SYSTEMS AND METHODS FOR DETERMINING FAULT LOCATIONS IN AN ELECTRICITY NETWORK TECHNICAL FIELD There is disclosed systems and method for determining fault locations in an electricity network, for example, fault location in a distribution network, such as high impedance faults. BACKGROUND Power system networks are the backbone of any economy. The sustained development in infrastructure has led to rapid development of power system networks. A power system network can be divided into three main categories namely, generation, transmission, and distribution. Most of the generation is done by centralized generating stations. The transmission system transports the power from generating stations to load centres by overhead lines or high voltage underground lines. It is the distribution network that does the final job of power delivery to the load centres. The routing and design of distribution networks are determined by the nature, location of the network and type of load that needs to be supplied. Power systems have evolved over the years to keep up with ever-increasing demand. To meet increasing demand and maintain reliable operation, power systems have become complex networks. Advanced monitoring and smart grid technology are being deployed to meet the system requirements. One of the main goals of any power system operation is identification of system faults. Under any circumstances, a fault in the system is a serious threat, which can endanger its integrity. Devices like relays and circuit breakers are used as protection devices in case of faults in the system. Various sensors like current transformers (CTs) and potential transformers (PTs) are used to monitor the system parameters. Low impedance faults (LIF) have very low fault resistance which makes them easily detectable by conventional over-current protection relays whereas high impedance faults (HIF) have low fault current. Often due to the intermittent nature of high impedance faults, detection of these faults has always been a challenge for power companies. It is important to detect and locate high impedance faults as they may present a danger to people and may be precursor to more serious faults. Medium voltage electricity distribution networks experience a wide range of faults that can compromise electricity supply and/or human safety. Existing technology is not capable of effective or reliable fault location detection for many categories of fault. Electricity distribution companies expend considerable time and effort locating and managing network faults so as to improve operation and safety of those networks. While low impedance faults are rapidly cleared by line protection systems, high impedance faults go undetected until they cause obvious effects, such as fire ignition or visual reports. Fire ignition is a problem that can result in disconnecting power at locations with high fire risk (such as bush fires). As such, high impedance faults can cause fires and public safety issues. Therefore, to quickly and effectively locate and manage faults that are not detected by existing technology would be of benefit. The ability to implement wide-spread monitoring through low-cost sensing hardware, and to analyse the collected data to rapidly determine useful information about system faults and their location is a high value proposition. SUMMARY There is described systems, methods and data for determining a fault location in an electricity distribution network. The systems, methods and data described may be useable to quickly and effectively locate and manage faults. The systems, methods and data described may be usable to implement wide-spread monitoring through low-cost sensing hardware. The systems, methods and data described may be usable to analyse collected data to rapidly determine useful information about system faults and their location. In one example, there is provided a method for determining a fault location in an electricity distribution network. Such a method may comprise receiving one or more electrical signals (e.g. voltage or current signals) obtained from one or more measurement locations at the electricity distribution network. The one or more electrical signals may have been influenced by a fault at the network. The method may decompose each electrical signal to derive one or more constituent frequency components. Such constituent frequency components may be constructable to approximate (or estimate) the electrical signal. The method may provide a circuit model of a portion of the electricity distribution network. Such circuit model may comprise an expected fault location. The method may establish one or more predicted relationships between the fault location and the electrical signals using the circuit model. The method may derive one or more estimated fault locations to determine the fault location in the electricity distribution network. The method may derive each estimated fault location from a predicted relationship using a constituent frequency component decomposed from each of the respective electrical signals in the predicted relationship. Described herein, there is provided a method for determining a fault location in an electricity distribution network, the method comprises: receiving one or more electrical signals obtained from one or more measurement locations at the electricity distribution network, the one or more electrical signals having been influenced by a fault at the network; decomposing each electrical signal to derive one or more constituent frequency components, the constituent frequency components being constructable to approximate the electrical signal; providing a circuit model of a portion of the electricity distribution network comprising an expected fault location; establishing one or more predicted relationships between the fault location and one or more of the constituent frequency components of the electrical signals using the circuit model; and deriving one or more estimated fault locations using the predicted relationships to determine the fault location in the electricity distribution network. In some examples, the predicted relationships are established based additionally on fault impedance and the electrical signals (e.g. respective constituent frequency components) using the circuit model, and the method may include deriving one or more estimated fault locations and one or more estimated fault impedances using the predicted relationships to determine the fault location in the electricity distribution network. The one or more electrical signals influenced by the fault may comprise a damped transient signal portion, and the one or more constituent frequency components are derived, at least in part, from that damped transient signal portion. The one or more predicted relationships may be established based on one or more specific characteristics of the constituent frequency components, those characteristics being magnitude, frequency, phase angle or damping. The constituent frequency components may be selected based on natural resonant frequencies of the network. The predicted relationships may be based additionally on fault current injection magnitude and the electrical signals (e.g. respective constituent frequency components) using the circuit model, and the method may include deriving one or more estimated fault locations and one or more estimated fault current injection magnitudes using the predicted relationships to determine the fault location in the electricity distribution network. The fault current injection magnitude may be representative of a fault current injection resulting from an arcing event at the fault location. The fault current injection may create harmonic frequencies in the electrical signals of the network that are measurable at the measurement locations of the network. The one or more electrical signals influenced by a fault may comprise a steady state signal portion, and the one or more constituent frequency components are derived, at least in part, from that steady state signal portion. The one or more electrical signals may be voltage signals and/or current signals, and constituent frequency components may be separately decomposed for the voltage signals and/or the current signals, so as to derive a determined fault location in the electricity distribution network. The one or more constituent frequency components decomposed from one of the electrical signals may have common frequencies (or the same frequencies) to constituent frequency components decomposed from another of the electrical signals. Further, each estimated fault location may be derived from the said predicted relationship using constituent frequency components with a common frequency. Further, two or more of the constituent frequency components with different frequencies may be derived from each of the electrical signals, and wherein at least two estimated fault locations may each be derived from the said predicted relationship using constituent frequency components with a common frequency different to a common frequency of constituent frequency components used to derive another estimated fault location using the said predicted relationship. Further, two or more of the predicted relationships are established, and wherein at least two estimated fault locations may each be derived from a predicted relationship different to a predicted relationship used to derive another estimated fault location. Further, at least two of the estimated fault locations may be derived from different respective predicted relationships using respective constituent frequency components having a common frequency, and wherein the common frequency used in each of the predicted relationships is the same. Further, at least