Method and device for predicting future faults in an electric power system

Abstract

The present disclosure relates to a computer-implemented method (100) for predicting an impending conductor fault that causes a fault within a component of a power system. The method includes steps of obtaining data from components of the power system (101). Further, it includes steps of extracting features from anomalies in the data (103) and identifying patterns (104). Further, it includes a step of classifying anomalies based on the patterns so as to sort the anomalies into classes (105). Further, it includes a step of identifying anomalies (106) and a step of providing a prediction indicating a point in time at which at least one impending anomaly of a particular class will occur (107). Further, the method (100) includes a step of determining the likelihood that at least one impending anomaly will cause a fault within the power system (108) and a step of providing prediction information accessible to a user (109).

Description

Technical Field

[0001] The present disclosure relates to a method and an electronic device for predicting an impending conductor fault that causes a fault within a component of a power system.

Background Art

[0002] Generally, power systems such as power grids operate in a reactive manner. Therefore, operators and users of such systems would benefit by being able to operate in a proactive rather than a reactive approach.

[0003] Therefore, in most existing systems, users and operators can only notice faults that have already occurred. That is, most existing solutions only record what has already occurred. Therefore, the data collected is accumulated without being properly utilized. By being able to appropriately predict future faults in a power system or its components, significant savings in downtime and wasted energy can be achieved.

[0004] Although there is some prior art directed to identifying anomalies within a power system, there is none directed to predicting an impending fault that causes a fault within a power system. By being able to predict when an impending fault that causes a fault will occur, it is possible to plan maintenance appropriately and use the system at least until just before the predicted fault occurs.

[0005] Based on the above, there is room for improvement in the art for having a method and a device that can predict an impending fault that causes a fault within a power system.

[0006] Accordingly, there is room in the art to explore methods and devices for predicting upcoming conductor failures that cause faults in components of a power system. Specifically, the methods and devices should provide accurate predictions.

[0007] Among currently known solutions, some function well depending on the situation, but it is desirable to provide methods and devices that specifically meet the requirements regarding prediction accuracy. SUMMARY OF THE INVENTION

[0008] Accordingly, an object of the present disclosure is to provide an electronic device and a method for reducing, alleviating, or eliminating one or more of the deficiencies and drawbacks identified above.

[0009] This object is achieved by a method, a computer-readable storage medium, and an electronic device as defined within the scope of the appended claims.

[0010] The present disclosure is at least partially based on the insight that by providing improved methods, devices, and computer-readable storage media, upcoming conductor failures that cause faults within a power system can be accurately predicted, and thus their users / operators can act timely, for example through maintenance, thereby reducing the cost and downtime of the system and / or increasing the lifespan and energy efficiency of the system hardware.

[0011] The present disclosure provides a computer-implemented method for predicting an impending conductor fault that causes or may cause a fault within a component of a power system in the future. The method includes obtaining data (e.g., current data and historical data) from a component of the power system, the data including at least one of a current signal and a voltage signal. Further, the method includes detecting an (a plurality of) anomaly in the data, and extracting features from each anomaly. Further, the method includes identifying a pattern (e.g., occurrence, magnitude, or any other pattern) among the (a plurality of) anomalies, and classifying the anomalies based on the pattern so as to sort each anomaly into a class (in some aspects, the pattern may be identified after classification). Further, the method includes identifying an anomaly correlated with a conductor fault that causes a fault based on the extracted features. Further, the method includes providing a prediction indicating a point in time when at least one impending anomaly of a particular class will occur, and determining, based on the prediction, the likelihood that at least one impending anomaly will cause a fault within the power system. Additionally, the method includes providing information of the prediction for access by a user interface. In some aspects, the prediction may be accessible when the likelihood is higher than a pre-determined threshold value / in response to the likelihood being higher than a pre-determined threshold value. The threshold value may be a specific importance threshold or a specific likelihood ratio.

