Determining a value of an electric power flow characteristic of an electric power grid

Abstract

A method, apparatus, and system for determining a value of an electric power flow characteristic (e.g. grid inertia or topology) of an electric power grid. The grid comprises a plurality of power units each configured to consume and / or provide power to the grid. A change in power causes a respective change in a value of a first parameter of electric power flow (e.g. frequency or voltage phase angle) in the grid. A computing system determines a first plurality of sets of event data generated by a respective first plurality of data generating devices, each associated with a respective power unit. Each set of event data is indicative of i) a change in power by the associated power unit and ii) a time of the change, within a first time window. The computing system determines grid response data indicating values of the first parameter during the first time window, and determines the value of the electric power flow characteristic based on the determined sets of event data and the determined grid response data.

Description

Technical Field The present invention relates to a method, apparatus, and system for determining a value of an electric power flow characteristic of an electric power grid. Background Generated electrical power is distributed to consumers via an electricity distribution network or electric power grid. Electric power grids typically operate at a nominal grid frequency that is uniform throughout a synchronous area of the grid. For example, the UK mains supply nominally operates at 50 Hz AC. Grid operators are usually obliged to maintain the grid frequency to within predefined limits. For example, the UK electricity system should be kept within 1% of the nominal 50 Hz grid frequency. Large capacity power stations, such as nuclear or fossil fuel power stations, use spinning generators with relatively massive rotating parts that are powered to rotate at relatively high speeds at some multiple of the nominal grid frequency (so called spinning generation). In the course of their normal operation, these spinning generators store relatively large amounts of kinetic energy in the weight and speed of their spinning turbines. Small capacity providers, such as wind or solar farms often use DC-connected inverters to supply power to the grid at the nominal grid frequency, and typically store a much smaller amount of kinetic energy, or even no kinetic energy at all. A change in the balance between provision and consumption of electrical power to the grid (for example, if the total amount of provision cannot meet consumption during high demand periods, or if the provision from a power generator or interconnect fails) leads to a change in the load on the generators. This results in a change of the rotational speed of the spinning generators and a corresponding change in the operating frequency of the grid. Grid inertia is a measure of the amount of kinetic energy stored in the electric power grid and influences the rate at which the operating frequency of the grid changes in response to a change in balance of power provision and consumption in the grid. The grid inertia depends on the number and type of both generators and loads connected to the grid. For example, spinning generation such as in nuclear or fossil fuel power stations generally contribute to the grid inertia whereas renewable energy production methods like wind and solar generally do not. Additionally, large rotating power consumption machines such as used in factories generally contribute to inertia whereas typical domestic consumption, such as loading laptop batteries, generally does not. Due to the different mix of generators and loads in different areas of the electrical power grid, different areas of the electrical power grid can have different inertia values. The rate of change of grid frequency for a given change in power balance in a grid area with high inertia is less than it is in a grid area with low inertia. An estimate of grid inertia can be useful for grid operators in tailoring strategies for mitigating the effects of sudden changes in power balance, for example. It is also desirable to estimate values for electric power flow characteristics of the grid other than grid inertia, such as values indicative of a topology or operational state of the grid, for example. These can inform optimisation of the configuration and / or operation of the grid. Summary According to a first aspect of the present invention, there is provided a method of determining a value of an electric power flow characteristic of an electric power grid, the grid comprising a plurality of power units each configured to consume electric power from and / or provide electric power to the electric power grid, a change in consumption of electric power from or provision of electric power to the electric power grid by any particular one of the power units causing a respective change in a value of a first parameter of electric power flow in the electric power grid, the method comprising, at a computing system: determining a first plurality of sets of event data generated by a respective first plurality of data generating devices, each data generating device being associated with a respective one of a first plurality of the power units, each set of event data indicating i) a change in consumption of electric power from or provision of electric power to the electric power grid by the associated power unit and ii) a time of the change, wherein the time of the change indicated by each of the determined first plurality of sets of event data is within a first time window; determining grid response data, the grid response data indicating values of the first parameter of electric power flow in the grid as a function of time during the first time window; and determining, based the determined sets of event data, and based on the determined grid response data, the value of the electric power flow characteristic of the electric power grid. Optionally, the method comprises, for each of one or more of the first plurality of sets of event data: categorising the set of event data as belonging to a first category, responsive to determining that the change indicated by the set of event data results from a power flow change event that is correlated with one or more further power flow change events, the one or more further power flow change events each resulting in a change in consumption of electric power from or provision of electric power to the electric power grid by a respective one or more power units that are not of the first plurality of power units; and determining the value of the electric power flow characteristic of the electric power grid is based additionally on the categorisation of the one or more sets of event data as belonging to the first category. Optionally, the method comprises: excluding the one or more sets of event data categorised as belonging to the first category from being used in the determination of the value of the electric power flow characteristic of the electric power grid. Optionally, the method comprises parametrising the change indicated by the one or more sets of event data categorised as belonging to the first category with a second parameter, wherein the second parameter represents the proportion, of the total change in consumption or provision of power resulting from the correlated power flow change events, that the change indicated by the one or more sets of event data categorised as belonging to the first category represents; and determining the value of the electric power flow characteristic is based on the parameterised change indicated by the one or more sets of event data categorised as belonging to the first category. Optionally, the method comprises determining a value of the second parameter concurrently with the determination of the value of the electric power flow characteristic. Optionally, categorising the one or more of the first plurality of sets of event data as belonging to the first category comprises: analysing the first plurality of sets of event data to identify the one or more sets of event data. Optionally, analysing the first plurality of sets of event data to identify the one or more sets of event data comprises: for each of a plurality of second time windows within the first time window: determining a sum of the changes indicated by respective sets of event data that indicate a time within the second window; in response to a magnitude of the sum being larger than a first threshold value, categorising the sets of event data that indicate a time within the second time window as belonging to the first category. Optionally, determining the value of the electric power flow characteristic of the electric power grid comprises fitting the first plurality of sets of event data, or first data derived from the first plurality of sets of event data , to the grid response data, or second data derived from the grid response data. Optionally, determining the value of the electric power flow characteristic comprises: averaging the changes indicated by each of two or more of the first plurality of sets of event data or data derived therefrom; averaging a plurality of portions of the grid response data or data derived therefrom, wherein each portion of the response data or data derived therefrom corresponds to values of the first parameter within a third time window beginning at the time indicated by a respective