System for monitoring a circuit breaker

Abstract

The invention relates to a system for monitoring a circuit breaker, the system comprising: at least one sensor; and a processing unit; wherein the at least one sensor is configured to be positioned and utilized to obtain at least one time series sensor data of at least one first part of the operating circuit breaker; wherein the at least one sensor is configured to provide the at least one time series sensor data of the at least one first part of the operating circuit breaker to the processing unit; and wherein the processing unit is configured to determine whether at least one second part of the operating circuit breaker is operating correctly or the at least one second part of the operating circuit breaker has a fault, wherein the determination comprises an analysis of the at least one time series sensor data of the at least one first part of the operating circuit breaker by a trained neural network implemented by the processing unit.

Description

TECHNICAL FIELD

[0001] The present invention relates to a system for monitoring a circuit breaker, a system for monitoring a two- or three-phase switching device or control device, a method for monitoring a circuit breaker, a system for training a neural network for monitoring a circuit breaker, and a method for training a neural network for monitoring a circuit breaker. BACKGROUND

[0002] Problems of industrial assets such as circuit breakers can often be measured using time series sensor data. However, it is often difficult to extract relevant information from the signals, especially when the signal-to-noise ratio is low. It is also impractical for a human to continuously monitor such signals and it can be difficult to obtain sensor data from the required parts of such assets on site.

[0003] There is a need to address these problems. SUMMARY

[0004] It can therefore be advantageous to have an improved technology for monitoring assets such as circuit breakers.

[0005] In a first aspect, there is provided a system for monitoring a circuit breaker, the system comprising:

[0006] at least one sensor; and

[0007] a processing unit.

[0008] The at least one sensor is configured to be positioned and utilized to obtain at least one time series sensor data of at least one first part of the operating circuit breaker. The at least one sensor is configured to provide the at least one time series sensor of the at least one first part of the operating circuit breaker to the processing unit. The processing unit is configured to determine whether at least one second part of the operating circuit breaker is operating correctly or has a fault. The determination comprises an analysis of the at least one time series sensor data of the at least one first part of the operating circuit breaker by a trained neural network implemented by the processing unit.

[0009] In an example, the neural network is trained on the basis of at least one time series sensor data of at least one first part of a calibration circuit breaker, wherein the at least first part of the calibration circuit breaker is operating correctly. The neural network is further trained on the basis of at least one time series sensor data of at least one second part of the calibration circuit breaker, wherein the at least second part of the calibration circuit breaker is operating correctly. The calibration circuit breaker is of the same type or model as the operating circuit breaker.

[0010] In an example, the at least one time series sensor of the at least one first part of the calibration circuit breaker is obtained simultaneously with the at least one time series sensor of the at least one second part of the calibration circuit breaker.

[0011] In an example, the at least one sensor utilized to obtain the at least one time series sensor data of the at least one first portion of the calibration circuit breaker is of the same type or model as the at least one sensor utilized to obtain the at least one time series sensor data of the at least one first portion of the operating circuit breaker.

[0012] In an example, the at least one sensor utilized to obtain the at least one time series sensor data of the at least one first portion of the calibration circuit breaker is located at the same or equivalent at least one location as the at least one sensor utilized to obtain the at least one time series sensor data of the at least one first portion of the calibration circuit breaker.

[0013] In an example, the trained neural network is configured to determine at least one synthetic time series sensor data of the at least one second portion of the operating circuit breaker. The determination of the at least one synthetic time series sensor data comprises an analysis of the at least one time series sensor data of the at least one first portion of the operating circuit breaker by the trained neural network. The determination of whether the at least one second portion of the operating circuit breaker is operating correctly or the at least one second portion of the operating circuit breaker has a fault can then comprise a comparison of the at least one synthetic time series sensor data of the at least one second portion of the operating circuit breaker with the baseline saved data.

[0014] In an example, the baseline saved data comprises at least one time series sensor data of the at least one second portion of the circuit breaker, wherein the at least one second portion of the circuit breaker is operating correctly.

[0015] In an example, the baseline saved data comprises at least some of the following data: at least one time series sensor data of the at least one second portion of the calibration circuit breaker used in training the neural network, and / or at least one synthetic time series sensor data of the at least one second portion of the calibration circuit breaker generated by the neural network from at least some of the at least one time series sensor data of the at least one first portion of the calibration circuit breaker.

[0016] In an example, the processing unit is configured to determine that the at least one second portion of the operating circuit breaker has a fault on the basis that a distance measure between the at least one synthetic time series sensor data of the at least one second portion of the operating circuit breaker and the baseline saved data is equal to or greater than a threshold value.

[0017] In an example, the system comprises an output unit configured to output information indicating that the at least one second portion of the operating circuit breaker has a fault.

[0018] In an example, the neural network comprises a sequence-to-sequence model.