two of the estimated fault locations may be derived from different respective predicted relationships using respective constituent frequency components having a common frequency, and wherein the common frequency used in each of the predicted relationships is different. The one or more predicted relationships may be established using an inverse problem approach. The predicted relationships may be established as error minimisation problems wherein each estimated fault location is determined for corresponding constituent frequency components by varying expected values of the fault location so as to provide a minimum error in a corresponding error minimisation problem. The error minimisation problem may comprise: forming a surface plot using varying values of fault location for corresponding constituent frequency components; determining at least one minima of the surface plot, and deriving the determined fault location from an estimated fault location associated with the at least one minima. The method may provide a plurality of different estimated fault locations, and the fault location is determined from the estimated fault locations based on a sensitivity of a respective predicted relationship to a corresponding estimated fault location. In some examples, the sensitivity of the respective predicted relationship to the corresponding estimated fault locations may be determined based on the second differential of the minima of the error minimisation problem. The method may provide a plurality of different estimated fault locations, and the fault location is determined from the estimated fault locations based on initial magnitude, damping and/or energy of corresponding constituent frequency components used to derive a corresponding estimated fault location. The method may provide a plurality of different estimated fault locations, and the fault location is determined from a unified estimate fault location that is a weighted mean of the estimated fault locations, wherein a weighting of each estimated fault location correlate to: a sensitivity of a respective predicted relationship to a corresponding estimated fault location, and/or initial magnitude, damping and/or energy of corresponding constituent frequency components used to derive a corresponding estimated fault location. The fault at the network may be detected by reducing received electrical signals to zero- sequence components. The fault at the network may be detected by using a sliding window algorithm on the received electrical signals. The fault at the network may be a high impedance fault. In one example, there is provided a computer readable medium carrying instructions which, when executed by at least one processor of a system, cause the system to carry out the method according to the embodiments described herein. In one example, there is provided a system for determining a fault location in an electricity distribution network, the system comprises: at least one processor, wherein the processor is configured to: receive one or more electrical signals obtained from one or more measurement locations at the electricity distribution network, the one or more electrical signals having been influenced by a fault at the network; decompose each electrical signal to derive one or more constituent frequency components, the constituent frequency components being constructable to approximate the electrical signal; provide a circuit model of a portion of the electricity distribution network comprising an expected fault location; establish one or more predicted relationships between the fault location and one or more of the electrical signals using the circuit model; and derive one or more estimated fault locations to determine the fault location in the electricity distribution network, wherein each estimated fault location is derived from a predicted relationship of the predicted relationships using a constituent frequency component decomposed from each of respective electrical signals in the said predicted relationship. The term ‘comprising’ as used in this specification and claims means ‘consisting at least in part of’. When interpreting statements in this specification and claims which include the term ‘comprising’, other features besides the features prefaced by this term in each statement can also be present. Related terms such as ‘comprise’ and ‘comprised’ are to be interpreted in a similar manner. 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. BRIEF DESCRIPTION OF THE DRAWINGS Examples of the system and method will now be described by way of example only with reference to the accompanying figures in which: Figure 1 shows a representative circuit diagram of a single phase electricity distribution network; Figure 2 shows a damped transient signal portion of an electrical signal; Figure 3 shows a steady state signal portion of an electrical signal; Figure 4 shows a table of constituent frequency components decomposed from damped transient signal portions of electrical signals; Figure 5 shows a circuit model of a portion of the network shown in Figure 1; Figure 6 shows a table with estimated fault locations and estimated fault impedances; Figures 7A, 7B, 7C, 8A, 8B and 8C show surface plots at respective constituent frequency components; Figure 9 shows a circuit model of a portion of a three-phase network; Figures 10 shows a table of constituent frequency components decomposed from damped transient signal portions of electrical signals of a three-phase network; Figure 11 shows a table with estimated fault locations and estimated fault impedances of the three-phase network; Figures 12A and 12B shows odd and even harmonic portions of the steady state signal portion of Figure 3; Figure 13 shows another circuit model of a portion of a three-phase network with a current injection at the expected fault location; Figure 14 shows a table of constituent frequency components decomposed from steady state signal portions of electrical signals of a three-phase network; Figure 15 shows another table of constituent frequency components decomposed from steady state signal portions of electrical signals of a three-phase network; Figure 16 shows a table with estimated fault locations and estimated fault current injection magnitudes of a three-phase network; Figure 17 shows a table with estimated fault locations and estimated fault current injection magnitudes of a three-phase network; Figure 18 shows another circuit model of a portion of a three-phase network including capacitances in the portion; Figure 19 shows an example of a method of determining a fault location in use. DETAILED DESCRIPTION Disclosed herein are systems and methods for determining fault locations, for example, in electricity networks such as electricity distribution networks. It will also be understood that the system and methods described may be applied to other electricity networks, such as electricity transmission networks. Some examples described relate specifically to so-called high impedance faults, but it will be appreciated that the same system and methods may be used to determine the location of other faults. By way of an example, Figure 1 shows a representation of an electricity distribution network 100, which is presented in terms of a circuit diagram having electrical properties of the network 100, as will be understood. In this example, the network is a single-phase network and includes overhead lines (including lines 108, 110, 112) as well as underground cables (including lines 114, 116, 118). It will be understood that the systems and methods disclosed herein may also be applied to a three-phase network. As shown in Figure 1, an example fault 106 is shown as occurring on one of the overhead lines 108. The fault 106 may be categorised as a low impedance fault, medium impedance fault or a high impedance fault. In examples where the fault may be considered to be a low impedance fault, this may create fault currents significantly larger than the expected (e.g., maximum) load current of the network 100. These currents may be detected by protective devices or relays in the network 100, and remedial action taken immediately to protect the network and maintain safety. In contrast, high impedance faults (e.g. where current leaks undesirably from the network 100) may create fault currents that may be considered to be small compared to the load current of the network 100, and may in some cases need visual confirmation in order to identify and clear the fault from the network. High impedance faults may occur, for example, when the overhead line 108 comes in contact with trees or breaks and falls to the ground. These faults may also be considered as line-to-ground faults or may be arc faults, or the like. It may be desirable to best understand the location of such faults within a geographically expansive network so that rapid identification and remedial action can be taken. The following describes examples of determining the fault 106 locations in the network 100 and, in particular, where the fault 106 may be considered to be a high impedance fault. Signals at the distribution network 100 (e.g., signals that may be used to communicate electrical power and/or data) may vary over time depending on load conditions at the network 100. This is common and expected during normal operation of such networks 100. From time to time, however, changes in voltage, current, or changes in other signals or conditions in the network 100 (e.g., changes from the norm or beyond what would be expected) may be considered to be indicative of the occurrence of a fault condition at the network 100. As may be expected, monitoring of some or all of those signals or conditions may be helpful in order to