[0012] An advantage of the method is to provide an accurate prediction as to when a fault may occur. Thus, the method can efficiently and accurately predict which impending anomalies will cause a fault within the system. Thus, the maintenance of the system can be timely planned.

[0013] Furthermore, there is an advantage in the system that patterns are (individually) detected among the identified anomalies. Thus, other data that are not anomalies can be overlooked in the step of identifying patterns. As a result, there is a possibility that more efficient forecasting as a large amount of data can be ignored. In addition, by identifying specific anomalies that can cause a failure in the system, it is possible to more easily apply corrective measures when a specific harmful anomaly is identified.

[0014] The term "predicting" herein may be replaced with "forecasting". The term "conductor" may refer to a wire, cable, switch, or any other conductive element connected to a power system or a part thereof. Thus, the term "conductor" may refer to any conductive element of a power system that can cause the power system to fail at a future time. A conductor fault may refer to a physical defect or any other type of fault that affects the conductor.

[0015] The phrase "cause a failure" may refer to the fact that a fault can cause a failure in the power system at a future time.

[0016] In some aspects herein, the method may relate to a method for predicting an impending fault that causes a failure within a component of a power system. The fault can be any fault.

[0017] Data can be acquired in snapshots, and further data can be acquired from relays and PQ meters or any other suitable electrical measurement devices.

[0018] The components of the power system herein include transmission and distribution components and conductors.

[0019] The classifications can be short circuits, imbalances, ground faults, and cable faults.

[0020] The method can utilize at least one trained machine learning (ML) algorithm, preferably, in the steps of providing classification, anomaly identification, and prediction, at least one of a naive Bayes classifier, support vector machine (SVM) linear regression, logistic regression, artificial neural network (ANN), decision tree, random forest, K-nearest neighbor (KNN), and K-means clustering. Note that the ANN can include (i) a multi-layer perceptron (MLP), (ii) a recurrent neural network including long short-term memory and gated recurrent units, (iii) a convolutional neural network, and any combination thereof. Thus, each step of the method can be performed by utilizing a trained machine learning algorithm.

[0021] In the aforementioned method steps, the advantage of utilizing a trained machine learning algorithm is that more accurate predictions can be provided. The machine learning algorithm can predict how an anomaly develops into a failure over time.

[0022] The ML algorithm identifies anomalies based on a specific snapshot of data, where the ML algorithm can identify irregular events within the data. An anomaly can be defined based on a threshold, i.e., in the aspects of this specification, an anomaly can be identified based on the likelihood that an irregular event is an anomaly. Thus, the threshold can be set such that an event cannot be identified as an anomaly if a potential anomaly / irregular event has a likelihood below the threshold. The threshold can be set by the user or the ML algorithm.

[0023] Thus, the trained machine learning algorithm can not only predict future anomalies based on the extracted features and the trained learning data, but also predict when / which anomalies will cause a failure. However, for prediction, the anomalies correlated with failures are the subject.

[0024] The extracted features may include at least one of harmonicity, frequency deviation, phase shift, jump, amplitude, occurrence time, root mean square (RMS), duration, admittance, resistance, inductance, impedance, active power and reactive power, normalized voltage signal, and normalized current signal.

[0025] The classification step may further include, preferably all, one of determining the fault direction of each anomaly based on the measurement position associated with the anomaly. Further, the classification step may include estimating the distance of the anomaly based on the measurement position, and determining the position of the anomaly in the power system based on the estimated fault direction and distance.

[0026] By obtaining the fault direction, the distance and the position of the anomaly, the method can advantageously ignore / filter out / prioritize irrelevant / related anomalies in the step of identifying the anomaly, i.e., anomalies that may be caused at positions not correlated with the faulty conductor causing the fault or their electrical component systems can be ignored. Also, by determining the position, the maintenance of the system is simplified based on the fact that the user / operator can derive the position of the anomaly.

[0027] In some aspects herein, the position is determined based on the known grid topology or the estimated grid topology of the power system. Thus, faster position estimation is enabled. The position estimation can be performed based on rules or on a trained machine learning algorithm.