[0019] In a second aspect, a system for monitoring a two or three phase switching device or control device is provided, the system comprising two or three systems according to the first aspect, one of the two or three systems being directed to a circuit breaker of each of the two or three phases.

[0020] In a third aspect, a method for monitoring a circuit breaker is provided, the method comprising:

[0021] a) utilizing at least one sensor positioned to obtain at least one time series sensor data of at least one first part of the operating circuit breaker;

[0022] b) providing the at least one time series sensor of the at least one first part of the operating circuit breaker to a processing unit; and

[0023] c) determining, by the processing unit, whether at least one second part of the operating circuit breaker is operating correctly or has a fault, wherein the determining comprises: an analysis of the at least one time series sensor of the at least one first part of the operating circuit breaker by a trained neural network implemented by the processing unit.

[0024] In a fourth aspect, a system for training a neural network for monitoring a circuit breaker is provided, the system comprising:

[0025] - at least one first sensor;

[0026] - at least one second sensor; and

[0027] - a processing unit.

[0028] The at least one first sensor is configured to be positioned and utilized to obtain at least one time series sensor data of at least one first part of a calibration circuit breaker. The at least one first sensor is configured to provide the at least one time series sensor data of the at least one first part of the calibration circuit breaker to the processing unit. The at least one second sensor is configured to be positioned and utilized to obtain at least one time series sensor of at least one second part of the calibration circuit breaker. The at least one second sensor is configured to provide the at least one time series sensor of the at least one second part of the calibration circuit breaker to the processing unit. The processing unit is configured to train the neural network. The training of the neural network comprises: a utilization of the at least one time series sensor data of the at least one first part of the calibration circuit breaker and the at least one sensor data of the at least one second part of the calibration circuit breaker. The trained neural network is configured to determine, on the basis of an analysis of the at least one time series sensor data of the at least one first part of the operating circuit breaker, whether at least one second part of the operating circuit breaker is operating correctly or has a fault, for a calibration circuit breaker of the same type or model as the operating circuit breaker.

[0029] In a fifth aspect, a method for training a neural network for monitoring a circuit breaker is provided, the method comprising:

[0030] a1 ) utilizing at least one first sensor positioned to obtain at least one time series sensor data of at least one first part of a calibration circuit breaker;

[0031] b1 ) providing the at least one time series sensor of the at least one first part of the calibration circuit breaker to a processing unit;

[0032] c1 ) utilizing at least one second sensor positioned to obtain at least one time series sensor data of at least one second part of the calibration circuit breaker;

[0033] d1 ) providing the at least one time series sensor of the at least one second part of the calibration circuit breaker to the processing unit; and

[0034] e1 ) training, by the processing unit, a neural network, wherein the training of the neural network comprises utilizing the at least one time series sensor data of the at least one first part of the calibration circuit breaker and the at least one sensor data of the at least one second part of the calibration circuit breaker, and wherein the trained neural network is configured to determine, on the basis of analyzing at least one time series sensor data of at least one first part of an operating circuit breaker, whether the at least one second part of the operating circuit breaker is operating correctly or the at least one second part of the operating circuit breaker has a fault, and wherein the calibration circuit breaker is of the same type or model as the operating circuit breaker.

[0035] The above aspects and examples become apparent from and are elucidated with reference to the embodiments described hereinafter. BRIEF DESCRIPTION OF DRAWINGS

[0036] The exemplary embodiments are described below with reference to the following drawings:

[0037] Figure 1 A schematic diagram of an exemplary system for monitoring a circuit breaker is shown;

[0038] Figure 2 A schematic diagram of an exemplary system for monitoring a two- or three-phase switching device or control device is shown;

[0039] Figure 3 is a method for monitoring a circuit breaker;

[0040] Figure 4 A schematic diagram of an exemplary system for training a neural network for monitoring a circuit breaker is shown;

[0041] Figure 5A method for training a neural network for monitoring a circuit breaker is shown; and

[0042] Figure 6 A workflow of training a neural network is shown. DETAILED DESCRIPTION

[0043] Figure 1 An example of a system 10 for monitoring a circuit breaker is shown. The system comprises at least one sensor 20 and a processing unit 30. The at least one sensor is configured to be positioned and utilized to obtain at least one time series sensor data of at least one first part of an operating circuit breaker 40. The at least one sensor is configured to provide the at least one time series sensor data of the at least one first part of the operating circuit breaker to the processing unit. The processing unit is configured to determine whether at least one second part of the operating circuit breaker is operating correctly or the at least one second part of the operating circuit breaker has a fault. Determining correct operation or faulty operation comprises analysis of the at least one time series sensor data of the at least one first part of the operating circuit breaker by a trained neural network implemented by the processing unit.

[0044] For example, there can be one time series data from one sensor or two time series data taken in parallel in time from two sensors, etc. in this document.

[0045] In an example, the at least one first part of the operating circuit breaker is a main shaft of the circuit breaker.