determine what may be a fault 106, versus what may simply be normal operation. Further, power or data signals from the network 100 may themselves be influenced by fault conditions such that they contain information regarding such faults 106, particularly during high impedance fault conditions, which may not be immediately apparent (e.g., compared to larger transients associated with low impedance faults). An example of identifying possible faults may be by determining any deviations of electrical signals away from the zero sequence components that are present during normal operation, which may be indicative of a change in the network from the normal operation. In another example, a sliding window algorithm may be used, additionally or alternatively, to detect that a variation in signal properties that may be associated with a high impedance fault. This works by receiving continuous power or data signals from the network 100 and analysing those signals over time to determine any changes in the signals beyond what would be expected. Such signals from the network and methods exemplified above may be used to help identify that a fault has occurred, or provide some probability of occurrence, but they may not be able to determine where, at the network 100, such faults have occurred. As will be appreciated, the ability to provide an indication of location of such fault permits a network operator to allocate resources to fixing the fault at the fault location, thereby improving electricity distribution and reducing risks caused by the fault. In the method described, electrical signals influenced by a fault 106 are used (e.g., received directly or the information from those signals used) to determine a fault 106 location. Such signals may be obtained from one or more measurement locations at the electricity distribution network 100. It will be appreciated that the network 100 may comprise any number of measurement locations (and could likely have many such locations), but for the purposes of example – and with a single fault location to determine – Figure 1 shows two measurement locations 102 and 104 that are provided on the network 100. In some examples, it may be possible to use only a single measurement location and derive a fault location from that. Here, however, electrical signals, such as voltage signals, current signals, electric fields and/or magnetic fields associated with the power distribution at the two measurement locations on the network 100 may be measured during operation of the network 100. As an example, electrical signals may be continuously monitored to identify an occurrence of a fault. If a fault occurrence is identified, electrical signals indicative of a fault may be recorded to be further analysed to determine the location of the fault on the network 100. As shown, fault 106 occurs between the measurement locations 102, 104, which is also representative of the real world location of the fault 106 between those measurement locations 102, 104, but which would otherwise be unknown to the network operator or a maintenance team. It will be appreciated that the fault 106 may occur at any time and that, throughout the duration of the fault event, the electrical signals measured may have changing properties (e.g., changing waveforms). In some examples, such properties may mean that the signals, received over time can be considered to have two signal portions associated with the high impedance fault: a first signal portion may be associated with the initial fault event (e.g., in a transient manner), while a second signal portion may be associated with the longer-term fault condition (e.g., in a more steady state manner). When a fault 106 occurs in such a manner, the resulting signal associated with the initial fault event may be considered to provide a damped transient signal portion of the electrical signal. After the initial fault event, the resulting signal (e.g., waveforms of the signal) associated with the longer-term fault condition may become steadier and relatively non-damped and correspond to a steady state signal portion of the electrical signal. In some examples, an arcing event during a high impedance fault may influence the steady state signal portion of the electrical signal. When identified, aspects of the electrical signals may be categorised into the first and second portions, each associated with a different temporal aspect related to the fault 106. Information from each of the signal portions may be used together, or separately, to determine the fault location as will be described. Figure 2 shows an example of a damped transient signal portion 204 of an electrical signal 202, having been influenced by a fault in the network (e.g. fault 106). As can be seen, the damped portion 204 has an overall amplitude that decays over time and exhibits oscillatory but non-sinusoidal behaviour. Figure 3 shows an example of a steady state signal portion 302 of the electrical signal 202, having been influenced by a fault (e.g. fault 106). As can be seen, the steady state portion maintains a somewhat more consistent amplitude, and in this case is periodic but not entirely sinusoidal. It will be appreciated that an example electrical signal 202 may be derived from filtering and/or processing an electrical signal first measured from the network. As such, in an example, an electrical signal first measured from the network may be considered to be a combination of the electrical signal 202 and a fundamental signal of the network. Such a fundamental signal may have a frequency of 50Hz, for example. Here, and after having been received, each electrical signal that may have been influenced (or may expected to have been influenced) by a fault 106 are decomposed to derive one or more constituent frequency components. The derived constituent frequency components may be represented or characterised by one or more specific characteristics such as magnitude, frequency, phase angle or damping. The constituent frequency components would otherwise be constructable to approximate each electrical signal in a known manner. In an example where a voltage signal ^^ _^ and a current signal ^^ _^ are obtained from measurement location 102, the voltage signal is decomposed separately from the current signal to derive respective constituent frequency components for each of the signals ^^ _^ , ^^ _^ . To illustrate, Figure 4 shows Table 400 where each of the electrical signals ^^ _^ , ^^ _^ , ^^ _^ and ^^ _^ are decomposed into constituent frequency components of different frequencies, albeit each of the components have common frequencies. That is to say that the one or more constituent frequency components decomposed from one of the electrical signals can have the same or common frequencies to constituent frequency components decomposed from another of the electrical signals. To illustrate, Table 400 shows some constituent frequency components 402, 404 and 406 of the voltage signal ^^ _^ derived with respective frequencies of 50 Hz, 181 Hz and 303 Hz, while constituent frequency components 408, 410 and 412 of the voltage signal ^^ _^ are also derived with respective common frequencies of 50 Hz, 181, Hz and 303 Hz. Similarly, current signals ^^ _^ , ^^ _^ also have constituent frequency components with common frequencies of 50 Hz, 181, Hz and 303 Hz. It will be appreciated, that while the constituent frequency components may be characterised by a damping parameter, such a parameter may be negligible or zero in some cases. For example, the damping parameter for components 402 and 408 (components with frequencies of 50 Hz) may be zero. In some examples, the constituent frequency components, including components 402, 404, 406, 408, 410 and 412, as shown in Table 400 may be derived, at least in part, from a damped transient signal portion (e.g. signal portion 204) of one or more electrical signals ^^ _^ , ^^ _^ , ^^ _^ and ^^ _^ influenced by a fault. In some examples, constituent frequency components, including components 404, 406, 408, 410 and 412, of the damped transient signal portion (e.g. signal portion 204) are related to the natural resonant frequencies of the entirety of network 100 influenced by components on the network 100. Each of the constituent frequency components, including components 402, 404, 406, 408, 410 and 412, at the damped transient signal portion of respective electrical signals may be characterised by signal parameters of frequency, damping, magnitude as well as phase. The damping signal parameter is associated with the second-order time constant of the signal during the initial fault event. Here, it will be appreciated that although a signal portion may include constituent frequency components with zero damping (e.g. components 402 and 408), the signal portion may be considered a damped transient signal portion as it includes constituent frequency components with damping. Many methods may be used to decompose a damped transient signal portion (e.g. signal portion 204) of electrical signals ^^ _^ , ^^ _^ , ^^ _^ and ^^ _^ but, in this example, the Prony’s method may be implemented to decompose damped transient signal portion of electrical signals to derive at least one of the one or more constituent frequency components, such as components 402, 404, 406, 408, 410 and 412, with a damping characteristic (or parameter). Prony’s method is an extension of Fourier analysis and it estimates the frequency, magnitude, phase and damping of the constituent frequency components. The Prony’s method may be able to reduce any error between a function representing the damped transient signal portion, of signal ^^ _^ for example, and a function representing the combination of constituent frequency components, including components 402, 404 and 406 that are constructable to approximate signal ^^ _^ for example, with estimated magnitude, frequency, phase and damping. A damped transient signal portion of an electrical signal can be represented as a combination of exponential functions as:

equations 1 and 2 may be characterised by a number of signal parameters. The damping factor ^^ _^ , frequency ^^ _^ , magnitude ^^ _^ and phase ^^ _^ may be used as parameters of the constituent frequency components, and in this example may be determined from the roots ^^ _^ and ℎ _^ using the following equations, wherein the roots ^^ _^ and ℎ _^ may be determined using Prony’s method: equations 3 and 4 equations 5 and 6 Additionally or alternatively, the damped transient signal portion 204 of electrical signals ^^ _^ , ^^ _^ , ^^ _^ and ^^ _^ may also be decomposed to derive at least one of the one or more signal parameters of constituent frequency components, such as components 402, 404, 406, 408, 410 and 412, with a damping characteristic using Matrix Pencil (MP) methodology. Matrix pencil method may be considered to be an alternative to Prony’s method and essentially consists of solving an eigenvalue problem. In some examples, Matrix Pencil can be more resilient to noise than the Prony’s method, which may be beneficial in electricity networks. The application of Matrix Pencil requires consideration of the sampling rate used to obtain the electrical signals, windowing of the damped transient signal portion of the electrical signals and the order of the Matrix Pencil. The order defines the number of constituent frequency components derived from each electrical signal. For example, the sampling rate selected for measurements used to obtain electrical signals to be decomposed by the Matrix Pencil may be selected using the Nyquist Criterion, which is at least double the highest frequency component in the electrical signal. The windowing of the damped transient signal portion may be selected to at least capture a full cycle of the fundamental frequency of the electricity distribution network. For example, a window size of 20 ms may be selected for a fundamental frequency of 50 Hz. The windowing of the damped transient signal portion can also be selected based on a separation between two natural frequencies in an electrical signal. In an example, an electrical signal with two constituent frequency components at 660 Hz and 690 Hz are separated by 30 Hz. In this case a window size of 33.3 ms or longer may be considered suitable. The order of the Matrix Pencil may be selected such that constituent frequency components that are associated with a fault 106 are decomposed from respective electrical signals without decomposing constituent frequency components that are associated with noise frequencies or other frequencies not related to a fault 106. A constituent frequency component that is associated with a fault 106 may have a large initial magnitude, small damping and/or large energy. A Matrix Pencil with a high order may fit noise frequencies and/or other undesirable frequencies not related to a fault 106. A Matrix Pencil with a low order may not derive a sufficient number of constituent frequency components to substantially approximate the electrical signal resulting from a fault 106. This may result in errors in estimated magnitude and phases of constituent frequency components that have similar frequencies. After the constituent frequency components have been derived (e.g., and associated signal parameters), it may then be possible to derive one or more estimated fault locations using a circuit model of a portion of the electricity distribution network 100, as will be described. Such a circuit model is provided such that it comprises the expected fault 106 location. The circuit model may be considered to be a model comprising the line between the two measurement locations 102, 104 (e.g., a forward model of the line). Figure 5 shows an exemplary circuit model 500 that may be used and may be considered to be a portion of the electricity distribution network 100 between the two measurement points 102, 104 (see also Figure 1). The circuit model 500 shows the relevant parameters of resistances and inductances of the overhead line 108, and fault impedance and distances, ^^ and 1 − ^^, yet to be determined from the measurement locations 102, 104 to the fault 106. After having been determined, the distances, ^^, 1 − ^^, from the measurement locations 102, 104, to the fault 106 at the model 500 may correlate, in real life, to the fault location in the network 100. Using the circuit model 500, one or more predicted relationships between electrical signals and fault location are established. As such, predicted relationships between the fault location and one or more of the constituent frequency components, such as components 402, 404, 406, 408, 410 and 412, of the electrical signals are established, as will be described. The one or more predicted relationships may be based on one or more specific characteristics of the constituent frequency components, such as magnitude, frequency, phase angle or damping. In this particular example, the one or more predicted relationships may be established using an inverse problem approach. That is to say that the constituent frequency components, such as components 402, 404, 406, 408, 410 and 412, of electrical signals can be used to determine one or more estimated fault locations that could have influenced the electrical signals and constituent frequency components. Using the predicted relationships, one or more estimated fault locations are derived based on constituent frequency components. Each estimated fault location can be derived from a predicted relationship using a constituent frequency component decomposed from each of respective electrical signals in said predicted relationship, as will be described. Further, each estimated fault location can be derived from the said predicted relationship using constituent frequency components with a common frequency. Those estimated fault locations can then be used (e.g., in conjunction with an estimated fault impedance) to determine the fault location in the electricity distribution network. In this example, the predicted relationships are established using matrix equation 7. In this example, the predicted relationships are represented in the Laplace domain. The matrix equation shows two predicted relationships between the fault location and the electrical signals at the measurement locations 102, 104. In this example, the predicted relationships are established based on fault location and additionally on fault impedance and the electrical signals (and as such the constituent frequency components, such as components 402, 404, 406, 408, 410 and 412), using the circuit model 500. This means that the predicted relationships can be used to derive one or more estimated fault locations and one or more estimated fault impedances that may result in corresponding constituent frequency components. The fault location is represented by ^^ which also represents a normalised distance from the fault location to measurement location 102. The fault impedance is represented by ^^ _^ . The total inductance and total resistance between the two measurement points 102, 104 are represented by ^^ and ^^ respectively. The total inductance ^^ and total resistance ^^ may be known based on the design parameters of the network or may be approximated. The electrical signals and constituent frequency components decomposed, such as components 402, 404, 406, 408, 410 and 412, from respective electrical signals are represented by ^^ _^ , ^^ _^ , ^^ _^ and ^^ _^ , as above, and the damped transient signal portion (e.g. portion 204) of electrical signals ^^ _^ , ^^ _^ , ^^ _^ and ^^ _^ may be used to derive constituent frequency components. equation 7 The predicted relationships may also be established as error minimisation problems wherein each estimated fault location is determined for corresponding constituent frequency components by varying expected values of the fault location, e.g. between 0 to 1 of the normalised distance ^^, so as to provide a minimum error in a corresponding error minimisation problem. For example, the matrix equation 7 can be rearranged to form two predicted relationships as two error minimisation problems: 8 The impedance as ^^ _^ = ^^ + ^^ ^^. For each error minimisation problem, constituent frequency components may be used as values of ^^ _^ , ^^ _^ , ^^ _^ or ^^ _^ . The constituent frequency components used in this example are also characterised by damping. Line impedance ^^ _^ may be determined based on the design parameters of the network, as mentioned. By establishing a relationship using the fault location ^^ and fault impedance ^^ _^ (e.g., and varying expected or predicted values of location and/or impedance), a minimum error can be determined. The values of the fault location and fault impedance that provides the minimum error may be used to identify an estimated fault location and estimated fault impedance. At least two estimated fault locations can be derived from each predicted relationship using constituent frequency components of different frequencies. Each time an estimated fault location is derived from a predicted relationship using constituent frequency components of the same frequency or a common frequency. Such a common frequency being different to the common frequency of constituent frequency components used to derive another estimated fault location from the said predicted relationship. To illustrate, error minimisation problem represented by equation 8 can be solved by using values of ^^ _^ , ^^ _^ , ^^ _^ and ^^ _^ . The values of ^^ _^ , ^^ _^ and ^^ _^ may be based on constituent frequency components decomposed from respective electrical signals, such as components 402, 404, 406, 408, 410 and 412, shown in Table 400. A first estimated fault location can be derived from the error minimisation problem of equation 8 using the constituent frequency components with a common frequency of 50 Hz decomposed from respective electrical signals ^^ _^ , ^^ _^ and ^^ _^ . A second estimated fault location can be derived from the error minimisation problem of equation 8 using the constituent frequency components with a common frequency of 181 Hz decomposed from respective electrical signals ^^ _^ , ^^ _^ and ^^ _^ . The error minimisation problem of equation 8 can further be solved by constituent frequency components of respective electrical signals with the rest of the frequencies shown in Table 400 to derive a total of nine estimated fault locations 602 shown in Table 600 of Figure 6. In this example, each time equation 8 is solved, it is solved by constituent frequency components of the same frequency. Here, at least two estimated fault locations can also be derived from two or more separate or different predicted relationships. That is to say, each estimated fault location can be derived from a predicted relationship different to a predicted relationship used to derive another estimated fault location. To illustrate, both equation 8 and equation 9 can be solved by constituent frequency components of the same frequency (or common frequency) of respective electrical signals. This means that each time an equation is solved, the constituent frequency components of the same frequency are used to solve the equation. As an example, a first estimated fault location may be derived from equation 8 using a constituent frequency component decomposed from each of respective electrical signals ^^ _^ , ^^ _^ and ^^ _^ , those constituent frequency components with a common frequency of 50 Hz. A second estimated fault location may be derived from equation 9 using a constituent frequency component decomposed from each of respective electrical signals ^^ _^ , ^^ _^ and ^^ _^ , those constituent frequency components also with a common frequency of 50 Hz. In this example, a constituent frequency component of electrical signal ^^ _^ is used in one equation while a constituent frequency component of electrical signal ^^ _^ is used in the other equation. However, both the constituent frequency components from ^^ _^ and ^^ _^ have a common frequency of 50 Hz. To further illustrate, both equation 8 and equation 9 can be solved by constituent frequency components of all frequencies in Table 400. This means that each predicted relationship or error minimisation problem can be solved nine times, each time at a different frequency, to derive nine estimated fault locations, such as the nine estimated fault locations 602. For example, two estimated fault locations may first be derived by solving both equations 8 and 9 a first time by constituent frequency components of 181 Hz. Then, two more estimated fault locations may be further derived by solving both equations 8 and 9 a second time by constituent frequency components of 303 Hz and so forth. If both equations 8 and equation 9 are solved nine times, a total of 18 estimated fault locations 604 can be derived. To further illustrate, following this logic, at least two estimated fault locations may be derived from different respective predicted relationships using respective constituent frequency components having a common frequency, where the common frequency used in each of the predicted relationships is different. If desired, the error minimisation problems can also be solved by forming a surface plot using varying values of estimated fault location for corresponding constituent frequency components. At least one minima of the surface plot can be determined to derive at least one estimated fault location associated with the at least one minima. Figures 7a, 7b, 7c, 8a, 8b and 8c show example surface plots formed by varying fault location by varying expected values of the fault location ^^ and fault impedance ^^ _^ using equations 8 and 9 and constituent frequency components with frequencies of 0 Hz, 50 Hz, 181 Hz, 303 Hz, 329 Hz and 454 Hz from Table 400. The minima of the surface plots corresponding to the minimum errors of the error minimisation problems can be seen at the lowest regions on the surface plots. As demonstrated in Figure 6, a plurality of different estimated fault locations 604 may be derived by solving predicted relationships using a plurality of constituent frequency components. A plurality of different estimates of fault impedance 606 may also be derived. A fault location 106 in the electricity distribution network 100 can then be determined from the plurality of estimated fault locations 604. The fault location 106 may be determined from the plurality of estimated fault locations using, by way of one example, confidence measures associated with each of the estimated fault locations. The fault location may, alternatively or additionally, be determined from a unified estimate fault location that is a weighted mean of the estimated fault locations 604. The fault location 106 may be determined from the plurality of estimated fault locations 604 based on a sensitivity of each respective predicted relationship to a corresponding estimated fault location. The sensitivity of a predicted relationship to estimated fault location values also represents the strength of the relationship between fault location and the constituent frequency components (e.g the one or more characteristics of the constituent frequency components). This can be illustrated using error minimisation problem of equation 8. As expected values of fault location varied, e.g. between 0 to 1, the higher the rate the error value changes in equation 8, the higher the sensitivity of the predicted relationship. The higher the sensitivity, the more confidence is given to the estimate fault locations from the respective error minimisation problem or predicted relationship. In an example, the sensitivity of a predicted relationship to the fault location may be determined by determining the slope or first differential of the predicted relationship (e.g. at a zero-crossing). In another example, the sensitivity of the respective predicted relationship to the corresponding estimated fault locations may be determined based on the second differential of the minimum error value of an error minimisation problem. The minimum error value may also be a minimum absolute error value. The higher the first differential or second differential, the stronger the relationship is between the constituent frequency components and the fault location. Table 600 shows the nature 608 of minimum error values of respective estimated fault locations 604 and estimated fault impedances 606. Following this example, a minimum error value that is well defined has a corresponding predicted relationship with high sensitivity. A minimum error value that is weakly defined has a corresponding predicted relationship with low sensitivity. The fault location 106 may also be determined from the plurality of estimated fault locations 604 based on initial magnitude, damping and/or energy of corresponding constituent frequency components used to derive a corresponding estimated fault location 604. The higher the initial magnitude of the constituent frequency components used to solve the predicted relationships, the more confidence can be given to fault location estimates resulting from those constituent frequency components. The lower the damping of constituent frequency components, the more confidence can be given to the estimated fault locations resulting from those constituent frequency components. Estimated fault locations derived from constituent frequency components with large initial magnitude, but high damping may have less confidence. Estimated fault locations derived from constituent frequency components with a small initial magnitude, but low damping may have more confidence. More confidence can be set for estimated fault locations from constituent frequency components that have more energy. The energy of a constituent frequency component is partly governed by the initial magnitude and partly by the damping. The energy of a constituent frequency component can be determined using equation 10. equation 10 In an example, an estimated fault location may be selected to be the determined fault location based on both the sensitivity of a respective predicted relationship to the selected estimated fault location, and initial magnitude, damping and/or energy of corresponding constituent frequency components used to derive the selected estimated fault location. In an example where a unified fault location or impedance estimate is obtained, the weighting of each estimated fault location may correlate to the sensitivity of each respective predicted relationship to a corresponding estimated fault location, and/or initial magnitude, damping and/or energy of corresponding constituent frequency components used to derive a corresponding estimated fault location. In the example shown in Figure 6, the fault location may be determined from the 18 estimated fault locations 604 based on the sensitivity of a respective predicted relationship to a corresponding estimated fault location. As shown, estimated fault location 610 has a minimum error value that is well defined, indicating that a corresponding predicted relationship used to derive estimated fault location 610 has a high sensitivity. Therefore, estimated fault location 610 may be selected to be the determined