one of the two or more of the first plurality of sets of event data; and determining the value of the electric power flow characteristic based on the averaged changes or data derived therefrom and the averaged portions of response data or data derived therefrom. Optionally, the grid comprises one or more compensating devices configured to provide power to or consume power from the electric power grid in response to a change in a value of the first parameter in order to compensate the change in the value of the first parameter, and wherein the third time window is set to have a duration less than a response time in which the one or more compensating devices provide power to or consume power from the electric power grid in response to the change in the value of the first parameter. Optionally, determining the first plurality of sets of event data comprises: selecting the first plurality of sets of event data from a second plurality of sets of event data based on one or more attributes associated with each of one or more of the second plurality of sets of event data. Optionally, the one or more attributes comprise a location of the power unit associated with the set of event data, and determining the first plurality of sets of event data comprises: selecting the first plurality of sets of event data from the second plurality of sets of event data based at least in part on the respective locations associated therewith. Optionally, the first plurality of sets of event data are selected based at least in part on the respective locations associated therewith each belonging to a common geographical region or region of the electric power grid. Optionally, each set of event data is indicative of the location of the associated power unit. Optionally, the one or more attributes comprise one or more of a magnitude, rate, and duration of the change indicated by the set of event data, and determining the first plurality of sets of event data comprises: selecting the first plurality of sets of event data from the second plurality of sets of event data based at least in part on one or more of the respective magnitudes, rates, and durations of the respective changes indicated thereby. Optionally, the first plurality of sets of event data are selected based at least in part on one or more of the respective magnitudes, rates, and durations of the respective changes indicated thereby being greater than a threshold magnitude, threshold rate, and threshold duration, respectively. Optionally, determining the grid response data comprises: selecting the grid response data from a plurality of sets of grid response data based at least in part on a location at which the values of the first parameter indicated thereby were measured. Optionally, the first parameter of electric power flow in the grid is a frequency of electricity flowing in the grid and the electric power flow characteristic of the electric power grid is inertia or the first parameter of electric power flow in the grid is a difference between voltage phase angle at a first grid location and a second grid location and the electric power flow characteristic of the electric power grid is indicative of a current topology or operational state of the electric power grid. Optionally, the method comprises: receiving the first plurality of sets of event data from the respective data generating devices over a network. According to a second aspect of the present invention, there is provided a method of providing one or more sets of event data for determining a value of an electric power flow characteristic of an electric power grid, the grid comprising a plurality of power units each configured to consume electric power from and / or provide electric power to the electric power grid, the method comprising, at a data generating device associated with a respective power unit: determining a change in consumption of electric power from or provision of electric power to the electric power grid by the associated power unit; determining that one or more of a magnitude, rate, and duration of the change is greater than a threshold magnitude, threshold rate, and threshold duration, respectively; in response to determining that one or more of a magnitude, rate, and duration of the change is greater than a threshold magnitude, threshold rate, and threshold duration, respectively: generating a set of event data indicating i) the change in consumption of electric power from or provision of electric power to the electric power grid by the associated power unit and ii) a time of the change; and sending the set of event data to a computing system over a network. Optionally, the set of event data further indicates a location of the associated power unit. Optionally, determining the change in consumption of electric power from or provision of electric power to the electric power grid by the associated power unit comprises: measuring and / or controlling the change in the consumption of electric power from or provision of electric power to the electric power grid by the associated power unit. According to a third aspect of the present invention, there is provided apparatus configured to perform the method according to the first aspect or the second aspect. According to a fourth aspect of the present invention, there is provided a computer program comprising instructions which, when executed by a computer, cause the computer to perform the method according to the first aspect or the second aspect. According to a fifth aspect of the present invention, there is provided a system comprising an apparatus configured to perform the method according to the first aspect and the data generating devices; an apparatus configured to perform the method according to the second aspect and the computing system; or an apparatus configured to perform the method according to the first aspect and an apparatus configured to perform the method according to the second aspect. Further features and advantages of the invention will become apparent from the following description of preferred embodiments of the invention, given by way of example only, which is made with reference to the accompanying drawings. Brief Description of the Drawings Figure lisa flow diagram illustrating a method according to an example; Figure 2 is a schematic diagram illustrating an electric power grid according to an example; Figure 3A is a schematic diagram illustrating a plot of a power P provided to the electric power grid by a power unit, as a function of time t, according to an example; Figure 3B is a schematic diagram illustrating a plot of grid frequency f as a function of time t over the same time period as in Figure 3A, according to an example; Figure 4 is a schematic diagram illustrating sets of event data for multiple power units in a certain time period and a plot of the grid frequency in the time period, according to an example; Figure 5A is a schematic diagram illustrating a plot of an average of a plurality of sets of event data, according to an example; Figure 5B is a schematic diagram illustrating a plot of an of average of the grid frequency over a plurality of time windows corresponding to the plurality of sets of event data of Figure 5 A, according to an example; Figure 6 is a schematic diagram illustrating sets of event data for multiple power units in a certain time and a plot of the sum of power changes of the event data over each of a plurality of sub-time periods; Figure 7 is a schematic diagram illustrating regions of an electric power grid according to an example; Figure 8 is a flow diagram illustrating a method according to an example; Figure 9 is a schematic diagram illustrating a system according to an example; Figure 10 is a schematic diagram illustrating apparatus according to an example; Figure 11 is a graph illustrating a plot of grid frequency as a function of time according to an example model; Figure 12 is a bar chart illustrating the number of events for each power change according to an example simulation using the model; Figure 13 is a graph illustrating a plot of frequency of a function of time according to the example simulation; and Figure 14 is a histogram illustrating the number of times a certain values of grid inertia were determined according to the example simulation. Detailed Description Referring to Figure 1, there is illustrated a method of determining a value of an electric power flow characteristic of an electric power grid, according to an example. As described in more detail below with reference to Figures 2, the grid 200 comprises a plurality of power units 219 each configured to consume electric power from and / or provide electric power to the electric power grid 200. As described in more detail below with reference to Figures 3 A to 3C, a change in consumption of electric power from or provision of electric power to the electric power grid 200 by any particular one of the power units 219 causes a respective change in a value of a first parameter f, <Pij of electric power flow in the electric power grid 200. In broad overview, the method comprises, at a computing system 205: in step 102, determining a first plurality of sets of event data generated by a respective first plurality of data generating devices 218, each data generating device being associated with a respective one of a first plurality of the power units 219, each set of event data indicating i) a change in consumption of electric power from or provision of electric power to the electric power grid 200 