WO2022064350A1

WO2022064350A1 - Electricity distribution network fault detection system and method

Info

Publication number: WO2022064350A1
Application number: PCT/IB2021/058569
Authority: WO
Inventors: Anwarul Islam SIFAT; Fiona J Stevens MCFADDEN
Original assignee: Victoria Link Limited
Priority date: 2020-09-22
Filing date: 2021-09-21
Publication date: 2022-03-31

Abstract

An electricity distribution network fault detector 122 is provided. The fault detector may comprise a signal processor 124 configured to perform signal processing on magnetic field data, the magnetic field data representing at least one magnetic field generated by at least one electricity distribution line 112; a data selector 126 configured to receive transformed magnetic field data from the signal processor 124 and to select a time window of the received data; and a classifier 128 configured to receive at least one time window from the data selector 126 and to apply a classification to the at least one time window, the classification selected from a group of classifications.

Description

ELECTRICITY DISTRIBUTION NETWORK FAULT DETECTION SYSTEM AND

METHOD

FIELD OF INVENTION

The invention relates to a fault detection and classification system, particularly for use in detecting and classifying faults in an electricity distribution network.

BACKGROUND

Power lines are a critical component of electricity infrastructure. It is therefore very important to make sure that power lines are maintained to avoid any critical failure and ensure a high level of safety.

Faults in electricity distribution networks are an unwanted phenomena, with the potential to ignite fires, cause electrocution and/or damage the system itself. High-current Low- Impedance Faults (LIF) are typically detected and protected against using, for example overcurrent, distance, directional relays, and fuse. Existing techniques are able to detect the presence of LIF faults somewhere in the network. However, pinpointing the exact location of such faults within the network is more challenging.

Equally hazardous High Impedance Faults (HIF) are more challenging to detect due to the fault current being much lower than load currents. HIF faults go undetected until they cause some kind of obvious effect, for example a fire, low-impedance fault, or downed conductor. The detection of HIF is therefore still an ongoing distribution industry and research challenge.

Current Transformers (CTs) are used to monitor current just at the substation or key switching points. One drawback of a CT is the saturation of its magnetic core, resulting in a current chopping effect at the secondary coil. Furthermore, CTs are cost-prohibitive to implement for more widespread monitoring, for example beyond just the substation or key points.

It is an object of at least preferred embodiments to address at least some of the aforementioned disadvantages. An additional or alternative object is to at least provide the public with a useful choice. SUMMARY OF THE INVENTION

The disclosure relates to an electricity distribution network fault detector comprising a signal processor configured to perform signal processing on magnetic field data, the magnetic field data representing at least one magnetic field generated by at least one electricity distribution line; a data selector configured to receive transformed magnetic field data from the signal processor and to select a time window of the received data; and a classifier configured to receive at least one time window from the data selector and to apply a classification to the at least one time window, the classification selected from a group of classifications.

The term 'comprising' as used in this specification means 'consisting at least in part of. When interpreting each statement in this specification that includes the term 'comprising', features other than that or those prefaced by the term may also be present. Related terms such as 'comprise' and 'comprises' are to be interpreted in the same manner.

The group of classifications may include one or more of the group comprising: normal system state (NSS), low impedance fault (LIF), HIF high current arcing (HHCA), HIF low current (HLC).

The classifier may comprise a deep neural network (DNN) classifier block

The fault detector may further comprise a data collector configured to generate the magnetic field data representing the at least one magnetic field generated by the at least one electricity distribution line.

The data collector may comprise two Giant Magneto-Resistive (GMR) sensor heads mounted on a power pole.

The fault detector may further comprise an analog to digital converter (ADC) configured to receive the output from the data collector. The signal processor may be configured to receive measurement signals from the ADC.

The fault detector may further comprise a current estimation block configured to receive measurement signals from the ADC.

The fault detector may further comprise a current estimation block configured to receive instantaneous current output from the current estimation block. The signal processor may be configured to receive input from the current estimation block. The signal processor may be configured to output a dataset augmented with timefrequency analysed signals.

The signal processor may include a Hilbert transform (NT) block configured to calculate instantaneous amplitude (IA) and instantaneous phase (IP) angle on input signals.

The signal processor may include a discrete wavelet transform (DWT) block configured to decompose at least one input signal into plurality of frequency bands.

The signal processor may include a high pass filter (HPF) configured to remove at least some frequency components from the input signal to the DWT block.

The data selector may be configured to receive magnetic field data from the data collector and to receive the transformed magnetic field data from the signal processor.