fault location. The fault location is determined to be a normalised distance of ^^ = 0.25 from the fault location to measurement location 102. As shown in Figure 5, this will be related to a corresponding actual distance from the measurement location 102 to the expected fault location 106. The actual distance from the fault location to measurement location 102 may be determined from a known or estimated total length of overhead line 108. The determined fault location may then be used to by the network operator or maintenance team to repair or investigate the fault on the network 100 at the fault location determined from estimated fault location 610. In the event that the fault is not identified at that location, then the next probable location may be determined from estimated fault locations 604 using confidence measures for example. An alternative approach of using a unified estimated fault location, or the like, may also be used. While the above examples were described relative to a single-phase network, it will readily be appreciated that the similar methodology may be used for a three-phase electricity network. By way of an example, Figure 9 shows a circuit model 900 of a portion of an example three-phase electricity distribution network. The three-phase network comprises three overhead lines 912, 914 and 916. The three-phase network also includes measurement locations 902 and 904, from which electrical signal may be obtained that are associated with the power being distributed at the network. Electrical signals from any or all of the three overhead lines 912, 914 and 916 can be obtained from either of the measurement locations 902 or 904. Each of the electricals signals can be a voltage signal or current signal. The circuit model 900 also includes an expected fault 908 represented with an impedance that is occurring on line 914. Each of the electrical signals from the measurement locations 902 and/or 904 of the circuit model 900 can be decomposed to constituent frequency components, as before. Figure 10 shows a Table 1000 with example constituent frequency components, including components 1002, 1004 and 1006, for all of the electrical signals in a three-phase network of circuit model 900. These constituent frequency components may be characterised by one or more specific characteristics such as magnitude, frequency, phase angle or damping. These constituent frequency components, including components 1002, 1004 and 1006, in this example are decomposed from a damped transient portion (e.g. portion 204) of respective electrical signals related to the natural resonant frequencies of the entirety of the three-phase network of circuit model 900 influenced by components on the three-phase network. Therefore, the constituent components of this example each include a damping characteristic. Based on the circuit model 900, predicted relationships between the constituent frequency components, such as components 1002, 1004 and 1006, and the fault location can be established, as before. Such predicted relationships bay be established based on the one or more specific characteristics of the constituent frequency components. The predicted relationships of the three-phase network circuit model 900 can be established using matrix equation 11. In this example, the predicted relationships are also established based on fault location and additionally on fault impedance and the constituent frequency components, such as components 1002, 1004 and 1006, using the circuit model 900. The example matrix equation 11 is established using a modular matrix approach. The modular matrix approach involves splitting the circuit model 900 into individual sub-systems 906, 908 and 910. Sub- system 906 is defined as the section between measurement location 902 and expected fault 908. Sub-system 908 is the expected fault. Sub-system 910 is defined as the section between expect fault 908 and measurement location 904. A matrix equation can be derived from each sub-system and combined to form matrix equation 11. equation 11 The matrix equation 11 can be further rearranged to establish the predicted relationships as error minimisation problems shown by matrix equation 12. , and fault admittance ( ^^ _ி ) in equation 11 are replaced with line resistance and self inductance ( ^^ ^^ + ^^ ^^ ^^), mutual inductance ( ^^ ^^ ^^), and fault impedance ( ^^ _ி ) respectively. Matrix equation 12 shows six error minimisation problems derived based on the circuit model 900. However, only four of the error minimisation problems are relevant in establishing predicted relationships between the fault location and fault impedance of the expected fault 908, and the constituent frequency components, such as components 1002, 1004 and 1006. In other words, only four of the error minimisation problems are functions of fault impedance ( ^^ _ி ) and location ( ^^). The fault location ( ^^) also represents a normalised distance from the fault location to measurement location 902. One, some or all four of the relevant error minimisation problems or predicted relationships can be solved by constituent frequency components of one, some or all frequencies in Table 1000. In an example, Figure 11 shows Table 1100 where 32 estimated fault locations 1102 and estimated fault impedances 1104 are derived from the predicted relationships of matrix equation 12 and constituent frequency components of Table 1000. The fault location in this example may also be determined from the plurality of estimated fault locations 1102 using confidence measures associated with each of the estimated fault locations 1102. The fault location in this example may also, alternatively or additionally, be determined from a unified estimate fault location, e.g., that is a weighted mean of the estimated fault locations 1102. In the example shown in Figure 11, the fault location may be determined from the 32 estimated fault locations 1102 based on the sensitivity of a respective predicted relationship to a corresponding estimated fault location. As shown, estimated fault location 1106 has a minimum error that is well defined indicating that a corresponding predicted relationship used to derive estimated fault location 1106 has a high sensitivity. Therefore, estimated fault location 1106 may be selected to be the determined fault location. The fault location is determined to be a normalised distance of ^^ = 0.25 from the fault location to measurement location 902. The above examples were described relative to using the first signal portion, e.g. damped transient signal portion, of electrical signals to derive constituent frequency components to determine fault location. Additionally or alternatively, the fault location 106 may also be determined from constituent frequency components decomposed from the second signal portion, e.g. the steady state signal portion 302, of electrical signals associated with the longer-term fault condition. In some cases, a high impedance fault may be characterised by an arcing event at the fault location 106. Such arcing event create a time varying resistance that may be considered to dominate the characteristics of a fault current injection at the fault location. In other words, the nature of the fault itself can be approximated by – or may be observed to be similar to, or otherwise induce – current injection at the location of the fault 106 of the network 100 for example. Such current injection may create harmonic components in the electrical signals of the network that are measurable at the measurement locations of the network. These harmonic components can be represented by the constituent frequency components of the steady state signal portion of the electrical signal after the initial fault event. Further, the constituent frequency components may have frequencies that are multiples of the fundamental frequency of the network. For example, a fault injection current at a network with a fundamental frequency of 50 Hz could have harmonic frequencies at 100 Hz, 150 Hz and 200 Hz at the first few orders. The waveform of the steady state signal portion 302 of electrical signal 202 resulting from a fault may be a combination of waveforms of both odd harmonic 1202 and even harmonic 1204 portions as shown in Figures 12a and 12b for example. The odd harmonic portion 1202 of the steady state signal portion 302 has a symmetrical waveform and is constructable by odd harmonic components with frequencies at odd number multiples of the fundamental frequency at 150 Hz, 250 Hz and 350 Hz for example. The odd harmonic portion and/or the odd harmonic components of the electrical signal may be formed if an arc of an arcing event occurs at both polarities of network voltage in the line with the fault. The even harmonic portion 1204 has an asymmetrical waveform and is constructable by even harmonic components with frequencies at even number multiples of the fundamental frequency at 100 Hz, 200 Hz and 300 Hz for example. The even harmonic portion and/or the even harmonic components of the electrical signal may be formed if an arc of an arcing event occurs more at one (e.g. positive) polarity of network voltage than at the opposite (e.g. negative) polarity of network voltage in the line with the fault. Whether an arc occurs at both polarities of network voltage and/or whether an arc occurs more at one polarity than the other of the network may also be dependent on the geometry of the surfaces involved in the arcing event. Sharp points involved in the arcing event are more likely to form an arc at the positive part of the electrical cycle than smooth points in the arcing event. The constituent frequency components of the steady state portion 302 represent the odd and/or even