[0028] In some embodiments, the method may include determining a root cause of an identified anomaly correlated with a conductor fault, the root cause being determined based on a pattern or based on a pattern and a class. In some embodiments, the pattern may be compared to a historical pattern that causes a fault within a system where the cause of the fault is known (e.g., by a trained ML algorithm). The root cause may be determined based on the historical pattern. For example, the root cause may be environmental conditions (lightning, overgrown plants, or any other environmental condition) or a hardware defect. The root cause may be provided as information for the user in a method step that provides information.

[0029] In other embodiments herein (e.g., by an ML algorithm), the method may further include providing (to the user) data of at least one feature associated with each identified anomaly correlated with a conductor fault. The data may be, for example, at least one of the type, magnitude, occurrence, or any other type of data of the feature.

[0030] The step of providing information may include providing a ranking that indicates the scope of each anomaly contributing to the prediction. The ranking may be derived by a machine learning algorithm in the step of providing the prediction.

[0031] The features may be harmony, frequency deviation, phase shift, jump, amplitude, occurrence time, root mean square (RMS), duration, impedance, active power and reactive power, normalized voltage signal, normalized current signal, or any other feature.

[0032] An advantage of this is that it may provide a countermeasure that allows the user to suppress the anomaly without necessarily resolving the root cause.

[0033] The step of providing information may include an estimate of when the outage will occur, data on the anomaly causing the fault, and the location of the fault within the power system.

[0034] Furthermore, the prediction can be performed for prediction ranges of milliseconds, seconds, hours, within one month, or two to three months or three to six months.

[0035] Therefore, there is an advantage in that long-term prediction becomes possible and timely maintenance can be provided.

[0036] In some aspects herein, the method is directed to a method for predicting future cable faults that cause faults within components of a power system, specifically, cable faults for transmission and distribution.

[0037] In other aspects herein, the method is directed to a method for predicting future wire faults within components of a power system, where the wire faults can be winding wire faults, such as stator winding wire faults, preferably winding insulation faults. The method can predict future wire faults caused by degradation of insulating materials (e.g., coating materials covering conductors) in a power system or its components. Degradation of the insulating materials can occur due to electrical stress caused by switching surges that place the insulation of power cables / wires under high electrical stress.

[0038] Such data can be acquired by an electrical measurement device aimed at measuring anomalies and provided as a waveform input used to identify anomalies for use in the methods herein. Since the filtering can be machine learning-based, the machine learning component is aimed at filtering the correct amount of noise based on a trained learning algorithm that memorizes patterns from previous training data.

[0039] Therefore, in other aspects herein, the method can be aimed at predicting future cable faults that cause faults within components of a power system, such as causing a fault within a component of the power grid. The cables can be transmission and distribution cables.

[0040] It should be noted that the method steps herein can be executed in different orders and can be varied within the knowledge of those skilled in the art, and thus are not restricted to the order as disclosed herein. Furthermore, the method steps herein (specifically, the steps of identifying a pattern, classifying into a class, and identifying an anomaly correlated with a conductor fault that causes a failure) can be executed in parallel.

[0041] A computer-readable storage medium storing one or more programs configured to be executed by one or more control circuit elements of an electronic device is also provided, and the one or more programs include instructions for executing a method according to any aspect herein.

[0042] Furthermore, an electronic device is provided that includes one or more control circuit elements and a memory device storing one or more programs configured to be executed by the one or more control circuit elements, and the one or more programs include instructions for executing the method herein. In some aspects, the electronic device can be a power system.

[0043] Hereinafter, the present disclosure will be described non-limitingly and in more detail with reference to the exemplary embodiments shown in the accompanying drawings.