by the associated power unit 219 and ii) a time of the change, wherein the time of the change indicated by each of the determined first plurality of sets of event data is within a first time window T; in step 104, determining grid response data, the grid response data indicating values of the first parameter f, (pij of electric power flow in the grid 200 as a function of time during the first time window T; and in step 106, determining, based the determined sets of event data, and based on the determined grid response data, the value of the electric power flow characteristic of the electric power grid 200. This may allow for a value of the electric power flow characteristic (such as grid inertia) to be determined based on changes in power consumed from and / or provided to the grid 200 provided by events that occur in the electric power grid 200, as part of the usual, day-to-day use of the grid, for example. For example, such events may include a kettle or boiler turning on. Accordingly, there may be no need for the manufacture, deployment, operation and / or control of a dedicated power modulator to inject power pulses into the grid. Sets of event data indicative of the change in power and the time of these events are generated by data generating devices, such as smart meters or smart plugs, that are associated with the power units 219. Accordingly, by determining a value of the electric power flow characteristic on the basis of the determined sets of event data and the grid response data (indicative of grid frequency f, for example), the value of the electric power flow characteristic can be determined in a cost effective and efficient manner. Improved electric power flow characteristic value determination may therefore be provided. As mentioned, the method is for determining a value of an electric power flow characteristic of an electric power grid 200. Referring now to Figure 2, there is illustrated an electric power grid 200 according to an example. Supply of electricity from providers such as power stations, to consumers, such as domestic households and businesses, typically takes place via an electricity distribution network or electric power grid 200. In the example of Figure 2, the electric power grid (hereinafter ‘grid’) 200 comprises a transmission grid 202 and a distribution grid 204. The transmission grid 202 is connected to power generators 206, which may be nuclear plants or gas-fired plants, for example, from which it transmits large quantities of electrical energy at very high voltages (typically of the order of hundreds of kV), over power lines such as overhead power lines, to the distribution grid 204. These power generators 206 may also include larger-scale wind farms and / or solar farms. The transmission grid 202 is linked to the distribution grid 204 via a transformer 208, which converts the electric supply to a lower voltage (typically of the order of 50kV) for distribution in the distribution grid 204. The distribution grid 204 is connected via substations 210 comprising further transformers for converting to still lower voltages to local networks which provide electric power to power consuming devices connected to the electric power grid 200. The local networks may include networks of domestic consumers, such as a city network 212, that supplies power to domestic appliances within private residences 213 that draw a relatively small amount of power in the order of a few kW. Private residences 213 may also use electric vehicles, battery storage, heat pumps, air conditioning devices and photovoltaic devices 215 to provide relatively small amounts of power for consumption either by appliances at the residence or for provision of power to the grid. The local networks may also include industrial premises such as a factory 214, in which larger appliances operating in the industrial premises draw larger amounts of power in the order of several kW to MW. The local networks may also include networks of smaller power generators such as battery storage, solar and wind farms 216 that provide power to the electric power grid. Although, for conciseness, only one transmission grid 202 and one distribution grid 204 are shown in Figure 2, in practice a typical transmission grid 202 supplies power to multiple distribution grids 204 and one transmission grid 202 may also be interconnected to one or more other transmission grids 202. Electric power flows in the electric power grid 200 as alternating current (AC), which flows at a system frequency, which may be referred to as a grid frequency (typically 50 or 60 Hz, depending on country). The electric power grid 200 operates at a synchronized frequency so that the frequency is substantially the same at each point of the grid. The grid 200 may include one or more direct current (DC) interconnects 217 that provide a DC connection between the electric power grid 200 and other electric power grids. Typically, the DC interconnects 217 connect to the typically high voltage transmission grid 202 of the electrical power grid 200. The DC interconnects 217 provide a DC link between the various electric power grids, such that the electric power grid 200 defines an area which operates at a given, synchronised, grid frequency that is not affected by changes in the grid frequency of other electric power grids. For example, the UK transmission grid is connected to the Synchronous Grid of Continental Europe via DC interconnects. The electric power grid 200 may comprise one or more measurement devices 220 for measuring a value of a first parameter of electric power flow in the electric power grid 200, such as grid frequency f and / or voltage V. For example, the measurement device 220 may comprise a phasor measurement unit (PMU), which may be configured to measure one or more of a frequency, voltage, current, power, reactive power, and phase angle of electricity flowing in the electric power grid. As another example, the measurement device 220 may comprise an extensible measurement unit (XMU), which may be configured to collect voltage data from the grid and apply signal processing to determine one or more of a frequency, voltage, and voltage angle of electricity flowing in the electric power grid. As another example, the measurement device 220 may comprise other types of measurement devices such as a Power Meter or a Digital Fault Recorder (DFR). A DFR may be configured to sample and record first parameter values including but not limited to harmonics, frequency, and voltage levels of electricity flowing in the electric power grid. For example, a DFR may sample data captured by, for example, protection relays of the grid. As another example, the one or more measurement devices 220 may comprise a measurement device configured to measure first parameter values for power lines, such EMF (electromagnetic forces) and dynamic line rating measurement devices such as thermal, vibration, current sensors and radar and / or optical sensors. For example, a thermal sensor may measure the temperature of a power line, which may be indicative of the current flowing in the power line. As another example, optical sensors may measure current and / or voltage of electrical energy flowing in a power line, for example making use of the Pockels effect. Other measurement devices may be used, such as measurement devices configured to measure a synchrophasor and / or a Point on Wave of electricity flowing in the electric power grid. Other measurement devices may be used. The grid 200 comprises a plurality of power units 219. Each power unit 219 is configured to consume electric power from and / or provide electric power to the electric power grid 200. For example, each of the power generators 206, residences 213, photovoltaic devices 215, factory 214, and wind farms 216 may be an example of a power unit 219. Indeed, in examples, any device or other grid asset, or combination or subset of such grid assets, that consume electric power from and / or provide electric power to the electric power grid 200, may be a power unit 219. Each power unit 219 of a first plurality of the power units 219 is associated with a data generating device 218 configured to generate sets of event data. Each set of generated event data indicates i) a change in consumption of electric power from or provision of electric power to the electric power grid 200 by the associated power unit 219 and ii) a time of the change. For example, each set of event data may indicate i) an amplitude of a change in consumption of electric power from or provision of electric power to the electric power grid 200 by the associated power unit 219 and ii) a time of the change. For example, the event data may include the amplitude of the change, for example in the form of an amplitude value. The time of the change may be the time at which the change occurred, for example in the form of a timestamp associated with the change. As another example, the event data may include a series of values of power consumed or provided by the associated power unit 219 as a function of time over a time period including a change in the power consumed or provided by the associated power unit. In examples, as described in more detail below, the change may be measured by the data generating device 218 and / or may be a change that the data generating device 218 has controlled the associated power unit 219 to make. In examples, each data generating device 218 may be configured to determine that a change in provision or consumption of