A method of detecting faults in an electricity distribution network may comprise receiving magnetic field data representing at least one magnetic field generated by at least one electricity distribution line in the electricity distribution network; performing signal processing on the received magnetic field data; selecting at least one time window of the transformed magnetic field data obtained from the signal processing; and applying a classification to the at least one time window, the classification selected from a group of classifications.

The method may further comprise selecting the at least one time window from the transformed magnetic field data obtained from the signal processing and from magnetic field data obtained from a data collector.

A computer-readable medium may have stored thereon computer-executable instructions that, when executed by a processor, cause the processor to perform a method of detecting faults in an electricity distribution network. The method comprises: receiving magnetic field data representing at least one magnetic field generated by at least one electricity distribution line in the electricity distribution network; performing signal processing on the received magnetic field data; selecting at least one time window of the transformed magnetic field data obtained from the signal processing; and applying a classification to the at least one time window, the classification selected from a group of classifications. The group of classifications may include one or more of the group comprising: normal system state (NSS), low impedance fault (LIF), HIF high current arcing (HHCA), HIF low current (HLC).

The invention in one aspect comprises several steps. The relation of one or more of such steps with respect to each of the others, the apparatus embodying features of construction, and combinations of elements and arrangement of parts that are adapted to affect such steps, are all exemplified in the following detailed disclosure.

To those skilled in the art to which the invention relates, many changes in construction and widely differing embodiments and applications of the invention will suggest themselves without departing from the scope of the invention as defined in the appended claims. The disclosures and the descriptions herein are purely illustrative and are not intended to be in any sense limiting. Where specific integers are mentioned herein which have known equivalents in the art to which this invention relates, such known equivalents are deemed to be incorporated herein as if individually set forth.

In addition, where features or aspects of the invention are described in terms of Markush groups, those persons skilled in the art will appreciate that the invention is also thereby described in terms of any individual member or subgroup of members of the Markush group.

As used herein, '(s)' following a noun means the plural and/or singular forms of the noun.

As used herein, the term 'and/or' means 'and' or 'or' or both.

It is intended that reference to a range of numbers disclosed herein (for example, 1 to 10) also incorporates reference to all rational numbers within that range (for example, 1, 1.1, 2, 3, 3.9, 4, 5, 6, 6.5, 7, 8, 9, and 10) and also any range of rational numbers within that range (for example, 2 to 8, 1.5 to 5.5, and 3.1 to 4.7) and, therefore, all sub-ranges of all ranges expressly disclosed herein are hereby expressly disclosed.

These are only examples of what is specifically intended and all possible combinations of numerical values between the lowest value and the highest value enumerated are to be considered to be expressly stated in this application in a similar manner. In this specification where reference has been made to patent specifications, other external documents, or other sources of information, this is generally for the purpose of providing a context for discussing the features of the invention. Unless specifically stated otherwise, reference to such external documents or such sources of information is not to be construed as an admission that such documents or such sources of information, in any jurisdiction, are prior art or form part of the common general knowledge in the art.

In the description in this specification reference may be made to subject matter which is not within the scope of the appended claims. That subject matter should be readily identifiable by a person skilled in the art and may assist in putting into practice the invention as defined in the presently appended claims.

Although the present invention is broadly as defined above, those persons skilled in the art will appreciate that the invention is not limited thereto and that the invention also includes embodiments of which the following description gives examples.

BRIEF DESCRIPTION OF THE DRAWINGS

Preferred forms of the electricity distribution network fault detector will now be described by way of example only with reference to the accompanying figures in which:

Figure 1 shows an example of a fault detector configured to detect faults in an electricity distribution network;

Figure 2 shows a further example of the fault detector of figure 1;

Figure 3 shows an example of a method for detecting faults in an electricity distribution network;

Figure 4 shows an example method performed by the current estimation block of figure 2;

Figure 5 shows an example of a signal processing block from figure 2;

Figure 6 shows an example of the Deep Neural Network (DNN) classifier from figure 2;

Figure 7 shows an example of a data collection apparatus;

Figure 8 shows an example of data collection apparatus of figure 7 mounted on a power pole; and Figure 9 shows example hardware suitable for implementing components of figures 1 and 2.

DETAILED DESCRIPTION

Figure 1 shows an example of a fault detector configured to detect faults in an electricity distribution network. An electricity network 100 typically comprises components that support generation, transmission and distribution of electricity.

For example, a power plant 102 generates electricity.

Electricity transmission involves at least one transformer 104 that steps up voltage for transmission, a plurality of transmission lines 106 and 108 that carry electricity long distances, and at least one neighbourhood transformer 110 that steps down voltage.