Brief Description of the Drawings

[0044]

Fig. 1

Fig. 2

Fig. 3

Fig. 4

Fig. 5A

Fig. 5B

Fig. 6

Fig. 7

Fig. 8

Fig. 9

Mode for Carrying Out the Invention

[0045] In the following detailed description, some aspects of the present disclosure will be described. However, it should be understood that the features of different aspects are interchangeable between embodiments and can be combined in different ways, unless otherwise specified. In the following description, in order to provide a more complete understanding of the devices and methods provided, it will be apparent to those skilled in the art that the devices and methods can be implemented without these details, even though a number of specific details are shown. In other cases, well-known structures or functions will not be described in detail so as not to obscure the present disclosure.

[0046] In the following description of the exemplary embodiments, the same reference numerals denote the same or similar components.

[0047] Figure 1 schematically shows, in the form of a flowchart, a method 100 for predicting an impending conductor fault that causes a fault within a component of a power system. The method 100 includes a step (101) of obtaining data from a component of the power system, where the data includes at least one of a current signal and a voltage signal. The data can be obtained continuously. Further, the method includes detecting an anomaly in the data (102) and extracting features from the anomaly (103). The components of the power system can be cables, generators, relays, insulation, overhead lines.

[0048] Further, the method includes identifying a pattern among the anomalies (104) and classifying the anomalies based on the pattern so as to sort the anomalies into classes (105). Further, the method includes identifying the anomalies based on the extracted features that correlate with a conductor fault that causes a fault (106). Thus, in step 104, each of the anomalies can be compared to identify a pattern among the anomalies.

[0049] In addition, the method 100 provides a prediction (107) indicating the point in time at which at least one of the impending anomalies of a particular class (one of the classes mentioned) will occur. Further, the method 100 is directed to determining (108) the likelihood that at least one of the impending anomalies will cause a fault within the power system based on the prediction, and providing (109) information about the prediction accessible to the user if the likelihood is above a pre-determined threshold.

[0050] Figure 1 further shows that prior to step 102 of detecting an anomaly, the method 100 can include a step 101' of filtering frequency components above and / or below a particular frequency range, for example above / below a particular amplitude. This can be performed by using a frequency filtering device, preferably a band filter. The term "power system" as used herein can refer to any network of electrical components used for supplying (generating), transmitting, distributing, and consuming electrical power.

[0051] FIG. 2 shows an example of a current signal obtained from a component of a power system. Further, reference numerals 21 and 22 indicate portions within the graph that include anomalies. Thus, an anomaly can be a deviation from a general pattern.

[0052] Thus, method 100 can identify patterns within anomalies (e.g., anomalies 21, 22 in FIG. 2). Method 100 can classify anomalies based on the patterns. The classes can be short circuits, imbalances, ground faults (e.g., transient current, low ohm, high ohm, intermittent), cable faults, or any other suitable classification. Other classifications can be sag, swell, transient current, interruption, harmonics, frequency deviation, flicker. In some aspects, the classification can be a binary classification. The classification can be performed by a trained learning algorithm based on previous training.

[0053] Note that step 102 is directed to detecting the anomaly itself. On the other hand, step 106 is directed to identifying anomalies that correlate with an impending fault. Thus, method step 106 can be based on the extracted features, but in some aspects herein, can also be based on the particular class to which the anomalies are clustered. In some aspects herein, the phrase "impending fault" can refer to an impending fault within a particular time period or within any future time period.

[0054] The prediction step 107 of method 100 can predict how an anomaly will develop over time (e.g., time periods from milliseconds, seconds, one hour to three months), and the point in time at which an impending anomaly will cause a fault. The prediction can include the number, amplitude, and time of occurrence of anomalies in future cycles. Thus, the prediction can include an estimation of the remaining life of the electrical components of the power system.

[0055] In particular, it should be noted that the steps of classification 105, identification 106, and provision 109 may utilize at least one trained machine learning algorithm. The machine learning algorithm may include at least one of a naive Bayes classifier, a support vector machine SVM, linear regression, logistic regression, an artificial neural network ANN, a decision tree, a random forest, a k-nearest neighbor KNN, and k-means clustering.