power by the associated power unit 219 has occurred, and, responsive to the determination, generate event data indicating i) the change and ii) the time of the change. As described in more detail below, in some examples, the generation of the event data by the data generating device (and / or the sending the generated event data to the computing system 205) may be responsive to a determination by the data generating device 218 that one or more of a magnitude, rate, and duration of the change is greater than a magnitude threshold, rate threshold, and duration threshold, respectively. In examples, each data generating device 218 is configured to send the sets of event data to the computing system 205. For example, the data generating devices 218 may send the event data to the computing system 205 over a computer network (not shown in Figure 2, but see network 900 in Figure 9 described in more detail below) such as the Internet. In examples, the data generating devices 218 may be Intemet-of-Things (IOT) devices. In examples, the data generating device 218 may comprise a power measurement device configured to measure or meter electric power consumed from the grid 200 by the associated power unit and / or provided by the associated power unit 219 to the grid 200. The power measurement device may measure the power consumed and / or provided as a function of time, and the data generating device 218 may provide data representing this measurement to a computing system, such as the computing system 205, for example over the Internet. For example, a power unit 219 may be a domestic appliance such as a kettle, boiler, or electric radiator. A data generating device 218 associated with the power unit 219 may be a smart plug via which the power unit 219 is connected to the mains supply. The smart plug may monitor the consumption of electric power by the power unit 219 as a function of time, and provide this data (or data derived therefrom) to a computing system, such as the computing system 205, over the Internet. As another example, a power unit 219 may represent one or more power consumers and / or generators, such as those of a particular household 213, factory 214 or power generator 206, 216. A data generating device 218 associated with the power unit 219 may be a smart meter that measures electric power consumed by the power unit 219 from the grid 200 and / or electric power provided to the grid 200 by the power unit 219. The smart meter may monitor the consumption and / or provision of electric power by the power unit 219 as a function of time, and provide this data (or data derived therefrom) to a computing system, such as the computing system 205. As another example, a data generating device 218 associated with the power unit 219 may be a Phasor Measurement Unit (PMU) that measures electric power consumed by the power unit 219 from the grid 200 and / or electric power provided to the grid 200 by the power unit 219. The PMU may monitor the consumption and / or provision of electric power by the power unit 219 as a function of time, and provide this data (or data derived therefrom) to a computing system, such as the computing system 205. In examples, the data generating device 218 may comprise a control device configured to control a change in power consumption and / or provision of an associated power unit 219. For example, the control device may send a control signal to an associated power device 219 to change the power that the power unit 219 consumes from or provides to the grid 200 by a certain amount at a certain time. In these examples, the data generating device 218 may infer, from the control signal, that the power unit 219 changes the power by the certain amount at the certain time. The data generating device 218 may generate a set of event data indicative of the certain change and the certain time of the change, and provide this data to a computing system, such as the computing system 205, for example over the Internet. A change in consumption of electric power from or provision of electric power to the electric power grid 200 by any particular one of the power units 219 causes a respective change in a value of the first parameter f, (ptj of electric power flow in the electric power grid 200. For example, the change in power by a power unit 219 may cause a change in the grid frequency f and / or a change in the difference (pij between voltage phase angle at a first grid location i and voltage phase angle at a second grid location j. Other examples are possible. Referring to Figure 3A, there is illustrated an example change in power P provided to the grid 200 by a power unit 219. At time to, the power unit 219 is neither consuming power from nor providing power to the grid 200. This continues until time ti, at which point the power unit 219 provides power of amplitude A to the grid 200. The power unit 219 continues to provide power at amplitude A up until and beyond time t2. In this example, the change in power at time ti is instantaneous or practically instantaneous. Accordingly, the power P as a function of time t takes the form of a step function. In this example, the amplitude of the change in power consumed from and / or provided to the grid 200 is +A (where the ‘+’ represents that the change is an increase in power provided to the grid 200), and the time of the change is ti. The change in power results from a power flow change event, also referred to herein as an event. For example, the event may be that the power unit 219 has undergone a change in operating state, such as being switched on or off, that results in a change in power consumed by the power unit 219 from the grid 200 or provided to the grid 200 by the power unit 219. In this example, a data generating device 218 associated with the power unit 219 may generate a set of event data associated with the event. The set of event data may be indicative of i) the change in consumption of electric power from or provision of electric power to the electric power grid 200 by the associated power unit 219 resulting from the event and ii) a time of the change. Accordingly, in this example, the set of event data may be data indicative of the change +A and the time of the change ti. Referring to Figure 3B, there is illustrated an example of the respective change in grid frequency f caused by the power change of Figure 3A. At time to, the grid frequency f has an initial value fo. For example, the initial frequency fo may be the nominal frequency which, for example, in the UK would be 50Hz (although in other examples the initial frequency fo may be a value other than the nominal frequency). The frequency remains at the initial value fo until time ti, where the value of the grid frequency is fi = fo. At time ti, due to the increase in power +A provided to the grid 200, the grid frequency f increases. The grid frequency f continues to increase until, at time t2, it settles as a higher value f2. The increase in grid frequency f at time ti is not instantaneous. The rate of change of the frequency f resulting from the change in power P is dependent on the inertia of the grid 200. Accordingly, a value for the grid inertia can be determined based on the change A in power P and the rate of the resulting change of the grid frequency f. In examples, a change in consumption of electric power from or provision of electric power to the electric power grid 200 may cause a respective change in a value of a parameter of electric power flow in the electric power grid 200 other than frequency f (such a change in the difference <ptj between voltage phase angle at a first grid location i and voltage phase angle at a second grid location j, as described in more detail below), and characteristics other than grid inertia may be determined on the basis of the power change and a change in value of the first parameter (such as a value indicative of a current topology or operational state of the grid, as described in more detail below). As mentioned, the method of Figure 1 comprises, in step 102, determining a first plurality of sets of event data generated by a respective first plurality of data generating devices 218, each data generating device being associated with a respective power unit 219. Each set of event data is indicative of i) a change in consumption of electric power from or provision of electric power to the electric power grid 200 by the associated power unit 219 and ii) a time of the change. The time of the change indicated by each of the determined first plurality of sets of event data is within a first time window. In examples, the data generating devices 218 may send the sets of event data to the computing system 205, and the computing system 205 may store the sets of event data in a first database, such as a time series database (see e.g. database 902 in Figure 9). In examples, determining the first plurality of sets of event data may comprise retrieving, from the first database, sets of event data that indicate a time within the first time window. For example, the first time window may have a start time, such as 14:00 GMT and an end time, such as 14:30 GMT, and determining the first plurality of sets of event data may comprise retrieving sets of event data that indicate a time that is after the start time and before the end time. In examples, the first time window may be, for example, tens of minutes, such as 30 minutes. As mentioned, the method of Figure 1 comprises, in step 104, determining grid response data, the grid response