's law) to modify the digital model (by PINN). Then, box 44 indicates the step of providing a second set of measurement data (e.g., stray flux distribution) from a second sensor. In step 46, using this second set of measurement data from step 42 and the modified digital model, the digital model is again modified (by the same or another PINN) based on (the same or another) set of equations (e.g., Maxwell's equations). The resulting (further) modified digital model then reaches box 48. Here, another set of measurement data (e.g., heat generation in the chamber during startup) is provided. Then, in step 50 (which is, for example, based on thermal equations), the additional modified digital model and another set of measurement data are processed again (by the same or another PINN). Finally, in step 52, the final product (e.g., temperature distribution on the output chamber) is output.

[0065] Alternatively, a single PINN can be used to replace the multi-step process, which implements all (different sets of) equations as a combined set of PDEs and processes all measurement data at once.

[0066] Figure 6 Method 60 for modifying the digital model is described. Step 61 provides a digital model of the electrical device based on the boundary conditions of the device's setup. Step 62 provides measurement data from sensors detecting the physical aspects of the electrical device. Step 63 modifies the digital model by processing the measurement data using a physical information neural network (PINN).

[0067] In this embodiment, the boundary conditions are geometric conditions and constraints.

[0068] In an embodiment, the steps of modifying the digital model include: creating additional boundary conditions and / or adjusting the boundary conditions.

[0069] In an embodiment, the step of changing the digital model by processing measurement data with PINN further includes: changing the digital model by processing the initial conditions of the digital model with PINN.

[0070] In an embodiment, the method further includes: providing the types of one or more output variables; and providing information about the modified digital model based on the types of one or more output variables.

[0071] In one embodiment, the method further includes: adjusting the electrical device according to the modified digital model.

[0072] In the embodiments, the electrical equipment is a transformer, a shunt reactor, a distribution transformer, a circuit breaker, a tap changer, a bushing, or a gas-insulated switchgear (GIS).

[0073] In this embodiment, the digital model is an acoustic model, and the measurement data is vibration data or acoustic data.

[0074] In this embodiment, the digital model is an electromagnetic model, the measurement data is electrical data, magnetic data, or electromagnetic data, and the sensor is a transient voltage-to-ground sensor, an electrical sensor, a magnetic sensor, or an electromagnetic sensor.

[0075] In this embodiment, the digital model is a stray flux model, and the measurement data is stray flux data.

[0076] In this embodiment, the digital model is a thermal model, the measurement data is thermal data, and the sensor is a thermal imager or a thermocouple.

[0077] In this embodiment, the digital model is a fluid dynamics model.

[0078] In the embodiments, the sensor is a fiber optic sensor, a static sensor, a motion sensor, a camera, or a drone.

[0079] In this embodiment, the measurement data is one or more images.

[0080] In an embodiment, the method further includes: providing second measurement data from a second sensor that detects a second physical aspect of the electrical device, wherein the step of changing the digital model by processing the measurement data by a PINN further includes: changing the digital model by processing the second measurement data by a second PINN.

[0081] In the embodiments, the second physical aspect and the physical aspect refer to the same physical measure, and wherein the second PINN is PINN.

[0082] In an embodiment, the method further includes: providing third measurement data from a third sensor of a third physical aspect of the detection electrical device, wherein the third physical aspect and the physical aspect involve different physical quantities; and further modifying the modified digital model by processing the third measurement data with a third PINN using a partial differential equation (PDE) involving the third physical aspect, wherein the PINN uses a different PDE involving the physical aspect.

[0083] In this embodiment, the boundary conditions on which the digital model is based are not a complete representation of the setup of the electrical apparatus.

[0084] While various embodiments of this disclosure have been described above, it should be understood that they are presented by way of example only and not by way of limitation. Similarly, various figures may depict exemplary architectures or configurations, provided to enable those skilled in the art to understand the exemplary features and functionality of this disclosure. However, such persons will understand that this disclosure is not limited to the illustrated exemplary architectures or configurations, but can be implemented using various alternative architectures and configurations. Additionally, as those skilled in the art will understand, one or more features of one embodiment may be combined with one or more features of another embodiment described herein. Therefore, the breadth and scope of this disclosure should not be limited to any of the exemplary embodiments described above.

[0085] It should also be understood that any reference to elements in this document using names such as "first," "second," etc., generally does not restrict the number or order of those elements. Rather, these names may be used herein as a convenient means of distinguishing between two or more elements or instances of elements. Therefore, references to the first element and the second element do not imply that only two elements may be used, or that the first element must somehow precede the second element.