[0056] Furthermore, the extracted features may include at least one of harmonicity, frequency deviation, phase shift, jump, amplitude, occurrence time, root mean square (RMS), duration, impedance, active power and reactive power, and normalized voltage and current signals.

[0057] Figure 3 schematically shows a part of the grid topology 100 of the power system in an exemplary manner. In particular, the classification step 104 (schematically shown in Figure 1) determines the fault directions 31, 31' of each anomaly based on the (at least one) measurement location 30 associated with the anomaly (e.g., from which direction the anomaly occurs as seen from the measurement location 30), estimates the distance d1 of the anomaly based on the measurement location 30, determines the location 32 of the anomaly in the power system based on the estimation of the fault direction and distance, and may include.

[0058] Figure 3 shows a part of the lattice topology 300 that can determine the fault directions 31, 31' based on a specific (e.g., known) measurement position 30 of an anomaly. The fault direction can indicate from which direction 31, 31' the anomaly occurs. In Figure 3, the true fault direction is shown as 31'. Thus, based on the estimated fault direction and specific voltage and / or current values, the distance of the anomaly can be estimated, thereby enabling position estimation and deriving the position 32 of the anomaly. In this way, specific components of the power system that are at risk of failure can be identified. In other words, based on the above, the position 32 can be a position within the power system where a future conductor fault that causes a fault within a specific component of the power system / or the power system itself can occur. It should be noted that in some aspects herein, based on the classification step 105, the method can determine the type of electrical component from which the anomaly originates. Thus, determining the position can further be based on the determined electrical component. For example, the feature extraction step can be performed after classification. In other words, the method can determine the position 32 of the anomaly within the power system based on the fault direction, distance estimation, and the type of electrical component from which the anomaly originates, and the type of electrical component is determined within the classification step.

[0059] The wire fault can be a wire fault caused by the deterioration of the insulating material within the power system.

[0060] The information providing step 109 can include an estimate of when the outage will occur, data on the anomaly causing the fault, and the position of the fault within the power system. Further, in some aspects herein, the information is transmitted as a notification / warning to the user / operator of the power system. In some aspects, the information can be provided to be visually emphasized to the user / operator.

[0061] FIG. 4 shows an electronic device 1 comprising one or more control circuit elements 2 and a memory device 3 storing one or more programs configured to be executed by the one or more control circuit elements 2, the one or more programs including instructions for performing a method 100 of any aspect herein for predicting an impending conductor fault that causes a failure within a component of a power system (or the power system itself).

[0062] The at least one memory device 3 may comprise any form of volatile or non-volatile computer-readable memory including, but not limited to, persistent storage, solid-state memory, remotely attached memory, magnetic media, optical media, random access memory (RAM), read-only memory (ROM), mass storage media (such as a hard disk), removable storage media (such as a flash drive, compact disk (CD), or digital video disk (DVD)), and / or any other volatile or non-volatile, non-transitory device that stores usable information, data, and / or instructions, readable and / or computer-executable memory device.

[0063] The control circuit element 2 can be arranged to execute an instruction set in the memory device 3 to operate the method 100. The control circuit element 2 can be of any suitable type, such as a microprocessor, a digital signal processor (DSP), an ASIC (application specific integrated circuit), an FPGA (field programmable gate array), a graphics processing unit (GPU), or a combination thereof, or other similar processing means arranged to execute an instruction set. The computer-readable storage medium can be of a non-volatile and / or volatile type, and a temporary or non-temporary type, such as, for example, RAM, EEPROM, flash disk, etc. Note that the memory unit 3 can be integrated with the control circuit element 2. The control circuit element 2 can include any suitable type of input / output interface 5 and communication interface, such as an Ethernet, I2C bus, RS232, CAN bus, etc., a wireless communication technology such as IEEE802.11-based or cellular-based technology, or other communication protocols according to the application scope. The communication interface can be used to receive signals, software updates, and instruction messages. Further, the communication interface can be used to communicate results, messages, status reports, etc. to external devices and control units such as a control station or a server via a network, for example, via a public or private network.