[0046] In an example, the at least one first sensor for obtaining the at least one time series sensor of the at least one first part of the operating circuit breaker comprises one or more of: an acceleration sensor, a main shaft angle sensor.

[0047] In an example, the at least one second part of the operating circuit breaker is a movable contact and / or a push rod of the circuit breaker.

[0048] According to an example, the neural network is trained on the basis of at least one time series sensor data of at least one first part of a calibrated circuit breaker. The data is obtained when the at least first part of the calibrated circuit breaker is operating correctly. The neural network is also trained on the basis of at least one time series sensor data of at least one second part of the calibrated circuit breaker. The data is obtained when the at least one second part of the calibrated circuit breaker is operating correctly. The calibrated circuit breaker is of the same type or model as the operating circuit breaker.

[0049] For example, there can be one time series data from one sensor or two time series data acquired in parallel in time from two sensors, etc. there can also be several examples of such data acquired over a time period of several operations of the calibration circuit breaker to build a database of training data, where the calibration circuit breaker is operating correctly to train the neural network. Only the "healthy" data is used to train the neural network, where the circuit breaker is operating correctly.

[0050] In an example, the at least one first part of the calibration circuit breaker is a main shaft of the circuit breaker.

[0051] In an example, the at least one first sensor of the at least one time series sensor utilized to obtain the at least one time series sensor data of the at least one first part of the calibration circuit breaker comprises one or more of: an acceleration sensor, a main shaft angle sensor.

[0052] In an example, the at least one second part of the calibration circuit breaker is a movable contact and / or a push rod of the circuit breaker.

[0053] In an example, the at least one second sensor 70 of the at least one time series sensor utilized to obtain the at least one time series sensor data of the at least one second part of the calibration circuit breaker comprises one or more of: a position sensor, a speed sensor.

[0054] According to an example, the at least one time series sensor data of the at least one first part of the calibration circuit breaker is obtained simultaneously with the at least one time series sensor data of the at least one second part of the calibration circuit breaker.

[0055] According to an example, the at least one sensor 20, 60 utilized to obtain the at least one time series sensor data of the at least one first part of the calibration circuit breaker is of the same type or model as the at least one sensor 20 utilized to obtain the at least one time series sensor of the at least one first part of the operating circuit breaker.

[0056] According to an example, the at least one sensor utilized to obtain the at least one time series sensor data of the at least one first part of the calibration circuit breaker is located at the same or equivalent at least one location as the at least one sensor utilized to obtain the at least one time series sensor of the at least one first part of the operating circuit breaker.

[0057] According to an example, the trained neural network is configured to determine at least one synthetic time series sensor data for the at least one second portion of the operating circuit breaker. The determination of the synthetic data comprises an analysis of the at least one time series sensor data for the at least one first portion of the operating circuit breaker by the trained neural network. Then, the determination of whether the at least one second portion of the operating circuit breaker is operating correctly or the at least one second portion of the operating circuit breaker has a fault can comprise a comparison of the at least one synthetic time series sensor data for the at least one second portion of the operating circuit breaker with the baseline saved data.

[0058] According to an example, the baseline saved data comprises at least one time series sensor data for the at least one second portion of the circuit breaker, wherein the at least one second portion of the circuit breaker is operating correctly.

[0059] According to an example, the baseline saved data comprises at least some of the at least one time series sensor data for the at least one second portion of the calibration circuit breaker used for training the neural network.

[0060] According to an example, the baseline saved data comprises at least one synthetic time series sensor data for the at least one second portion of the calibration circuit breaker generated by the neural network from at least some of the at least one time series sensor data for the at least one first portion of the calibration circuit breaker.

[0061] According to an example, the baseline saved data comprises at least one synthetic time series sensor data for the at least one second portion of the calibration circuit breaker generated by the neural network from at least some of the at least one time series sensor data for the at least one first portion of the calibration circuit breaker.

[0062] According to an example, the processing unit is configured to determine that the at least one second portion of the operating circuit breaker has a fault on the basis that a distance measure between the at least one synthetic time series sensor data for the at least one second portion of the operating circuit breaker and the baseline saved data is equal to or greater than a threshold value.

[0063] In an example, the distance measure comprises a root mean square error.

[0064] In an example, the processing unit is configured to determine that the at least one second part of the operating circuit breaker is operating correctly in case a distance measure between the at least one synthetic time series sensor data of the operating at least one second part of the circuit breaker and the baseline holding data is smaller than a threshold value.

[0065] According to an example, the system comprises an output unit configured to output information indicating that the at least one second part of the operating circuit breaker has a fault.

[0066] According to an example, the neural network comprises a sequence-to-sequence model.

[0067] Figure 2 An example of a system 100 for monitoring a two or three phase switching device or control device is shown. The system 100 comprises two or three systems 10 as described with respect to Figure 1 However, one processing unit can process sensor data from all sensors of each phase.