harmonic components of the current injection and is dependent on the surfaces involved in an arcing event (e.g. surfaces on the line and surfaces in the surrounding environment of the network such as asphalt, concrete, grass or sod). The constituent frequency components decomposed from the steady state signal portion of the electrical signal may be characterised by the signal parameters of frequency, magnitude, and phase angle. It is not required for the constituent frequency components of the steady state signal portion 302 of electrical signals to be characterised by damping. Since the constituent frequency components derived from the steady state signal portion are not required to be characterised by damping, the constituent frequency components may be derived from electrical signals using Fast Fourier Transform (FFT) analysis, Fourier Series Expansion or other methods that do not factor in damping when determining the frequencies involved in a signal. Figure 13 shows another circuit model 1300 of a portion of an example three-phase electricity distribution network. Similar to circuit model 900, the three-phase network comprises three overhead lines 1312, 1314 and 1316. The three-phase network also includes measurement locations 1302 and 1304, from which electrical signal may be obtained that are associated with the power being distributed at the network. Electrical signals from any or all of the three overhead lines 1312, 1314 and 1316 can be obtained from either of the measurement locations 1302 or 1304. Each of the electrical signals can be a voltage signal or current signal. The circuit model 1300 also includes an expected fault 1308 represented as a current injection occurring on line 1314. The current injection may be representative of an arcing event at the fault location. The current injection may be characterised by one or more specific characteristics such as magnitude. The harmonic components (odd and/or even) of the current injection can influence the electrical signals of the network that are measurable at the measurement locations 1302 and/or 1304 of the network. Each of the electrical signals from the measurement locations 1302 and/or 1304 of the circuit model 1300 can be decomposed to constituent frequency components, in a similar manner to before. In this example, these one or more constituent frequency components are derived, at least in part, from the second signal portion, e.g., the steady state signal portion 302, of respective received electrical signals. In this particular example, these constituent frequency components represent the harmonic components (odd and/or even) of the current injection representative of an arcing fault. Figure 14 shows Table 14 with example constituent frequency components, including components 1402, 1404, and 1406, decomposed from the steady state signal portion 302 of all of the electrical signals in a three-phase network representing odd harmonic components. Figure 15 shows Table 15 with example constituent frequency components, including components 1502, 1504, and 1506, decomposed from the steady state signal portion of all of the electrical signals in a three-phase network representing even harmonic components. The constituent frequency components of the electrical signals shown in Tables 14 and 15 all have frequencies that are multiples of 50 Hz. This is because the harmonic components of the current injection have frequencies that are multiples of the 50 Hz fundamental frequency of the network. Using the circuit model 1300, predicted relationships between electrical signals, fault location and fault current injection can be established. As such, predicted relationships may be established between fault location and additionally on fault current injection and the constituent frequency components of those electrical signals, such as components 1402, 1404, 1406, 1502, 1504 and 1506. The predicted relationships may be established based on one or more specific characteristics of the constituent frequency components. In some examples, the predicted relationships may be established based on the specific characteristics (e.g. magnitude) of the fault current injection. This means that the predicted relationships may be usable to derive one or more estimated fault locations and one or more estimated fault current injection (e.g. magnitudes). The predicted relationships of the three- phase network circuit model 1300 can be established using matrix equation 13. Matrix equation 13 can be established using a similar approach to Matrix equation 11. The matrix equation 13 can be further rearranged to establish the predicted relationships as error minimisation problems shown by matrix equation 14. Matrix equation 14 shows six error minimisation problems derived based on the circuit model 1300. In this case, only three of the error minimisation problems are relevant in establishing predicted relationships between the fault location and the constituent frequency components. Four of the error minimisation problems are relevant in establishing predicted relationships between the fault current injection magnitude and the constituent frequency components. The values of the fault location and/or fault current injection magnitude that provides the minimum error in each relevant error minimisation problem may be used to identify an estimated fault location and/or estimated fault injection current magnitude. One or all four of the relevant error minimisation problems or predicted relationships can be solved by constituent frequency components of one or all frequencies in Tables 1400 and 1500, such as components 1402, 1404, 1406, 1502, 1504 and 1506.18 estimated fault locations 1602 and 24 estimated fault current injection magnitudes 1604 may be derived from the predicted relationships of matrix equation 14 and constituent frequency components of Tables 1400 or 1500 as shown in Table 1600 of Figure 16. In this example, the fault location may be determined from the plurality of estimated fault locations 1602 using confidence measures associated with each of the estimated fault locations 1602. The fault location may, alternatively or additionally, be determined from a unified estimate fault location that may be a weighted mean of the estimated fault locations 1602. Alternatively or additionally, the predicted relationships established as error minimisation problems may be combined and solved by constituent frequency components of one or all frequencies in Tables 1400 and 1500, such as components 1402, 1404, 1406, 1502, 1504 and 1506. The combination of the relevant error minimisation problems may be used to derive a unified error that is the sum of errors of all the relevant error minimisation problems divided by the number of relevant minimisation problems. In this example, the values of the fault location and fault injection current magnitude that provide the minimum unified error may be used to identify an estimated fault location and estimated fault current injection (e.g. fault current injection magnitude). Table 1700 of Figure 17 shows the estimated fault locations 1702 derived using the unified error of the combination of predicted relationships. The estimated fault locations are respective to the frequencies of constituent frequency components in Tables 1400 and 1500. Out of the six estimated fault locations 1702, a fault location may then be determined from the plurality of estimated fault locations 1702 using confidence measures associated with each of the estimated fault locations 1702. The fault location may, alternatively or additionally, be determined from a unified estimate fault location that may be a weighted mean of the estimated fault locations 1702. In the example the fault location may be determined from the six estimated fault locations 1702 based on the sensitivity of a respective predicted relationship to a corresponding estimated fault location. As shown, estimated fault location 1704 has a minimum error that is well defined indicating that the combination of predicted relationships used to derive estimated fault location 1704 has a high sensitivity. Therefore, estimated fault location 1704 may be selected to be the determined fault location. The fault location is determined to be a normalised distance of ^^ = 0.25 from the fault location to measurement location 1302. Figure 18 shows another circuit model 1800 of a portion of an example three-phase electricity network. The circuit model 1800 includes three overhead lines 1812, 1814 and 1816 with measurement locations 1802 and 1804 that may measure electrical signals from each of the three overhead lines 1812, 1814 and 1816. Further, capacitances between the lines and between the lines and the ground are demonstrated by capacitances 1818, 1820, 1822 are considered within the circuit model 1800. The circuit model 1800 also includes an expected fault 1808 with an impedance 1824 that is occurring on line 1814. Alternatively, the expected fault 1808 may modelled with a current injection occurring on line 1814 instead of impedance 1824. Each of the electrical signals from the measurement locations 1802 and/or 1804 of the circuit model 1800 can be decomposed to constituent frequency components, in a similar manner to before. Further, predicted relationships between the constituent frequency components and the fault location can be established using circuit model 1800, as before. A modular matrix approach may again be used to derive predicted relationships as a matrix equation. The modular matrix approach in this case involves splitting the circuit model 1800 into five individual sub-systems 1806, 1808, 1810, 1826 and 1828. Where the capacitances between measurement location 1802 and expected