data indicating values of the first parameter of electric power flow in the grid 200 as a function of time during the first time window. For example, the values of the first parameter may be obtained from, or derived from the output of, one or more of the measurement devices 220. For example, a PMU may measure a frequency of electricity flowing in the electric power grid. In examples, the one or more measurement devices 220 may send measured values of the first parameter as a function of time to the computing system 205. For example, the one or more measurement devices 220 may send a stream of data points to the computing system, each data point including a value of the first parameter and a time at which the value was measured. The computing system 205 may store the measured values in a second database, such as a time series database (see e.g. the second database 904 of Figure 9). In examples, determining the grid response data may comprise retrieving, from the second database, measured values of the first parameter that indicate a measurement time within the first time window. For example, determining the grid response data may comprise retrieving measured values of the first parameter having a measurement time that is after the start time and before the end time of the first time window. As an example, referring to Figure 4, there is illustrated a plurality of sets of event data Ei to Eio in a certain time period and a plot of the grid frequency f in the time period. Each set of event data Ei to Eio is indicative of a change in power (represented by the size and sign of the illustrated step function) and the time of the change (represented by the location of the illustrated step function along the time axis t). Two of the sets of event data Ei and Eio have been generated by a first data generating device DEV. 1 associated with a first power unit 219 (not shown in Figure 4). Three of the sets of event data E2, Ee and Es have been generated by a second data generating device DEV. 2 associated with a second power unit 219 (not shown in Figure 4). Two of the sets of event data Es and E9 have been generated by a third data generating device DEV. 3 associated with a third power unit 219 (not shown in Figure 4). Three of the sets of event data E4, Es and E? have been generated by an nth data generating device DEV. N associated with an nth power unit 219 (not shown in Figure 4). It will be appreciated that while only four data generating devices and ten sets of event data are shown in Figure 4 for clarity, in examples there may be hundreds or thousands or more of data generating devices and hundreds or thousands or more of sets of event data. There is a first time window T, which starts at time t3 and ends at time t4. Each of the sets of event data Ei to Ee indicates a time that is within a first time window T, whereas each of the sets of event data E? to Eio indicates a time that is outside of the first time window T. In examples, the sets of event data Ei to Ee are determined as the first plurality of sets of event data. In this example, the values of the grid frequency f as a function of time in the first time window T (that is, values having a measurement time between time t3 and time t4) are determined as the grid response data. The changes in power provision and / or consumption represented by the sets of event data Ei to Ee cause changes in the measured values of the first parameter (e.g. grid frequency) in the first time window T. It will be appreciated that the sets of event data Ei to Ee need not, and in some examples do not, represent all of the power change events that occur in the grid 200 within the first time window T. Indeed, there may be other changes in power provision and / or consumption by other power units of the grid 200 that also influence the measured values of the first parameter in the first time window T. However, the influence of the power changes represented by the sets of event data Ei to Ee on the measured values of the first parameter in the first time window T can nonetheless be determined, for example as described in more detail below. Accordingly, a value of an electric power flow characteristic of the electric power grid associated with the first time window T can nonetheless be determined on the basis of the determined sets of event data Ei to Ee and the determined grid response data, for example as described in more detail below. As mentioned, the method of Figure 1 comprises, in step 106, determining, based on the determined sets of event data, and based on the determined grid response data, the value of the electric power flow characteristic of the electric power grid. This may be done in different ways. Some example ways in which this may be done are described in more detail below. To illustrate the methods, in the following examples the first parameter is taken as the grid frequency / and the electric power flow characteristic is taken as grid inertia H. However, it will be appreciated that in other examples other first parameters and other electric power flow characteristics may be used, with the appropriate modifications being made. In examples, determining the value of the electric power flow characteristic (e.g. inertia) may comprise fitting the determined sets of event data or data derived from the determined sets of event data to the determined grid response data or data derived from the determined grid response data. For example, this may comprise using the determined sets of event data in a model of a response of the first parameter (e.g. grid frequency) to the power changes, and fitting the modelled response to the determined grid response data. The model may be parameterised by the electric power flow characteristic (e.g. inertia), and the fitting may comprise optimising the value of the electric power flow characteristic in the model, thereby to determine the value of the electric power flow characteristic of the grid. For example, behaviour of the grid 200 in the first time window T may be modelled as follows. Let N denote the total number of power flow change events that occur in the time window T, with each individual event being denoted by the subscript n. The power change resulting from each event can be mathematically represented as a step function en(t): enW = Anu(t - Tn~) (1) where t is time, An is the amplitude of the power change of the nth event (which may be positive for increases in power provided to the grid or decreases in power consumed from the grid and negative for increases in power consumed from the grid or decreases in power provided to the grid), Tn is the time of the power change of the nth event, and u(t) is given by: , . f L t >0 “ 10, t <o An example of a step function en(t) is represented in Figure 3A, where in this example the amplitude An is A, and the time of the power change rn is ti. The total power functionp(t) resulting from the power change of all N of the events during the first time window T is: p(t) = Sn=l^nW(t-Tn) (3). Taking the grid frequency f as an example of the first parameter: the relationship between power change p(t) and grid frequency / in an electric power grid is given by the swing equation: P(O = —(4) / 0 ot where H is the grid inertia, and / ois the nominal grid frequency, which for the UK grid is 50 Hz, for example. Substituting equation (3) into equation (4), taking the time integral of both sides, and rearranging, yields the frequency response f(t) to the power function: / (0 = £ Sn=i Anu(t - Tn) dt (5) which in turn yields: / (0 = Sn=i Anm(t - rn) (6) where ft, t >0 - to, t <0 The inertia H may be determined by solving equation (6) for H. In examples, the inertia H may be determined by fitting data derived from the determined sets of event data to the determined grid response data. For example, the amplitude An and the time Tn of each determined set of event data may be substituted into a respective term on the right hand side of equation (6), and the left hand side of equation (6) may be provided by the determined grid response data, in this case the measured grid frequency as a function of time. The value of inertia H may then be varied so as to fit the right hand side of equation (6) to the left hand side of equation (6). For example, the value of inertia H may be varied so as to minimise a difference between the right hand side and the left hand side of equation (6). For example, a least squares fitting procedure may be used. For example, the value of inertia H that minimises the difference may be determined as the inertia of the grid 200. Other fitting procedures may be used. As mentioned, in examples, the determined sets of event data may not represent all of the power changes that occur in the first time window T, and there may well be other power changes occurring for which the time and amplitude is not known. This can be taken into account by expanding equation (6) to include a term representing power changes for which the event data is not known: / (0 = (En=i Anm(t - t„) + En=W1+1 Anm(t - tJ) (8) where Ni is the total number of power change events for which a determined set of event data is provided, and TV is the total number of power change events occurring in the first time period T. The second term in the right hand side of equation (8) is unknown and may be treated as a noise. The inertiaHmay be determined from equation (8) in a similar way as described above for equation (6). Again, the inertia H may be determined by fitting the determined sets of event data