[0068] Figure 3 A method 200 for monitoring a circuit breaker in its basic steps is shown. The method comprises:

[0069] In a utilizing step 210, also referred to as step a), at least one sensor is utilized, the at least one sensor being positioned to obtain at least one time series sensor of at least one first part of an operating circuit breaker;

[0070] In a providing step 220, also referred to as step b), the at least one time series sensor of the at least one first part of the operating circuit breaker is provided to a processing unit; and

[0071] In a determining step 230, also referred to as step c), it is determined by the processing unit whether the at least one second part of the operating circuit breaker is operating correctly or whether the at least one second part of the operating circuit breaker has a fault, wherein the determining comprises analyzing the at least one time series sensor data of the at least one first part of the operating circuit breaker by a trained neural network implemented by the processing unit.

[0072] In an example, the neural network is trained on the basis of at least one time series sensor data of at least one first part of a calibration circuit breaker and at least one time series sensor data of at least one second part of the calibration circuit breaker, wherein the at least one first part of the calibration circuit breaker is operating correctly and wherein the at least second part of the calibration circuit breaker is operating. The calibration circuit breaker is of the same type or model as the operating circuit breaker.

[0073] In an example, the at least one timing sensor of the at least one first portion of the calibration circuit breaker is obtained at the same time as the at least one timing sensor of the at least one second portion of the calibration circuit breaker.

[0074] In an example, the at least one sensor used to obtain the at least one timing sensor data of the at least one first portion of the calibration circuit breaker is the same type or model as the at least one sensor used to obtain the at least one timing sensor data of the at least one first portion of the operating circuit breaker.

[0075] In an example, the at least one sensor used to obtain the at least one timing sensor data of the at least one first portion of the calibration circuit breaker is located at the same or equivalent at least one location as the at least one sensor used to obtain the at least one timing sensor data of the at least one first portion of the operating circuit breaker.

[0076] In an example, the method comprises determining, by the trained neural network, at least one synthetic timing sensor data of the at least one second portion of the operating circuit breaker. The determining comprises analysing, by the trained neural network, the at least one timing sensor data of the at least one first portion of the operating circuit breaker. The determining of whether the at least one second portion of the operating circuit breaker is operating correctly or the at least one second portion of the operating circuit breaker has a fault can then comprise comparing the at least one synthetic timing sensor data of the at least one second portion of the operating circuit breaker with the baseline saved data.

[0077] In an example, the baseline saved data comprises at least one timing sensor data of the at least one second portion of the circuit breaker, wherein the at least one second portion of the circuit breaker is operating correctly.

[0078] In an example, the baseline saved data comprises at least some of: the at least one timing sensor data of the at least one second portion of the calibration circuit breaker used in training the neural network, and / or the at least one synthetic timing sensor data of the at least one second portion of the calibration circuit breaker generated by the neural network from at least some of the at least one timing sensor data of the at least one first portion of the calibration circuit breaker.

[0079] In an example, the method comprises determining, by the processing unit, that the at least one second portion of the operating circuit breaker has a fault on the basis that a distance measure between the at least one synthetic timing sensor data of the at least one second portion of the operating circuit breaker and the baseline saved data is equal to or greater than a threshold value.

[0080] In an example, the distance measure comprises a root mean square error.

[0081] In an example, the method comprises determining, by the processing unit, that the at least one second part of the operating circuit breaker is operating correctly on the basis that a distance measure between the at least one synthetic time series sensor data of the at least one second part of the operating circuit breaker and the baseline holding data is equal to or greater than a threshold value.

[0082] The threshold values for determining a fault and determining correct operation can be the same or different. Thus, there can be a “black or white” of correct operation or fault, in which case remedial action needs to be taken and the circuit breaker is immediately stopped from operating if necessary. Or, there can be a “black or grey or white” in that the measure can be below one threshold value and the operation is correct or above another threshold value and the operation has a fault (and remedial action needs to be taken) or is between the two threshold values and the circuit breaker can continue to operate but can be more closely monitored or scheduled for service in the near future.

[0083] In an example, the method comprises outputting, by the outputting unit, information indicating that the at least one second part of the operating circuit breaker has a fault.

[0084] In an example, the neural network comprises a sequence-to-sequence model.

[0085] Figure 4 An example of a system 300 for training a neural network for monitoring a circuit breaker is shown. The system comprises at least one first sensor 20, 60, at least one second sensor 70, and a processing unit 30, 80. The at least one first sensor is configured to be positioned and to obtain at least one time series sensor data of at least one first part of a calibration circuit breaker 50. The at least one first sensor is configured to provide the at least one time series sensor data of the at least one first part of the calibration circuit breaker to the processing unit. The at least one second sensor is configured to be positioned and to obtain at least one time series sensor data of at least one second part of the calibration circuit breaker. The at least one second sensor is configured to provide the at least one time series sensor data of the at least one second part of the calibration circuit breaker to the processing unit. The processing unit is configured to train a neural network. Training the neural network comprises utilizing the at least one time series sensor data of the at least one first part of the calibration circuit breaker and the at least one time series sensor data of the at least one second part of the calibration circuit breaker. The trained neural network is then configured to determine, on the basis of an analysis of time series sensor data of at least one first part of an operating circuit breaker, whether the at least one second part of the operating circuit breaker is operating correctly or the at least one second part of the operating circuit breaker has a fault. The calibration circuit breaker is of the same type or model as the operating circuit breaker.