In relation to electricity distribution, a plurality of distribution lines 112 carry electricity to houses. At least one transformer 114 on a pole steps down electricity before it enters a house 116.

In an embodiment, data collector 118 measures at least one magnetic field generated by distribution lines 112. One example of a data collector 118 includes a non-contact sensor head that includes Giant Magneto-Resistive (GMR) sensors. The sensor head is attached to a power pole and measures magnetic fields generated by distribution lines 112 supported by the power pole. Examples of data collectors are further described below.

In an embodiment, current estimator 120 obtains a reasonable estimate of current flowing in the overhead lines above the sensors. Current estimator 120 takes as input a mathematical representation of magnetic fields from line currents. Current estimator 120 inverts the mathematical representation of magnetic fields from line currents using a Bio Savart law and spatial layout of lines and sensors. In an embodiment the calculations are performed using matrix multiplication.

The estimate of current output by current estimator 120 has the potential to include errors caused for example by errors in actual versus assumed geometry and/or sensor mis-alignment. Nevertheless, data obtained from the current estimator 120 can be used by Electricity Distribution Businesses (EDBs) to manage the loadings on their overhead line networks. This could be for operational reasons, asset management, or planning purposes. In an embodiment, fault detector 122 is configured to detect and classify faults using a pre-trained classifier 128. Fault detector 122 may include a signal processor 124 that applies a function to the magnetic field data obtained by the data collector 118. Fault detector 122 may also take as input data from current estimator 120. One example of a suitable function is the Hilbert Transform or Wavelet Transform function.

In an embodiment, data selector 126 combines the data obtained from current estimator @120, transformed magnetic field data obtained from signal processor 124 and/or raw magnetic field data obtained from data collector 118 into a "wider" data time-series.

Data selector 126 selects a time window of data to examine.

In an embodiment, classifier 128 takes as input a time window of data from data selector 126. Classifier 128 applies a Deep Learning Classifier to the input data to classify a data window. Examples of classifications applied to a data window include one or more of:

• normal system state (NSS);

• low impedance fault (LIF);

• HIF high current arcing (HHCA);

• HIF low current (HLC).

Normal system state is the no fault state and includes normal system events, for example switching on or off of loads. HIF high current arcing (HHCA) and HIF low current (HLC) both represent specific stages of HIF present. There may be, for example, up to ten different HIF stages and 1 LIF stage, one of which can be assigned to a data window. The HIF fault stage can vary depending on the fault surface material.

Figure 2 shows at 200 a further example of the fault detector 100 (see figure 1). Shown in figure 2 is an example of data flow between different computation blocks.

In an embodiment data collector 118 (see figure 1) is provided by two GMR sensor heads 202 mounted on a power pole. Figure 8 shows an example of the data collector 118 in the form of two GMR sensor heads mounted on a power pole.

The two GMR sensor heads each comprise two GMR sensor chips mounted on a PCB board. The two sensor heads therefore provide a four-channel output.

In an embodiment, an analog to digital converter (ADC) 204 reads the output from GMR sensors 202 with the help of a single board computer 206 may includeat least some of the components of computer system 900 (see Figure 9). These components include memory which may include main memory 906, ROM 908 and/or storage device 910. The components may also include processor 904. Stored in memory is processor-executable instructions that, when executed by processor 904, cause processor 904 to drive the ADC 204 at a 28 kHz sampling rate and save the data to memory.

The memory of single board computer 206 has stored thereon additional processorexecutable instructions that, when executed by the processor, cause the processor to perform additional functions. Three of these additional functions include software modules shown in figure 2 as current estimation block 208, current summation block 210, and signal processing block 212.

Current estimation block 208 and signal processing block 212 are examples of current estimator 120 and signal processor 124 respectively (see figure 1).

Current Estimation Block 208 may be configured to estimate three phase currents. For 4- wire systems, current estimation block 208 may additionally be configured to estimate neutral current. The calculations performed by current estimation block 208 may be made from GMR sensor data using spatial geometry of the overhead lines. Further explanation of current estimation block 208 is provided below with regard to figure 3.

An output from current estimation block 208 includes instantaneous current output. This output may optionally be passed to current summation block 210. Block 210 transforms individual conductor waveforms received from current estimation block 208 into a zero

Output from current summation block 210 is input to signal processing block 212. As shown in figure 2, the output from current estimation block may alternatively be input to signal processing block 212 without the summation processing performed by current summation block 210.