to the determined grid response data. For example, the amplitude An and the time in of each determined set of event data (that is, up to Ni) may be substituted into a respective term on the right hand side of equation (8), and the left hand side of equation (8) may be provided by the determined grid response data, in this case the measured grid frequency as a function of time. The inertia H may then be varied so as to fit the right hand side of equation (8) to the left hand side of equation (8). For example, the inertia H may be varied so as to minimise a difference between the right hand side and the left hand side of equation (8). For example, a least squares fitting procedure may be used. For example, the following optimisation problem may be formulated on the basis of a least squares fitting approach: ™n (Sn=l Anm(t - T„) + Sn^i+l Anm(t - Tn)) - f (t)) (9) and may be solved by varying H until the minimisation term is minimised. As mentioned, in some examples, determining the value of the electric power flow characteristic of the electric power grid (in this example inertia H) may involve fitting the data derived from the determined sets of event data to the grid response data. In some examples, this fitting may involve applying a matched filter to the grid response data with respect to the data derived from the determined sets of event data. A matched filter correlates a first signal with a second signal to detect the presence of the first signal in the second signal, and to determine a scaling of the first signal in the second signal. In the present example, the matched filter may correlate the modelled frequency response to the known power changes in the first time window (i.e. Sn=i Anm(t — t„) ) with the grid frequency measured in the first time window (i.e. fit)) to detect the f presence of the former in the latter, and to determine a scaling (i.e. —) of the former in 2H the latter. In this case, by determining the scaling, the value of inertia can be determined. Mathematically, in this example, by applying a matched filter, the grid inertia H in the first time window T is given by: fo 2 (10) where zC.t') = 'Ln1=1Anm(t ~Tn) (11) In some examples, the value of the electric power flow characteristic may be determined based on an average of the changes in power indicated by the determined sets of event data and an average of portions of the grid response data. For example, determining the value of the electric power flow characteristic may comprise averaging the changes in power indicated by the determined sets of event data; averaging a plurality of portions of the response data; and determining the value of the electric power flow characteristic based on the averaged changes and the averaged portions of response data. In this example, each portion of the response data corresponds to values of the first parameter (e.g. frequency f) within a third time window beginning at the time indicated by a respective one of the determined sets of event data. For example, as illustrated in Figure 4, the corresponding portion of the grid response data for the fourth set of event data E4 corresponds to values of the first parameter (grid frequency f in the example of Figure 4) within a third time window T’ beginning at the time indicated by the fourth set of event data E4. Similarly, for each of the other sets of event data Ei, E2, E3, Es, Ee, the corresponding portion of the grid response data corresponds to values of the first parameter (grid frequency f in the example of Figure 4) within a respective third time window (not shown in Figure 4) beginning at the time indicated by the set of event data. Each third time window has a certain duration Tr. In examples, the duration Tr of each third time window may be the same. In this example, an average of the amplitudes of each of the sets of event data Ei to Ee may be determined; an average of the respective portions of the grid response data may be determined; and a value of the electric power flow characteristic (in this case inertia H) may be determined based on the average of the amplitudes and the average of the portions of the grid response data. In the present example, this may be done using the swing equation (see equation (4)). For example, the average amplitude may be substituted for p(t) in equation (4), the rate of change of frequency of the 5 f averaged portions of the grid response data may be determined and substituted for — in equation (4). The value of inertia H may then be determined from equation (4) accordingly. It will be appreciated that in cases where sets of event data include both positive and negative amplitudes, the absolute value of the each amplitude may be used for the average of the amplitudes, and the portions of grid response data that correspond to negative amplitudes may be accordingly sign compensated before the portions of grid response data are averaged. In examples, an approach to determining the value of the electric power flow characteristic may be mathematically represented as follows. In this example, the first parameter is grid frequency f and the electric power flow characteristic is grid inertia H, however it will be understood that in other examples other first parameters and electric power flow characteristics may be used, with appropriate modifications being made. The average absolute power change P(t) of the power changes indicated in the determined sets of event data in the first time window T may be given by: PiQ = ^Nn^\An\u^ (12). The function P(t) of equation (12) is illustrated in Figure 5A. At times less than zero, P(t) is zero, and at times greater than or equal to zero, P(t) is Sn=i lAi I • In Figure 5 A, Tr is the duration of the third time window. The average sign compensated frequency response F(t) to the power change is given by: F(t) = ^-Sn=l5^(^n) / (Tn + 0 (13). In equation (13), the factor f(rn + t) indicates the measured frequency beginning at the time t„ of the nth power change event. This acts to aggregate all of the portions of grid response data to the same start time of t=0. The function F(t) of equation (13) is illustrated in Figure 5B. The function F(t) begins as time t=0 and increases, representing that an increase in power provided to the grid increases the grid frequency. In Figure 5B, Tr is the duration of the third time window. The grid inertia H may be determined, for example, by applying a matched filter to F(t) with respect to P(t), in a similar way to as described above. For example, the inertia H may be given by: fo JorP2mdt 2 (14) where Tr is the duration of the third time window, and F'(t) is the rate of change of the average sign compensated frequency response F(t) and is given by: = + t) (15). IV i In examples, for each of the determined sets of event data, the indicated power change may have a duration that is at least as long as the duration Tr of the third time window. In this case, the average absolute power change P(t) of equation (12) may be given instead by: = (16) IV1 and this expression of P(t) may be used instead in equation (14). This may simplify determination of inertia H in equation (14). In examples, the grid 200 may comprise one or more compensating devices (not shown) configured to provide power to or consume power from the electric power grid in response to a change in a value of the first parameter in order to compensate the change in the value of the first parameter. For example, the compensating devices may be configured to, in response to a drop in the grid frequency, provide power to the grid in order to restore the grid frequency to its nominal value. In this case, the response of the first parameter to any particular power change event may be impacted by the influence of the compensating devices, which has the potential to negatively impact the accuracy of the electric power flow characteristic. However, the one or more compensating devices may have a response time or typical response time in which the one or more compensating devices provide power to or consume power from the electric power grid in response to the change in the value of the first parameter, and before which no response is provided. For example, in the UK grid, the typical response time for compensating devices is around 5 to 10 second seconds. In examples, the third time window may be set to have a duration Tr less than the response time in which the one or more compensating devices provide power to or consume power from the electric power grid in response to the change in the value of the first parameter. For example, for the UK grid, the duration of the third time window may be set to around 5 seconds or less. This may help ensure that the frequency response of the grid to any particular power change event is not impacted by the influence of the compensating devices, which may, in turn, improve the accuracy with which the electric power flow characteristic is determined. As mentioned, in examples, the determined sets of event data may not represent all of the power changes occurring in the grid 200 in the first time window T. In some cases, some power flow change events occurring in the grid may be correlated with one another. For example, two or more events that are correlated with one another may occur at the same time, for example at precisely the same time, or practically so. For example, this may occur at certain times of day. For example, many appliances may be programmed to switch on at a certain time of day. As another example, this may occur in