[0086] In an example, the at least one first part of the operating circuit breaker is a main shaft of the circuit breaker.

[0087] In an example, the at least one first sensor for obtaining at least one time series sensor data of at least one first portion of the operating circuit breaker comprises one or more of: an acceleration sensor, a main shaft angle sensor.

[0088] In an example, the at least one second portion of the calibration circuit breaker is a movable contact and / or a push rod of the circuit breaker.

[0089] In an example, the at least one first portion of the calibration circuit breaker is a main shaft of the circuit breaker.

[0090] In an example, the at least one first sensor for obtaining at least one time series sensor of at least one first portion of the calibration circuit breaker comprises one or more of: an acceleration sensor, a main shaft angle sensor.

[0091] In an example, the at least one second portion of the calibration circuit breaker is a movable contact and / or a push rod of the circuit breaker.

[0092] In an example, the at least one second sensor 70 for obtaining at least one time series sensor data of at least one second portion of the calibration circuit breaker comprises one or more of: a position sensor, a speed sensor.

[0093] In an example, the neural network is trained on at least one time series sensor data of at least one first portion of the calibration circuit breaker 50 when the at least one first portion of the calibration circuit breaker is operating correctly and at least one time series sensor data of at least one second portion of the calibration circuit breaker when the at least one second portion of the calibration circuit breaker is operating correctly.

[0094] In an example, the at least one time series sensor of at least one first portion of the calibration circuit breaker is obtained simultaneously with the at least one time series sensor of at least one second portion of the calibration circuit breaker.

[0095] In an example, the at least one sensor 20, 60 for obtaining at least one time series sensor data of at least one first portion of the calibration circuit breaker is of the same type or model as the at least one sensor 20 for obtaining at least one time series sensor of at least one first portion of the operating circuit breaker.

[0096] In an example, the at least one sensor utilized to obtain at least one time series sensor data of at least one first portion of the calibration circuit breaker is located at the same or equivalent at least one position as the at least one sensor used to obtain at least one time series sensor data of at least one first portion of the operating circuit breaker.

[0097] Figure 5 A method 400 for training a neural network for monitoring a circuit breaker is shown in its basic steps. The method comprises:

[0098] In a utilizing step 410, also referred to as step al), at least one first sensor is utilized, the at least one first sensor being positioned to obtain at least one time series sensor of at least one first portion of a calibration circuit breaker;

[0099] In a providing step 420, also referred to as step bl), the at least one time series sensor of at least one first portion of a calibration circuit breaker is provided to a processing unit;

[0100] In a utilizing step 430, also referred to as step cl), at least one second sensor is utilized, the at least one second sensor being positioned to obtain at least one time series sensor data of at least one second portion of a calibration circuit breaker;

[0101] In a providing step 440, also referred to as step dl), the at least one time series sensor data of at least one second portion of a calibration circuit breaker is provided to a processing unit; and

[0102] In a training step 450, also referred to as step el), a neural network is trained by the processing unit. Training the neural network comprises utilizing the at least one time series sensor data of at least one first portion of a calibration circuit breaker and the at least one sensor data of at least one second portion of a calibration circuit breaker. The trained neural network is then configured to determine, on the basis of an analysis of time series sensor data of at least one first portion of an operating circuit breaker, whether at least one second portion of the operating circuit breaker is operating correctly or whether at least one second portion of the operating circuit breaker has a fault. The calibration circuit breaker is of the same type or model as the operating circuit breaker.

[0103] In an example, the at least one first portion of the operating circuit breaker is a main shaft of the circuit breaker.

[0104] In an example, the at least one first sensor for obtaining the at least one time series sensor of the at least one first portion of the operating circuit breaker comprises one or more of: an acceleration sensor, a main shaft angle sensor.

[0105] In an example, the at least one second portion of the operating circuit breaker is a movable contact and / or a push rod of the circuit breaker.

[0106] In an example, the at least one first portion of the calibration circuit breaker is a main shaft of the circuit breaker.

[0107] In an example, the at least one first sensor for obtaining the at least one time series sensor of the at least one first portion of the calibration circuit breaker comprises one or more of: an acceleration sensor, a main shaft angle sensor.

[0108] In an example, the at least one second part of the calibration circuit breaker is a movable contact and / or a push rod of the circuit breaker.

[0109] In an example, the at least one second sensor 70 for obtaining the at least one timing sensor data of the at least one second part of the calibration circuit breaker comprises one or more of the following: a position sensor, a speed sensor.