Signal processing block 212 is further described below with reference to figure 5. Signal processing block 212 may produce an output comprising a dataset suitable for classification purposes. For example, the measurement signals may be augmented with time-frequency analysed signals. These time-frequency analysed signals may provide information about the instantaneous properties of the data time series related to the characteristics of the type of fault being detected. Outputs from current estimation block 208 and signal processing block 212 may be stored in database 214 maintained on single board computer 206. Data may be retrieved from database 214 and reported, for example, as instantaneous phase current 216.

In an embodiment the data stored in database 214 is used for fault detection and/or classification. A time window of output from signal processing block 212 is retrieved from database 214 and fed as input to an Artificial Neural Network Classifier, shown for example as Deep Neural Network (DNN) Classifier block 218. Block 218 is an example of classifier 128 (see figure 1). Block 218 is further described below (see figure 6).

An output from classifier 218 may include a plurality of output classes. A decision framework 220 may select at least one predicted class from the set of output classes. The predicted class(es) may represent a response to which an operator of electricity network 100 (see figure 1) would need to act differently.

For example, decision framework 220 may be configured to more narrowly classify a set of 13 classes coming out of DNN classifier block 218 into 4 classes for which the operator of an electricity network would need to act differently in response. An example set of four classes includes:

• Normal system state (NSS)

• Low impedance fault (LIF)

• HIF high current arcing (HHCA)

• HIF low current (HLC).

It will be appreciated that this set of classes is flexible and may be altered according to requirements of a particular operator of electricity network 100.

Figure 3 shows an example of method 300 for detecting faults in an electricity distribution network forming part of a network such as network 100 (see figure 1).

In an embodiment, method 300 includes receiving 302 measurement signals. These measurement signals may be received in real time and/or in batches for example. The measurement signals include for example the data that is input to signal processing block 212 (see figure 2). The data may include for example, outputs from analog to digital convertor 204, current estimation block 208 and/or current summation block 210.

Signal processing block 212 (see figure 2) may perform 304 signal processing on the measurement signals. Data selector 126 (see figure 1) may select 306 a time window. After the signal processing block 212 has processed the signals, data selector 126 takes a time window of the processed signals, currently of 3 cycles, and passes that to Deep Neural Network (DNN) classifier block 218.

Deep Neural Network (DNN) classifier block 218 performs 308 classification on the data.

Figure 4 shows an example method 400 performed by current estimation block 208 (see figure 2).

Method 400 receives 402 as input the x and z axis magnetic field data from two 2D sensor heads 202 (see figure 2). These fields are the cross coupled field from all overhead lines carrying the phase currents A, B, C and if it is a 3-phase 4wire system the Neutral current.

The inputs may include for example, a reading from GMR sensor x-axis (Bx) and a reading from GMR sensor z-axis (Bz). These values are represented by the matrix The individual phase current magnetic field vectors are decoupled through matrix

multiplication.

The Biot-Savart law describes the mathematical relation between current and magnetic flux density. The magnetic flux density is directly proportional to the source current and inversely proportional to the source distance. For a single line conductor, the equation relates a current value to its associated field or vice versa.

Method 400 includes receiving 404 a set of user input variables. In an embodiment, the user input variables comprise a coefficient matrix that is calculated offline using the overhead line's spatial configuration geometry. The coefficient values may be calculated, for example, at least partly based on the vertical and horizontal distances of the sensors below the overhead lines. The coefficient values may be calculated at least partly based on line sag. The coefficient values may also be calculated as a function of the orientation of the sensor (i.e. horizontal x or vertical z) on the sensor head.

In an embodiment, the coefficient values are calculated at least partly from a combination of the vertical and horizontal distances of the sensors below the overhead lines, line sag, and the orientation of the sensor (i.e. horizontal x or vertical z) on the sensor head. The set of user input variables may be provided as a matrix C. The inverse of the user input coefficient matrix C¹ is then found. Matrix multiplication is then performed 406 as follows:

Method 400 outputs 408 a set of instantaneous phase currents that have been decoupled from the GMR sensor signals, and neutral current in the case of a 3-phase 4wire system, received at step 402 above. As described above, the instantaneous phase currents may be provided as a zero sequence waveform for a 3-phase system in the form I_o = or residual current for a 3 phase 4-wire system in the form

The instantaneous phase currents may be passed to current summation block

210. Alternatively, the instantaneous phase currents and/or neutral current may be passed directly to the signal processing block 212 (see figure 2).

Figure 5 shows an example of signal processing block 212 (see figure 2). Block 212 takes as input measurement signals 502. These measurement signals may include x-axis and z-axis sensor signals from one 2D sensor head 202 (see figure 2). Alternatively or in addition, block 212 may take as input a zero-sequence current signal from block 210 for example that is calculated from two 2D sensor heads.