fault 1808 is represented by sub-system 1826 and the capacitances between measurement location 1804 and the expected fault 1808 is represented by sub-system 1828. Sub-system 1808 represents the expected fault including capacitances. The line impedances between measurement locations 1802, 1804 and the expected fault 1808 are represented by sub-systems 1806 and 1810 respectively, similar to the line impedances 906 and 910 of circuit model 900. Similar to the above, six error minimisation problems may also be derived. In this case, where the circuit model 1800 also includes the effects of capacitance, all six error minimisation problems are relevant in establishing predicted relationships between the fault location and fault impedance of the expected fault 1808, and the constituent frequency components. In the case where the expected fault 1808 is modelled with a current injection occurring on line 1814 instead of impedance 1824, predicted relationships between the fault location and fault current injection magnitude of the expected fault 1808 are established. One, some, or all six of the relevant error minimisation problems or predicted relationships can be solved by constituent frequency components derived from the electrical signals from measurement locations 1802 and/or 1804 of the circuit model 1800, to derive a plurality of estimated fault locations. The constituent frequency components may be derived from the transient signal portion 204 of the electrical signals. In the case where the expected fault 1808 is modelled with a current injection occurring on line 1814 instead of impedance 1824, the constituent frequency components may be derived from the steady-state signal portion 302 of the electrical signals. Alternatively or additionally, the predicted relationships may be established as a combination of all six error minimisation problems to derive a unified error. The values of the fault location that provide the minimum unified error may be used to identify an estimated fault location. The fault location may be determined from the plurality of estimated fault locations using confidence measures associated with each of the estimated fault locations similar to above. The fault location may, alternatively or additionally, be determined from a unified estimate fault location that may be a weighted mean of the estimated fault locations similar to above. Figure 19 show examples of the present method of for determining a fault location in an electricity distribution network in use. Electrical signals of a fault event in the electricity distribution network are received as fault data 1902. A signal classifier 1904 windows electrical signals into the damped transient signal portions 1906 and the steady-state signal portions 1922 of the electrical signals. One or more constituent frequency components can be decomposed from each of the damped transient signal portions 1906 using the Matrix Pencil 1908. Alternatively, Prony’s method can be used as above. Constituent frequency components that are from one electrical signal having the same frequencies as constituent frequency components decomposed from another of the electrical signals may be selected at step 1910. Constituent frequency components of larger energy, initial magnitude and/or low damping may also be selected at step 1910. A circuit model 1912 is provided as demonstrated by circuit models 500, 900 and 1800. Estimates of the fault location and impedance 1914 are derived from predicted relationships between the constituent frequency components, and fault location and fault impedance established from the circuit model. Constituent frequency components of multiple frequencies can be used to solve multiple predicted relationships to derive multiple estimates of the fault location and impedance 1916 in the network. The multiple fault estimates are weighted and assessed for confidence 1918 based on the quality of the constituent frequency components and/or based on the sensitivity of expected relationships. A unified fault location and impedance 1920 may be determined based on the weighted fault estimates. Alternatively, the fault location and/or impedance can be determined based on the estimates with the highest confidence level. One or more constituent frequency components can also be decomposed from each of the steady state signal portions 1922 using Fourier Transform1924. Constituent frequency components are selected at step 1910. Constituent frequency components that are from one electrical signal having the same frequencies as constituent frequency components decomposed from another of the electrical signals may be selected at step 1910. Constituent frequency components of larger energy, initial magnitude and/or low damping may also be selected at step 1910. A circuit model 1926 is provided as demonstrated by circuit model 1300. Estimates of the fault location and current injection magnitude 1928 are derived from predicted relationships between the constituent frequency components, and the fault location and fault current injection magnitude established from the circuit model. These estimates may also be assessed for confidence to determine the fault location and/or current injection magnitude based on the estimates with the highest confidence level. These estimates may also be weighted to produce a unified fault location and/or current injection magnitude. By determining the fault location, a network operator may allocate resources to fixing the fault at the fault location. Therefore, improving the consistency of electricity distribution and reducing risks caused by the fault. It will be readily appreciated that the methodology described in the examples above may be implemented based on real time monitoring, for example. This means that electrical signals associated with the power being distributed at the network may be directly received from the measurement locations on the network to detect an occurrence of a fault at the network and determine a location of the fault. It will also be readily appreciated that electrical signals associated with the power being distributed at the network may be recorded, e.g., to a database from the measurement locations, prior to being received to detect an occurrence/location of a fault at the network. The electrical signals may be received and decomposed at the network and used to determine a fault location prior to being indicated to a network operator. Alternatively, the electrical signals may be received by a network operator to be decomposed and used to determine a fault location, as will be appreciated. Further, while in the above example both current and voltage signals have been used in other examples only of those signal types may be used, or indeed derivatives of some or both of those signals may be used. A processor, on the network or at the location of the network operator for example, may be configured or programmed to implement the methodology described in the examples above. Furthermore, the methodology described in the examples above may be implemented by hardware, software, firmware, middleware, microcode, or any combination thereof. When implemented in software, firmware, middleware or microcode, program code or code segments that may be used to perform the methodology described in the examples above may be stored in a machine-readable medium such as a storage medium or other storage(s). A code segment may represent a procedure, a function, a subprogram, a program, a routine, a subroutine, a module, a software package, a class, or any combination of instructions, data structures, or program statements. A code segment may be coupled to another code segment or a hardware circuit by passing and/or receiving information or data such as electrical signals, parameters of constituent frequency components and values of fault location, impedance or current injection magnitude. Information or data, etc. may be passed, forwarded, or transmitted via any suitable means including memory sharing, message passing, token passing, network transmission, etc. In the foregoing, a storage medium may represent one or more devices for storing data, including read-only memory (ROM), random access memory (RAM), magnetic disk storage mediums, optical storage mediums, flash memory devices EPROM memory, EEPROM memory, registers, hard disk, a removable disk, a CD- ROM, or any other form of storage medium known in the art for storing information. The storage medium may be located on the network 100 or at the location of the network operator for example. The terms "machine readable medium" and "computer readable medium" include, but are not limited to portable or fixed storage devices, optical storage devices, and/or various other mediums capable of storing, containing or carrying instruction(s) and/or data. The methodology described in the examples above may be implemented or performed with a general purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic component, discrete gate or transistor logic, discrete hardware components, or any combination thereof. A general purpose processor may be a microprocessor, but in the alternative, the processor may be any conventional processor, controller, microcontroller, circuit, and/or state machine. A processor may also be implemented as a combination of computing components, e.g., a combination of a DSP and a microprocessor, a number of microprocessors, one or more microprocessors in conjunction with a DSP core, or any other such configuration. The foregoing description includes examples of methods and systems for determining fault locations in an electricity network, or the like. Modifications may be made thereto without departing from the scope of the invention.