[0110] In an example, the method comprises that the neural network is trained on the basis of the at least one timing sensor data of the at least one first part of the calibration circuit breaker when the at least one first part of the calibration circuit breaker is in correct operation and the at least one timing sensor data of the at least one second part of the calibration circuit breaker when the at least one second part of the calibration circuit breaker is in correct operation.

[0111] In an example, the method comprises that the at least one timing sensor data of the at least one first part of the calibration circuit breaker is obtained simultaneously with the at least one timing sensor data of the at least one second part of the calibration circuit breaker.

[0112] In an example, the at least one sensor for obtaining the at least one timing sensor data of the at least one first part of the calibration circuit breaker is of the same type or model as the at least one sensor for obtaining the at least one timing sensor data of the at least one first part of the operating circuit breaker.

[0113] In an example, the at least one sensor utilized for obtaining the at least one timing sensor data of the at least one first part of the calibration circuit breaker is located at the same or equivalent at least one position as the at least one sensor used for obtaining the at least one timing sensor data of the at least one first part of the operating circuit breaker.

[0114] The system for monitoring a circuit breaker, the system for monitoring a two- or three-phase switching device or control device, the method for monitoring a circuit breaker, the system for training a neural network for monitoring a circuit breaker and the method for training a neural network for monitoring a circuit breaker will now be described in particular detail, in which reference is made to Figure 6 .

[0115] The inventors realized that sequence-to-sequence (Seq2seq AI models) that have been used to translate one language into another language can be used in a completely different way to monitor circuit breakers, for example, as used in medium voltage switching devices or control devices. The inventors realized that when two sensor measurements are related, a properly trained Seq2seq model can be utilized in order to "translate" the information of one sensor into the information of the other sensor. Moreover, since these AI models are more accurate for input data that is similar to the data used in the training of them, and since healthy data is usually more readily available, such a seq2seq model can be trained using only or mainly healthy data to translate the output of one sensor into the output of the other sensor, which is more readily available than fault data required to train existing AI based device monitoring systems. Subsequently, one can expect a large "translation" error to be related to a problem in the asset since abnormal sensor information would lead to a larger "translation" error. A distance metric can be used to understand whether the deviation is large enough to raise an alarm.

[0116] This also means that, for example, circuit breakers, where a movable contact monitoring system is required, but is difficult to monitor directly in the field due to the required sensors and their required location, can make use of the new monitoring technology. A Seq2seq AI model can be trained based on sensor data acquired for a calibration circuit breaker, where these sensors directly measure the movement of the movable contact and / or the plunger. However, at the same time, other sensors are used to monitor the main shaft of the circuit breaker. Herein, it is much easier to monitor the main shaft of the circuit breaker in terms of the simplicity of the required sensors and the convenience of their location. The Seq2seq AI model is then trained to accurately translate the main shaft sensor data into the movable contact and / or plunger sensor data. For this data generation, a calibration circuit breaker is confirmed to be operating correctly.

[0117] The trained Seq2seq AI model can then be used in the field for an operating circuit breaker. Herein, only the sensors monitoring the main shaft are required as input to the Seq2seq AI model, whose output is actually synthetic sensor data that the movable contact and / or plunger sensors would have obtained. The synthetic sensor data can then be compared to the database sensor data (for the movable contact and / or plunger) of a healthy operating circuit breaker to determine whether the synthetic data matches the database data within limits to indicate whether the movable contact and / or plunger is operating correctly or has a fault.

[0118] The above relates to a movable contact / plunger of a circuit breaker with respect to one set of sensor data and the main shaft of the circuit breaker for other sensor data, but different parts of the circuit breaker can be monitored.

[0119] Accordingly, a new technique is provided to monitor mechanical and electrical faults, such as contact erosion of a circuit breaker.

[0120] Figure 6 An overview workflow of the technology described above is provided. In essence, the system records time-series data from at least two sensors. The data is fed to a seq2seq model that learns to transform the information of one sensor data to another sensor data during the training phase. At production, the system is used to continuously provide such transformation for data not used during training. If the transformation error exceeds a certain level, an alarm is triggered using some distance metric.

[0121] While the application has been illustrated and described in detail in the drawings and foregoing description, such illustration and description is to be considered illustrative or exemplary and not restrictive. The application is not limited to the disclosed embodiments. Other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practising the claimed application, from a study of the drawings, the disclosure, and the dependent claims.