Block 212 may include one or both of a Hilbert transform (HT) block 504 and a discrete wavelet transform (DWT) block 506. Where block 212 includes both HT block 504 and DWT block 506, HT block 504 and DWT block 506 may be executed in parallel or sequentially.

The HT block 504 may be configured to calculate instantaneous amplitude (IA) 508 and Instantaneous phase (IP) angle 510 calculated on each input signal. The input signal to HT block 504 may pass through a high pass filter (HPF) 512 to remove at least some frequency components, for example any 50Hz fundamental frequency components.

The DWT block 506 may be configured to decompose at least one input signal into a plurality of frequency bands, called levels (L). The DWT block 506 may decompose each input signal into a plurality of levels. Because there is a down-sampling between successive levels, the coefficients from each level of signals may be linearly interpolated (1D-LI) to regain the length of the original signal. The frequency bands are shown at 514. The bands 514 may include, for example, level 1, level 2, level 3, and so on to level n. HT block 504 and DWT block 506 are configured to take as input measurement signals 502 which may include x-axis and z-axis sensor signals from one 2D sensor head 202 (see figure 2) and/or the zero-sequence current signal from block 210 that is calculated from two 2D sensor heads.

An output from block 212 may include a time-series matrix 516, where the raw original measurement signals 518 have been augmented with time-frequency analysis-based transformations of each measurement signal, such as the outputs of DWT block 506 and/or HT block 504.

This augmentation increases the dimensionality of the data by expanding each 1-D data time-series into a >2D data time-series, with information about the instantaneous properties of the data time-series that are related to the characteristics of the type of fault being detected, which are not readily apparent in the raw data. Examples of instantaneous properties of the data time-series calculated include properties such as instantaneous amplitude and phase angle signals 520 obtained from HT block 504. Further examples include DWT coefficients 522 in the form of instantaneous frequency content obtained from DWT block 506 that decomposes a signal into different frequency bands 514.

Figure 6 shows an example of a Deep Neural Network (DNN) Classifier block 218. Also shown in figure 6 is an example of decision framework 220 that is configured to select at least one output class from a set of output classes from classifier 218 (see figure 2).

DNN Classifier block 218 may include two or more cascading layers of Deep Neural Networks and a decision layer. The DNN classifier block 218 may for example be based on an original Deep Sense framework using CNN+LTSM+ANN. A possible modification may include CNN+BiGRU+ANN.

As shown in figure 6, DNN classifier block 218 may take as input some or all of the data generated by HT 504 and DWT 506. The input may include time-series matrix 516, in which the raw original measurement signals 518 have been augmented with time- frequency analysis-based transformations of each measurement signal, such as the outputs of DWT block 506 and /or HT block 504 ( see figure 5). The data may be processed by a concatenator 602 that is configured to package the data ready for input to DNN classifier block 218.

DNN classifier block 218 may implement a DNN classification algorithm, the objective of which is to learn the hierarchical local and global features of the data. The features of the data may include, for example, fault and other signatures.

Typically, input data to a DNN classifier has multiple features. These features are the function of a small number of neighbouring time samples in the input data. A convolutional neural network (CNN) is capable of processing these local features. The convolution ID (convlD) network filters scan through the time domain, therefore, mapping the temporal features.

A typical DNN input dataset has numerous samples, where similar patterns can appear in multiple instances. A recurrent neural network (RNN) could learn these repeated instances and adjust the parameters accordingly. In this learning method, a network learns the long-term contextual dependency between input data instances.

The newer improved version of RNN includes long short-term memory (LSTM) and gated recurrent unit (GRU) networks.

Convolutional neural network and recurrent network models can be configured in numerous ways. These models can be used as standalone models or combined to make a hybrid model.

DNN classifier block 218 may be based on the original DeepSense framework which used CNN followed by GRU then ANN. Block 218 may include a modification CNN + Bidirectional (Bi) GRU+ANN for potentially better performance.

In CNN + Bi-GRU, the combination of the convolution block (CNN) with the GRU to make a hybrid model has the potential to enable the model within DNN classifier block 218 to learn time-related and sequential patterns from the input.

A further block of the hybrid model comprises neural network (ANN) block. The ANN block may include a feed-forward network with three fully connected layers to generalize the features from the GRU memory block.

The first two dense layers of the ANN block may comprise dot multiplication of input and randomly initialized weight matrix. The output may be fitted by a non-linear activation function to generalize the input patterns. A third layer may comprise a generalization of the logistic regression function or Softmax. The third layer is a further layer of the neural network block to perform multi- class classification. The SoftMax layer predicts the multi- class states of an input signal from learned features, i.e., local correlation and global contextual features.