[0064] Each memory device 3 may also store data that can be retrieved, manipulated, created, or stored by the control circuit element 2. The data may include, for example, local updates, parameters, training data, extracted features, anomalies, patterns, classes, and data of voltage and current. As shown in FIG. 4, the control circuit element may include a machine learning component / engine 4 arranged to store and process previous data, real-time data, prediction data (i.e., trained learning algorithms), trained learning algorithms (and / or models, components, data utilized within the trained learning algorithms). However, in some embodiments, the trained machine learning algorithms and the machine learning component 4 may be stored within a cloud computing device 8 accessible by the control circuit element 2. FIG. 4 shows that in some embodiments, the control circuit element 2 may include the machine learning component 4 and implement at least one trained learning algorithm based on data from the memory device 2. The data may also be stored within one or more databases or the cloud computing device 8. The one or more databases may be connectable to the electronic device 1 via a communication network.

[0065] FIG. 4 further shows that data 9 may be obtained from the electrical components 6 of the power system along the method step 101 as shown in FIG. 1. Further, FIG. 4 also shows that the user / operator 7 may receive information provided by the electronic device at their respective electronic devices. The information may be a prediction of when at least one upcoming anomaly of a particular class will occur, and the upcoming anomaly has a likelihood above a pre-determined threshold that causes a fault within the power system at a particular point in time. The particular point in time at which the fault will occur may also be provided by the electronic device 1.

[0066] Figures 5A through 5B and 6 through 9 illustrate the execution and aspects of method 100 and electronic device 1 disclosed herein. The purpose of Figures 5A through 8 is to further explain the disclosure presented herein and the advantages associated therewith. It should be noted that the simulation / test results are based on, but not limited to, the aspects for the purposes of the disclosure herein and may be varied within the present disclosure.

[0067] Figure 5A shows a graph demonstrating that method 100 and device 1 herein demonstrate the classification of anomalies with 99% accuracy (displayed by graph A). Graph B in Figure 5A shows the loss during the training of a trained machine learning algorithm. Further, Figure 5B shows a graph having a prioritization scheme indicating the range to which each anomaly contributes to the prediction. The anomaly I0 in Figure 5B is the anomaly that contributes up to the largest range to the prediction.

[0068] Similarly, Figure 6 shows a user interface for a user indicating expected explainability. The method and device herein may provide information to the user by the user interface. Figure 6 shows that each predictability A1 for the prediction of fault occurrences A2 and A3 shown in Figure 6 is shown in the user interface for the user. Figure 6 shows that the method and device herein may provide a prediction as shown in graph g1 in Figure 6. As shown in the figure, there are a plurality of anomalies A1 - A3 within the prediction. Further, Figure 6 exemplarily shows, as shown in graph g2, the anomaly that most contributes to the prediction of anomaly A2.

[0069] Figure 7 shows in graph form the importance of groups of prediction features based on corresponding phases. Thus, the method may provide data of at least one feature to which each identified anomaly correlated with a conductor fault relates, and a range of contribution in which at least one feature must have an anomaly. In Figure 7, feature 1 contributes most to the prediction. Figure 8 schematically shows over a certain time period a simulation of step 107 that provides a prediction as to when at least one upcoming anomaly of a particular class will occur.

[0070] Further, FIG. 8 also shows that information is provided in the form of a warning 109. Thus, more specifically, FIG. 8 shows the low output of the trained machine learning algorithm (see the dotted line in FIG. 8), and the binary prediction as a result of the decision step 108 in any aspect of the method 100 herein. Thus, the solid line indicates that a warning is provided based on the likelihood that at least one of the coming anomalies will cause a fault in the power system exceeding a threshold value that has been pre-determined based on the prediction of the trained machine learning algorithm. Thereby, as shown in FIG. 9, a warning is provided at a point in time prior to the actual fault.