Claims

1. A system (10) for monitoring a circuit breaker, the system comprising: - at least one first sensor; and - a processing unit; wherein the at least one first sensor is configured to be positioned and utilized to obtain at least one time series sensor data of at least one first part of an operating circuit breaker (40), wherein the at least one first part of the operating circuit breaker is a main shaft of the circuit breaker; wherein the at least one first sensor utilized to obtain the at least one time series sensor data of the at least one first part of the operating circuit breaker comprises one or more of: an acceleration sensor, a main shaft angle sensor; wherein the at least one first sensor is configured to provide the at least one time series sensor data of the at least one first part of the operating circuit breaker to the processing unit; wherein the processing unit is configured to determine whether at least one second part of the operating circuit breaker is operating correctly or whether the at least one second part of the operating circuit breaker has a fault, wherein the at least one second part of the operating circuit breaker is a movable contact and / or a push rod of the circuit breaker, characterized in that the determination comprises an analysis of the at least one time series sensor data of the at least one first part of the operating circuit breaker by a trained neural network implemented by the processing unit; wherein the neural network is trained on the basis of at least one time series sensor data of the at least one first part of a calibration circuit breaker (50), wherein the at least one first part of the calibration circuit breaker is operating correctly, and at least one time series sensor data of the at least one second part of the calibration circuit breaker, wherein the at least one second part of the calibration circuit breaker is operating correctly, and wherein the calibration circuit breaker is of the same type or model as the operating circuit breaker, and wherein the at least one first sensor utilized to obtain the at least one time series sensor data of the at least one first part of the calibration circuit breaker comprises one or more of: an acceleration sensor, a main shaft angle sensor, and wherein the at least one second part of the calibration circuit breaker is a movable contact and / or a push rod of the circuit breaker.

2. The system according to claim 1, wherein the at least one time series sensor of the at least one first part of the calibration circuit breaker is obtained simultaneously with the at least one time series sensor of the at least one second part of the calibration circuit breaker.

3. The system according to any one of claims 1 to 2, wherein at least one sensor utilized to obtain the at least one time series sensor data of the at least one first part of the calibration circuit breaker is of the same type or model as the at least one first sensor utilized to obtain the at least one time series sensor data of the at least one first part of the operating circuit breaker.

4. The system of any one of claims 1 to 2, wherein at least one sensor utilized to obtain the at least one time series sensor data of the at least one first portion of the calibration circuit breaker is located at the same or equivalent at least one location as the at least one sensor utilized to obtain the at least one time series sensor data of the at least one first portion of the operating circuit breaker.

5. The system of any one of claims 1-2, wherein the trained neural network is configured to determine at least one synthetic time-series sensor data of the at least one second portion of the operating circuit breaker, wherein the determining comprises: the analysis of the at least one time series sensor data of the at least one first portion of the operating circuit breaker by the trained neural network, and wherein the determination of whether the at least one second portion of the operating circuit breaker is operating correctly or the at least one second portion of the operating circuit breaker has a fault comprises a comparison of the at least one synthetic time series sensor data of the at least one second portion of the operating circuit breaker to baseline saved data.

6. The system of claim 5, wherein the baseline saved data comprises at least one time series sensor data of the at least one second portion of a circuit breaker, wherein the at least one second portion of the circuit breaker is operating correctly.

7. The system of claim 6, wherein the baseline saved data comprises at least some of the following data: the at least one time series sensor data of the at least one second portion of the calibration circuit breaker used in training the neural network, and / or at least one synthetic time series sensor data of the at least one second portion of the calibration circuit breaker generated by the neural network from at least some of the at least one time series sensor data of the at least one first portion of the calibration circuit breaker.

8. The system of claim 5, wherein the processing unit is configured to determine that the at least one second portion of the operating circuit breaker has a fault on the basis that a distance metric between the at least one synthetic time series sensor data of the at least one second portion of the operating circuit breaker and baseline saved data is equal to or greater than a threshold value.

9. The system of any one of claims 1 to 2, 6 to 8, wherein the system comprises an output unit configured to output information indicating that the at least one second portion of the operating circuit breaker has a fault.

10. The system of any one of claims 1 to 2, 6 to 8, wherein the neural network comprises a sequence-to-sequence model.

11. A system (100) for monitoring a two or three phase switching or control device, the system comprising two or three systems (10) according to any one of claims 1 to 10, one of the two or three systems (10) being for a circuit breaker of each of the two or three phases.

12. A method (200) for monitoring a circuit breaker, the method comprising: a) utilizing at least one first sensor positioned to obtain at least one time series sensor data of at least one first part of an operating circuit breaker, wherein the at least one first part of the operating circuit breaker is a main shaft of the circuit breaker, and wherein the at least one first sensor utilized to obtain the at least one time series sensor data of the at least one first part of the operating circuit breaker comprises one or more of: an acceleration sensor, a main shaft angle sensor; b) providing the at least one time series sensor data of the at least one first part of the operating circuit breaker to a processing unit; and c) determining, by the processing unit, whether at least one second part of the operating circuit breaker is operating correctly or whether the at least one second part of the operating circuit breaker has a fault, wherein the at least one second part of the operating circuit breaker is a movable contact and / or a push rod of the circuit breaker, characterized in that the determining comprises analyzing, by a trained neural network implemented by the processing unit, the at least one time series sensor data of the at least one first part of the operating circuit breaker, wherein the neural network is trained on the basis of at least one time series sensor data of the at least one first part of a calibration circuit breaker (50) and at least one time series sensor data of the at least one second part of the calibration circuit breaker, wherein the at least one first part of the calibration circuit breaker is operating correctly, wherein the at least one second part of the calibration circuit breaker is operating correctly, and wherein the calibration circuit breaker is of the same type or model as the operating circuit breaker, and wherein the at least one first sensor utilized to obtain the at least one time series sensor data of the at least one first part of the calibration circuit breaker comprises one or more of: an acceleration sensor, a main shaft angle sensor, and wherein the at least one second part of the calibration circuit breaker is a movable contact and / or a push rod of the circuit breaker.