The DNN classifier block 218 may generate an output that includes a plurality of output classes. In one example, DNN classifier block 218 generates an output comprising 13 output classes. These 13 output classes may include 11 categories of the power system fault and 2 categories of system event signals. The classes may be assigned numerical values selected from a range from zero (0) to twelve (12). It will be appreciated that the number of classes may be varied.

The classes of HIF are formed on the basis of the progression of different and unique stages of the fault that occur for different fault surface materials. The classes exhibit different fault current characteristics due to the physio-chemical properties of the fault surface and the changes in those during a fault. These influence the characteristics of the measurement signals and their time-frequency characteristics.

In an embodiment, decision framework 220 applies a selection process to more narrowly classify the 13 classes coming out of the DNN classifier block 218 into 4 classes for which the operator of an electricity network would need to act differently in response. These four classes may include, for example, Normal system state (NSS) 604, Low impedance fault (LIF) 606, HIF high current arcing (HHCA) 608, and HIF low current (HLC) 610. It will be appreciated that four selected classes are shown by way of example. The number of these operator-relevant selected classes is flexible and can be altered according to the requirements of a particular operator.

One example process followed by decision framework 220 is shown in figure 6. If the output from DNN classifier block 218 is no fault (0) or load switching (6) 612, then it is classified as NSS (normal system state) 604. If it is a low impedance fault (5) 614 then it falls into LIF 606. If it is either negative (3) or stable arc (4) 616, then it falls into HHCA (HIF high current arcing) 608. All other classes of output from the DNN classifier block 218 fall into HLC (HIF low current) 610.

Figure 7 shows at 700 one example of GMR sensor 202 (see figure 2). The data collector comprises a 2D sensor head 700 that uses Giant Magneto-Resistive (GMR) sensors to detect magnetic field. GMR is a quantum mechanical magnetoresistance effect that is observed in some multi-layered structures and pertains to a change in electrical resistance depending on layer magnetisation. u GMR sensors are an attractive option due to their non-contact sensing ability at high sensitivity levels, low cost, miniature size, and wide frequency bandwidth. A GMR sensor implemented in the vicinity of overhead power lines can capture the alternating magnetic field generated by current. Because of their cost and if used in a non-contact way low cost of installation, they therefore have the potential to enable much more widespread monitoring of line loadings and also be used as the basis for detection of faults in distribution networks.

Sensor head 700 may comprise two single-axis GMR sensors arranged in a way to measure magnetic fields in two dimensions. A first GMR sensor 702 may be oriented to produce a measurement in the x-axis, while a second GMR sensor 704 may be oriented to produce a measurement in the z-axis.. These sensors capture the alternating magnetic field generated by a current carried in an overhead wire. In an embodiment, the sensors are positioned on respective RGB boards housed in a plastic case.

Another example of a data collector comprises a 3D sensor head (not shown). The 3D sensor head comprises three GMR sensors that are used to take measurements in 3D.

Figure 8 shows an example of GMR sensors from figure 1 mounted on a power pole. In an embodiment a first sensor head 802 is spaced apart from a second sensor head 804 on a power pole 806. Sensor head 802 for example may be positioned 0.4m below, and sensor head 804 may be positioned 1.1m below, the centre point of the overhead lines.

Sensor head 802 and sensor head 804 do not require direct contact with the overhead line. Positioned as shown in figure 8, they result in four measurements of the overhead line.

Where a data collector comprises a 3D sensor head, a single 3D sensor head may be positioned 0.6 below the centre point of the overhead lines for example. The 3D sensor head does not require direct contact with the overhead line. For current estimation, there is a requirement for more than one sensor head. For fault detection from magnetic fields there may be a requirement for only one 2D sensor head.

The techniques described above for fault monitoring have the potential to provide wide- spread monitoring through low-cost sensing hardware. The techniques further have the potential to provide the ability to detect the presence of HIF faults and locate them within the network from the sensor data using sophisticated algorithms embodied in software.

According to one embodiment, The techniques described herein implemented by one or generalized computing systems programmed to perform the techniques pursuant to program instructions in firmware, memory, other storage, or a combination. Specialpurpose computing devices may be used, such as desktop computer systems, portable computer systems, handheld devices, networking devices or any other device that incorporates hard-wired and/or program logic to implement the techniques.

Figure 9 is a block diagram that illustrates a computer system 900 upon which components shown in figure 1 may be implemented. Examples of components include for example data collector 118, current estimator 120, fault detector 122, signal processor 124, data selector 126, and classifier 128.