response to certain other events. For example, where an energy provider indicates the start of a period of lower electricity pricing, many appliances may consume power from the grid in response. As mentioned, the first plurality of sets of event data (on the basis of which the electric power flow characteristic is determined) indicate power changes by the first plurality of power units 219. However, if one or more of these indicated power changes results from an event that is correlated with another event from a power unit that is not part of the first plurality of power units, this has the potential to negatively impact the accuracy of the electric power flow characteristic determination. For example, it may be that one hundred power change events, by a respective one hundred different power units 219, are correlated with one another, and occur at the same time. Each event results in a power change by the respective power unit 219. However, if only fifty of these power units 219 are in the first plurality of power units, then only fifty of the power changes resulting from only fifty of the correlated events will be indicated in the first plurality of sets of event data. In this case, the power change resulting from all one hundred of the correlated events occurring in the grid at that exact time may be underrepresented by the first plurality of sets of event data. Accordingly, in examples, the method may comprise, for each of one or more of the first plurality of sets of event data: categorising the set of event data as belonging to a first category. Specifically, a set of event data may be categorised as belonging to the first category responsive to determining that the change indicated by the set of event data results from an event that is correlated with one or more further events, the one or more further events resulting in a power change by a respective one or more power units that are not of the first plurality of power units. The determination of the value of the electric power flow characteristic of the electric power grid may then be based additionally on the categorisation of the one or more sets of event data as belonging to the first category. This allows for correlation of events to be taken into account in the determination of the electric power flow characteristic, which may, in turn, help improve the accuracy of the determined electric power flow characteristic. Hereinafter, sets of event data that are categorised as belonging to the first category may also be referred to as first category sets of event data. In some examples, the method may comprise excluding the sets of event data categorised as belonging to the first category from being used in the determination of the value of the electric power flow characteristic of the electric power grid. For example, the first category sets of event data may be removed from the determined first plurality of sets of event data, and the determination of the value of the electric power flow characteristic may be performed, for example as described above, on the basis of the resulting sets of event data. This may help ensure that sets of event data indicative of power changes resulting from correlated events that are not fully represented in the sets of event data are not included in the determination of the electric power flow characteristic, which may, in turn, help improve the accuracy of the electric power flow characteristic determination. In some examples, the sets of first category event data may not be excluded from the determination of the value of the electric power flow characteristic, but instead the power changes indicated thereby may be parameterised in the determination of the value of the electric power flow characteristic. For example, the method may comprise parametrising the change indicated by the one or more sets of event data categorised as belonging to the first category with a second parameter k. Specifically, the second parameter k may represent the proportion, of the total change in consumption or provision of power resulting from the correlated power change events, that the change indicated by the one or more sets of event data categorised as belonging to the first category represents.. Determining the value of the electric power flow characteristic may then be based on the parameterised change in consumption or provision of electric power indicated by the categorised one or more sets of event data. For example, in examples where the first parameter is frequency f and the electric power flow characteristic is inertia H, equation (8) may be re-written to include the power changes resulting from correlated events as a separate term parameterised by k: = (^n=l ~ Tn) + AnTTl(t - Tn) + Sn=JV1+l “ ^n)) (17). In equation 17, the power change events n=l to No are those resulting in power changes included in the determined sets of event data which have not been categorised as belonging to the first category , the power change events n=No+l to Ni are those resulting in power changes included in the determined sets of event data which have been categorised as belonging to the first category, and the power change events n=Ni+l to N are those events occurring in the first time period T but which result in power changes that are not included in the determined sets of event data. As above, the latter of these three terms is unknown and may be treated as noise. The parameter k represents the proportion, of the total power change resulting from the correlated events, that the power change resulting from events n=No+l to Ni represent. For example, the parameter k may be greater than zero and less than or equal to 1. The power change events n=No+l to Ni may all occur at the same time, and hence rn may be the same for each of these events. In examples, the parameter k may be known, or an estimate of the parameter k may be known. For example, it may be known or estimated that the power change resulting from events n=No+l to Ni represent 10% of the total correlated power change occurring at that time. In this case, the parameter k may be set to 0.1 In cases where the parameter k is known, or an estimate of the parameter k is known, this may be included into equation (17) and equation (17) may be solved for inertia H, for example in a similar way as described above for equation (8). Including the known value or estimate of the value of k into the determination of the electric power flow characteristic may improve the accuracy of the determination of the value of the electric power flow characteristic. In examples, the parameter k may not be known. In this case, the method may comprise determining a value of the second parameter k concurrently with the determination of the value of the electric power flow characteristic (e.g. inertia / / ) For example, equation (17) may be solved for both H and k. For example, the H and k may be determined by fitting the determined sets of event data to the determined grid response data. For example, the amplitude An and the time rn of each determined set of event data (that is, corresponding to events up to Ni) may be substituted into the appropriate term on the right hand side of equation (17), and the left hand side of equation (17) may be provided by the determined grid response data, in this case the measured grid frequency as a function of time. The inertia / / and the second parameter k may then be varied so as to fit the right hand side of equation (17) to the left hand side of equation (17). For example, the inertia H and the second parameter k may be varied so as to minimise a difference between the right hand side and the left hand side of equation (17). For example, a least squares fitting procedure may be used. For example, the following optimisation problem may be formulated on the basis of a least squares fitting approach: , . 2 + iX^Wo+iAi"i(t-Li)) - / (o) / / \Z.ri x. K u / I (18) and may be solved by varying H and k until the minimisation term is minimised. This may allow for both a value of the grid inertia / / , and the proportion k, of the total power change resulting from the correlated events, that the power change resulting from events n=No+l to Ni represents, to be determined. In examples, while the parameter k may not itself be known, there may be known one or more constraints on the value that parameter k can take. Accordingly, in some examples, the variation of the parameter k in the minimisation of the minimisation term of equation (18) may constrained according to the one or more known constraints. In examples, the one or more sets of event data that belong to the first category may be known. For example, it may be known that a certain subset of the determined sets of event data indicate a power change resulting from an event that is correlated with a further event that is by a power unit that is not part of the first plurality of power units (and hence which causes a power change that is not indicated in the first plurality of sets of event data). This known subset may be categorised as belonging to the first category accordingly, and this categorisation taken into account in the determination of the electric power flow characteristic, as described above. In examples, the one or more sets of event data that are categorised may not be known in advance. In these examples, or otherwise, categorising the one or more of the determined sets of event data as belonging to the first category may comprise analysing the first plurality of sets of event data to identify