13. A system (300) for training a neural network for monitoring a circuit breaker, the system comprising: - at least one first sensor; - at least one second sensor (70); and - a processing unit; wherein the at least one first sensor is configured to be positioned and utilized to obtain at least one time series sensor data of at least one first part of a calibration circuit breaker (50), wherein the at least one first part of the calibration circuit breaker is a main shaft of the circuit breaker, and wherein the at least one first sensor utilized to obtain the at least one time series sensor data of the at least one first part of the calibration circuit breaker comprises one or more of: an acceleration sensor, a main shaft angle sensor; wherein the at least one first sensor is configured to provide the at least one time series sensor data of the at least one first part of the calibration circuit breaker to the processing unit; wherein the at least one second sensor is configured to be positioned and utilized to obtain at least one time series sensor data of at least one second part of the calibration circuit breaker, wherein the at least one second part of the calibration circuit breaker is a movable contact and / or a push rod of the circuit breaker, and wherein the at least one second sensor utilized to obtain the at least one time series sensor data of the at least one second part of the calibration circuit breaker comprises one or more of: a position sensor, a speed sensor; wherein the at least one second sensor is configured to provide the at least one time series sensor data of the at least one second part of the calibration circuit breaker to the processing unit; and characterized in that the processing unit is configured to train a neural network, wherein the training of the neural network comprises utilization of the at least one time series sensor data of the at least one first part of the calibration circuit breaker and the at least one sensor data of the at least one second part of the calibration circuit breaker, and wherein the trained neural network is configured to determine, on the basis of an analysis of at least one time series sensor data of the at least one first part of an operating circuit breaker, whether the at least one second part of the operating circuit breaker is correctly operating or whether the at least one second part of the operating circuit breaker has a fault, and wherein the calibration circuit breaker is of the same type or model as the operating circuit breaker, and wherein the neural network is trained on the basis of the at least one time series sensor of the at least one first part of the calibration circuit breaker when the at least one first part of the calibration circuit breaker is correctly operating, and the neural network is trained on the basis of the at least one time series sensor of the at least one second part of the calibration circuit breaker when the at least one second part of the calibration circuit breaker is correctly operating.

14. A method (400) for training a neural network for monitoring a circuit breaker, the method comprising: al) utilizing at least one first sensor positioned to obtain at least one time series sensor data of at least one first part of a calibration circuit breaker, wherein the at least one first part of the calibration circuit breaker is a main shaft of the circuit breaker, and wherein the at least one first sensor utilized to obtain the at least one time series sensor data of the at least one first part of the calibration circuit breaker comprises one or more of: an acceleration sensor, a main shaft angle sensor; bl) providing the at least one time series sensor data of the at least one first part of the calibration circuit breaker to a processing unit; c1) utilizing at least one second sensor, said at least one second sensor being positioned to obtain at least one time series sensor data of at least one second part of said calibration circuit breaker, wherein said at least one second part of said calibration circuit breaker is a movable contact and / or a push rod of said circuit breaker, and wherein said at least one second sensor utilized to obtain said at least one time series sensor data of said at least one second part of said calibration circuit breaker comprises one or more of: a position sensor, a speed sensor; d1) providing said at least one time series sensor data of said at least one second part of said calibration circuit breaker to said processing unit; and e1) characterized in that said method comprises training a neural network by said processing unit, wherein said training of said neural network comprises utilizing said at least one time series sensor data of said at least one first part of said calibration circuit breaker and said at least one sensor data of said at least one second part of said calibration circuit breaker, and wherein said trained neural network is configured to determine, on the basis of an analysis of at least one time series sensor data of said at least one first part of an operating circuit breaker, whether said at least one second part of said operating circuit breaker is operating correctly or said at least one second part of said operating circuit breaker has a fault, and wherein said calibration circuit breaker is of the same type or model as said operating circuit breaker, and wherein said neural network is trained on the basis of said at least one time series sensor of said at least one first part of said calibration circuit breaker when said at least one first part of said calibration circuit breaker is operating correctly, and said neural network is trained on the basis of said at least one time series sensor of said at least one first part of said calibration circuit breaker when said at least one second part of said calibration circuit breaker is operating correctly.

Images