The computer system 900 includes a bus 902 or other communication mechanism for communicating information, and a processor 904 coupled with the bus 902 for processing information. The processor 904 may be, for example, a general-purpose microprocessor.

The computer system 900 also includes a main memory 906, such as a random access memory (RAM) or other dynamic storage device, coupled to the bus 902 for storing information and instructions to be executed by the processor 904. The main memory 906 may also be used for storing temporary variables or other intermediate information during execution of instructions to be executed by the processor 904. Such instructions, when stored in non-transitory storage media accessible to the processor 904, render the computer system 900 into a special-purpose machine that is customized to perform the operations specified in the instructions.

The computer system 900 further includes a read only memory (ROM) 908 or other static storage device coupled to the bus 902 for storing static information and instructions for the processor 904. A storage device 910, such as a magnetic disk or optical disk, is provided and coupled to the bus 902 for storing information and instructions.

The computer system 900 may be coupled via the bus 902 to a display 912, such as a computer monitor, for displaying information to a computer user. An input device 914, including alphanumeric and other keys, is coupled to the bus 902 for communicating information and command selections to the processor 904. Another type of user input device is a cursor control 916, such as a mouse, a trackball, or cursor direction keys for communicating direction information and command selections to the processor 904 and for controlling cursor movement on the display 912. This input device typically has two degrees of freedom in two axes, a first axis (e.g., x) and a second axis (e.g., y), that allows the device to specify positions in a plane. ces somputer system 900 may implement the techniques described herein using customized hard-wired logic, one or more ASICs or FPGAs, firmware and/or program logic which in combination with the computer system causes or programs the computer system 900 to be a special-purpose machine. According to one embodiment, the techniques herein are performed by the computer system 900 in response to the processor 904 executing one or more sequences of one or more instructions contained in the main memory 906. Such instructions may be read into the main memory 906 from another storage medium, such as the storage device 910. Execution of the sequences of instructions contained in the main memory 906 causes the processor 904 to perform the process steps described herein. In alternative embodiments, hard-wired circuitry may be used in place of or in combination with software instructions.

The term "storage media" as used herein refers to any non-transitory media that store data and/or instructions that cause a machine to operation in a specific fashion. Such storage media may include non-volatile media and/or volatile media. Non-volatile media includes, for example, optical or magnetic disks, such as the storage device 910.

Volatile media includes dynamic memory, such as the main memory 906. Common forms of storage media include, for example, a floppy disk, a flexible disk, hard disk, solid state drive, magnetic tape, or any other magnetic data storage medium, a CD- ROM, any other optical data storage medium, any physical medium with patterns of holes, a RAM, a PROM, an EPROM, a FLASH-EPROM, NVRAM, any other memory chip or cartridge.

Storage media is distinct from but may be used in conjunction with transmission media. Transmission media participates in transferring information between storage media. For example, transmission media includes coaxial cables, copper wire, and fiber optics, including the wires that include the bus 902. Transmission media can also take the form of acoustic or light waves, such as those generated during radio-wave and infra-red data communications.

Various forms of media may be involved in carrying one or more sequences of one or more instructions to the processor 904 for execution. For example, the instructions may initially be carried on a magnetic disk or solid state drive of a remote computer. The remote computer can load the instructions into its dynamic memory and send the instructions over a network connection. A modem or network interface local to the computer system 900 can receive the data. The bus 902 carries the data to the main memory 906, from which the processor 904 retrieves and executes the instructions. The instructions received by the main memory 906 may optionally be stored on the storage device 910 either before or after the by the processor 904 The computer system 900 also includes a communication interface 918 coupled to the bus 902. The communication interface 918 provides a two-way data communication coupling to a network link 920 that is connected to a local network 922. For example, the communication interface 918 may be an integrated services digital network (ISDN) card, cable modem, satellite modem, or a modem to provide a data communication connection to a corresponding type of telephone line. Wireless links may also be implemented. In any such implementation, the communication interface 918 sends and receives electrical, electromagnetic, or optical signals that carry digital data streams representing various types of information.

The network link 920 typically provides data communication through one or more networks to other data devices. For example, the network link 920 may provide a connection through the local network 922 to a host computer 924 or to data equipment operated by an Internet Service Provider (ISP) 926. The ISP 926 in turn provides data communication services through the world wide packet data communication network now commonly referred to as the "Internet" 928. The local network 922 and Internet 928 both use electrical, electromagnetic, or optical signals that carry digital data streams. The signals through the various networks and the signals on the network link 920 and through the communication interface 918, which carry the digital data to and from the computer system 900, are example forms of transmission media.