[0071] FIG. 9 shows the method 100 being executed over a period of time in an exemplary manner, showing that the method can continuously perform feature extraction and classification based on data (historical and live data). Further, the method 100 can be based on the extraction and classification executing method steps 104, 106 - 109. The method can identify patterns based on a recent time window during the step of identifying patterns, and can also identify patterns based on periodic and regular trends obtained from historical data.

Claims

1. A computer implementation method (100) for predicting an upcoming fault in a conductor that will cause a failure within a component of a power system, (101) Obtaining data from the components of the power system, wherein the data includes at least one of a current signal and a voltage signal. Detecting anomalies in the aforementioned data (102), Extracting features from the aforementioned anomalies (103), Identifying patterns among the aforementioned abnormalities (104), Classifying the anomalies based on the pattern so as to sort the anomalies into classes (105), Based on the extracted features, identify anomalies that correlate with conductor failures causing malfunctions (106), (107) To provide a prediction indicating the time at which at least one of a specific class of upcoming anomalies will occur, Based on the above prediction, the likelihood of at least one of the incoming anomalies causing a failure in the power system is determined (108), Providing the user with information on the predictions (109), Computer implementation methods, including those mentioned above.

2. The method according to claim 1 (100), which utilizes at least one trained machine learning algorithm.

3. The method according to claim 2 (100), wherein the machine learning algorithm utilizes at least one of the following in the steps of classification (105), identification (106), and provisioning (109): a simple Bayes classifier, a support vector machine (SVM), linear regression, logistic regression, an artificial neural network (ANN), a decision tree, a random forest, a K nearest neighbor (KNN), and K mean clustering.

4. The method according to any one of claims 1 to 3 (100), wherein the extracted features include at least one of harmonicity, frequency deviation, phase shift and jump, amplitude, time of occurrence, root mean square RMS, duration, impedance, admittance, resistance, inductance, active power and reactive power, normalized voltage and current signals.

5. The aforementioned classification step (104) is, The direction of the malfunction for each anomaly is determined based on the measurement location associated with the aforementioned anomaly, Estimating the distance of the abnormality based on the aforementioned measurement position, Based on the estimation of the fault direction and distance, the location of the anomaly within the power system is determined. The method according to any one of claims 1 to 3, further comprising (100).

6. The method according to claim 5 (100), wherein the position is determined based on a known or estimated grid topology of the power system.

7. The method according to any one of claims 1 to 3 (100), wherein the conductor fault is a conductor fault caused by deterioration of the insulating material in the power system.

8. The method (100) according to any one of claims 1 to 3, wherein the step of providing information (109) includes an estimate of when the outage will occur, data of the anomaly causing the failure, and the location of the failure in the power system.

9. The method (100) according to any one of claims 1 to 3, wherein the prediction (107) is performed for a prediction range of within one month, or two to three months, or three to six months.

10. The method according to any one of claims 1 to 3 (100), comprising determining the root cause of the identified anomaly that correlates with a conductor fault causing a failure (108a), wherein the root cause is determined based on the pattern.

11. The method according to any one of claims 1 to 3 (100), wherein the information provision step includes providing data on at least one feature relating to each identified anomaly that correlates with a conductor failure.

12. The method according to claim 11 (100), wherein providing the aforementioned information (109) includes providing a priority scheme indicating the range of each anomaly that contributes to the prediction.

13. A computer-readable storage medium for storing one or more programs configured to be executed by one or more control circuit elements (2) of an electronic device (1), A computer-readable storage medium comprising one or more programs including instructions for performing the method (100) described in any one of claims 1 to 3.

14. An electronic device (1) comprising one or more control circuit elements (2), and a memory device (3) that stores one or more programs configured to be executed by the one or more control circuit elements (2), The electronic device (1) comprises one or more programs, each containing instructions for performing the method (100) described in any one of claims 1 to 3.

Images