[0086] Additionally, those skilled in the art will understand that information and signals can be represented using any of a variety of different techniques and methods. For example, data, instructions, commands, information, signals, bits, and symbols, as referenced in the above description, can be represented by voltage, current, electromagnetic waves, magnetic fields or particles, light fields or particles, or any combination thereof.

[0087] Various modifications to the embodiments described in this disclosure will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other embodiments without departing from the scope of this disclosure. Therefore, this disclosure is not intended to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the novel features and principles disclosed herein, as set forth in the appended claims.

Claims

1. A method (60) for changing a digital model, the method (60) comprising: The digital model of the electrical device (16) is provided (61) based on the boundary conditions of the electrical device (16). Provide (62) first measurement data from the first sensor (14) that detects the first physical aspect of the electrical device (16); The digital model (63) is modified by processing the first measurement data using the first physical information neural network PINN; and Provide second measurement data from a second sensor that detects a second physical aspect of the electrical device (16), wherein the second measurement data is different from the first measurement data, the second sensor is different from the first sensor, and the second physical aspect is different from the first physical aspect; The step of modifying the digital model by processing the first measurement data with the first PINN further includes: modifying the digital model by processing the second measurement data with a second PINN, wherein the second PINN is different from the first PINN; and The method (60) further includes: adjusting the electrical device (16) according to the modified digital model.

2. The method according to claim 1, wherein, The boundary conditions are geometric conditions and constraints.

3. The method according to claim 1, wherein, The steps for modifying the digital model include: creating additional boundary conditions and / or adjusting the boundary conditions.

4. The method according to claim 1, wherein, The step of changing the digital model by processing the first measurement data by the first PINN further includes: changing the digital model by processing the initial conditions of the digital model by the first PINN.

5. The method of claim 1, further comprising: Provide the types of one or more output variables; Provide information about the modified digital model based on the type of one or more output variables; as well as Information is output on the display via a report (34).

6. The method according to any one of claims 1 to 5, in, The electrical device (16) is a transformer, a shunt reactor, a distribution transformer, a circuit breaker, a tap changer, a bushing, or a gas-insulated switchgear (GIS).

7. The method according to any one of claims 1 to 5, wherein, The digital model is an acoustic model, and the first measurement data is vibration data or acoustic data; or Wherein, the digital model is an electromagnetic model, the first measurement data is electrical data, magnetic data, or electromagnetic data, and the first sensor (14) is a transient voltage sensor to ground, an electrical sensor, a magnetic sensor (14), or an electromagnetic sensor; or Wherein, the digital model is a stray flux model, and the first measurement data is stray flux data; or Wherein, the digital model is a thermal model, the first measurement data is thermal data, and the first sensor (14) is a thermal imager or a thermocouple; or The digital model is a fluid dynamics model.

8. The method according to any one of claims 1 to 5, wherein, The first sensor (14) is an optical fiber sensor, a static sensor, a motion sensor, a camera, or a drone.

9. The method according to any one of claims 1 to 5, wherein, The first measurement data is one or more images.

10. The method according to any one of claims 1 to 5, further comprising: Provide third measurement data from a third sensor that detects a third physical aspect of the electrical device (16), wherein the third physical aspect and the first physical aspect involve different physical quantities; and The modified digital model is further altered by processing the third measurement data using a partial differential equation (PDE) involving the third physical aspect by a third PINN, wherein the first PINN uses a different PDE involving the first physical aspect.

11. The method according to any one of claims 1 to 5, wherein, The boundary conditions on which the digital model is based are not a complete representation of the setup of the electrical device (16).

12. A system for changing a digital model, the system (10, 12) comprising: The processor (13b) is configured to provide a digital model of the electrical device (16) based on the boundary conditions set by the electrical device (16); as well as A first sensor (14) is configured to detect a first physical aspect of the electrical device (16) to provide first measurement data; The processor (13b) is further configured to: The digital model is modified by processing the first measurement data using a first physical information neural network PINN; and Provide second measurement data from a second sensor that detects a second physical aspect of the electrical device (16), wherein the second measurement data is different from the first measurement data, the second sensor is different from the first sensor, and the second physical aspect is different from the first physical aspect; The method of modifying the digital model by processing the first measurement data with the first PINN further includes: modifying the digital model by processing the second measurement data with a second PINN, wherein the second PINN is different from the first PINN; and The system further includes the electrical device (16) configured to be adjusted according to the modified digital model.

Images