The computer system 900 can send messages and receive data, including program code, through the network(s), the network link 920, and communication interface 918. In the Internet example, a server 930 might transmit a requested code for an application program through the Internet 928, ISP 926, local network 922, and communication interface 918. The received code may be executed by the processor 904 as it is received, and/or stored in the storage device 910, or other non-volatile storage for later execution.

The foregoing description of the invention includes preferred forms thereof. Modifications may be made thereto without departing from the scope of the invention, as defined by the accompanying claims.

Claims

1. An electricity distribution network fault detector comprising: a signal processor configured to perform signal processing on magnetic field data, the magnetic field data representing at least one magnetic field generated by at least one electricity distribution line; a data selector configured to receive transformed magnetic field data from the signal processor and to select a time window of the received data; and a classifier configured to receive at least one time window from the data selector and to apply a classification to the at least one time window, the classification selected from a group of classifications.

2. The fault detector of claim 1 wherein the group of classifications includes one or more of the group comprising: normal system state (NSS), low impedance fault (LIF), HIF high current arcing (HHCA), HIF low current (HLC).

3. The fault detector of claim 1 or claim 2 wherein the classifier comprises a deep neural network (DNN) classifier block

4. The fault detector of any one of claims 1 to 3 further comprising a data collector configured to generate the magnetic field data representing the at least one magnetic field generated by the at least one electricity distribution line.

5. The fault detector of claim 4 wherein the data collector comprises two Giant Magneto-Resistive (GMR) sensor heads mounted on a power pole.

6. The fault detector of claim 4 or claim 5 further comprising an analog to digital converter (ADC) configured to receive the output from the data collector.

7. The fault detector of claim 6 wherein the signal processor is configured to receive measurement signals from the ADC.

8. The fault detector of claim 6 or claim 7 further comprising a current estimation block configured to receive measurement signals from the ADC.

9. The fault detector of claim 8 further comprising a current estimation block configured to receive instantaneous current output from the current estimation block.

10. The fault detector of claim 9 wherein the signal processor is configured to receive input from the current estimation block

11. The fault detector of any one of claims 1 to 10 wherein the signal processor is configured to output a dataset augmented with time-frequency analysed signals.

12. The fault detector of any one of claims 1 to 11 wherein the signal processor includes a Hilbert transform (HT) block configured to calculate instantaneous amplitude (IA) and instantaneous phase (IP) angle on input signals.

13. The fault detector of any one of claims 1 to 12 wherein the signal processor includes a discrete wavelet transform (DWT) block configured to decompose at least one input signal into plurality of frequency bands.

14. The fault detector of claim 13 wherein the signal processor includes a high pass filter (HPF) configured to remove at least some frequency components from the input signal to the DWT block.

15. The fault detector of any one of claims 4 to 10 wherein the data selector is configured to receive magnetic field data from the data collector and to receive the transformed magnetic field data from the signal processor.

16. A method of detecting faults in an electricity distribution network comprising: receiving magnetic field data representing at least one magnetic field generated by at least one electricity distribution line in the electricity distribution network; performing signal processing on the received magnetic field data; selecting at least one time window of the transformed magnetic field data obtained from the signal processing; and applying a classification to the at least one time window, the classification selected from a group of classifications.

17. The method of claim 16 wherein the group of classifications includes one or more of the group comprising: normal system state (NSS), low impedance fault (LIF), HIF high current arcing (HHCA), HIF low current (HLC).

18. The method of claim 16 or claim 17 further comprising selecting the at least one time window from the transformed magnetic field data obtained from the signal processing and from magnetic field data obtained from a data collector.

19. A computer-readable medium having stored thereon computer-executable instructions that, when executed by a processor, cause the processor to perform a method of detecting faults in an electricity distribution network, the method comprising: receiving magnetic field data representing at least one magnetic field generated by at least one electricity distribution line in the electricity distribution network; performing signal processing on the received magnetic field data; selecting at least one time window of the transformed magnetic field data obtained from the signal processing; and applying a classification to the at least one time window, the classification selected from a group of classifications.

20. The computer-readable medium of claim 19, wherein the group of classifications includes one or more of the group comprising: normal system state (NSS), low impedance fault (LIF), HIF high current arcing (HHCA), HIF low current (HLC).

21. The computer-readable medium of claim 19 or claim 20, the method further comprising selecting the at least one time window from the transformed magnetic field data obtained from the signal processing and from magnetic field data obtained from a data collector.