the one or more sets of event data. This analysis may be performed in different ways. In examples, this analysis may comprise an analysing the power change and / or the time of the change indicated by each of the first plurality of sets of event data. As one example, it may be known that correlated events are likely to occur at certain times of day, for example exactly at 14:00 and / or exactly 18:00. In this case, the analysis may comprise analysing the time indicated by each of the determined sets of event data to identify those sets of event data that indicate the certain times of day, for example exactly 14:00 and exactly 18:00. Those identified sets of event data may be accordingly categorised as belonging to the first category, and taken into account in the determination of the electric power flow characteristic, as described above. As another example, the analysis may comprise analysing the determined sets of event data to determine whether two or more of the power changes (for example, a threshold number or more of the power changes) indicated thereby occur at the same as one another. If it is determined that the two or more of the determined sets of event data indicate power changes that occur at the same time as one another, then it may be determined that the power change events resulting in these power changes are correlated with one another. From this, it may be inferred that there may be further such correlated events but by power units not included in the first plurality of power units (and hence which result in a power changes that are not indicated in the first plurality of sets of event data). These two or more sets of event data may therefore be categorised as belonging to the first category accordingly. Referring to Figure 6, there is illustrated an example of analysing determined sets of event data A20 to A26 within a first time window T to identify sets of event data A24 to A26 that indicate power changes resulting from events that are correlated with one another (and thereby to identify sets of event data A24 to A26 as belonging to the first category). In this example, the analysis may comprise, for each of a plurality of second time windows Ti to T4 within the first time window T: determining a sum LP of the power changes indicated by respective sets of event data that indicate a time within the second window; and in response to a magnitude of the sum XP being larger than a first threshold value Pi, categorising the sets of event data A24 to A26 that indicate a time within the second time window T4 as belonging to the first category. For example, as illustrated in Figure 6, the magnitude of the sum of the power changes in each of the second time windows Ti, T2 and T3 is less than the first threshold value Pi. Accordingly, the sets of event data A20 to A23 indicating a time within these second time windows Ti, T2 and T3 are not categorised as belonging to the first category. However, the magnitude of the sum of the power changes in the second time window T4is larger than the first threshold value Pi. Accordingly, the sets of event data A24 to A26 indicating a time within this second time window T4 are identified as indicating power changes resulting from correlated events, and hence are categorised as belonging to the first category. This may provide a computationally efficient way of identifying sets of event data belonging to the first category. In examples, as illustrated in Figure 6, the second time windows Ti to T4 may be overlapping, for example 50% overlapping. For example, the second half of Ti may overlap with the first half of T2, the second half of T2 may overlap with the first half of T3, and the second half of T3 may overlap with the first half of T4. The overlapping may help capture power changes resulting from correlated events more reliably, as this may reduce the impact of an outlier that may be present in any particular second time window and hence which may mask the presence of power changes resulting from correlated events. It will be appreciated that in other examples, other ways of analysing the first plurality of sets of event data to identify the one or more sets of event data belonging to the first category may be used. For example, a machine learning model trained to, based on an input of two or more sets of event data, produce an output indicative of whether the two or more sets of event data indicate power changes resulting from respective events that are correlated with one another, may be used. Other methods may be used. In examples, the determined first plurality of sets of event data may represent all of the sets of event data within the first time window T that are available to or obtainable by the computing system 205. However, in some examples, the determined first plurality of sets of event data may be selected from a larger, second plurality of sets of event data that are available to or obtainable by the computing system. That is, the second plurality of sets of event data may be filtered to determine the first plurality of sets of event data. In examples, the method may comprise obtaining a second plurality of sets of event data, for example, from a database. For example, each of the second plurality of sets of event data may indicate a time within the first time window T. The method may then comprise selecting the first plurality of sets of event data from the second plurality of sets of event data, thereby to determine the first plurality of sets of event data. In examples, determining the first plurality of sets of event data may comprise selecting the first plurality of sets of event data from a second plurality of sets of event data based on one or more attributes associated with each of one or more of the second plurality of sets of event data. As described in more detail below, this may allow for improved accuracy of the electric power flow characteristic determination and / or for the electric power flow characteristic determination to be tailored to one or more particular attributes. In examples, the one or more attributes comprise a location of the power unit 219 associated with the set of event data. In this case, determining the first plurality of sets of event data may comprises selecting the first plurality of sets of event data from the second plurality of sets of event data based at least in part on the respective locations associated therewith. This may allow for the value of the electric power flow characteristic to be determined for specific regions. In examples, the location may be a geographical location of the power unit 219 and / or a location of the power unit 219 within the grid 200. In examples, the region may be a geographical region and / or a region of the grid 200. Referring to Figure 7, there is illustrated an example electric power grid 700. The grid 700 comprises power units 219A-219D in different locations, as well as measurement devices 220A-220C in different locations. Specifically, a first power unit 219A and a second power unit 219B are located in a first region 702, and a first measurement device 220A is configured to measure value of the first parameter in the first region 702. A third power unit 219C and a fourth power unit 219D are located in a second region 704 and a second measurement device 220C is configured to measure value of the first parameter in the second region 704. A third measurement device 220B is configured to measure a value of the first parameter in a location in between the first region 702 and the second region 704. In examples, the first plurality of sets of event data may be selected based at least in part on the respective locations associated therewith each belonging to a common geographical region or region of the electric power grid 200. This may allow for a value of the electric power flow characteristic to be determined that is specific to that common geographical region or region of the grid 200. For example, sets of event data from the first power unit 219A and the second power unit 219B may be selected on the basis that their locations belong to a common region, specifically the first region 702. The determined value of the electric power flow characteristic may accordingly be specific to the first region 702 of the grid. As another example, sets of event data from the third power unit 219C and the fourth power unit 219D may be selected on the basis that their locations belong to a common region, specifically the second region 704. The determined value of the electric power flow characteristic may accordingly be specific to the second region 704 of the grid. In examples, the first plurality of sets of event data may be selected based on the respective locations associated therewith being distributed over multiple regions of the grid, for example over the entire grid. This may allow for a value of the electric power flow characteristic to be determined that is representative of multiple regions of the grid 200, for example representative of the entire grid. For example, sets of event data from the second power unit 219B in the first region 702 and the fourth power unit 219D in the second region 704 may be selected on the basis that their locations are distributed over the grid 700. The determined value of the electric power flow characteristic may accordingly be representative of the entire grid, as opposed to any specific one of the regions 702, 704. In examples, each set of event data may be indicative of the location of the associated power unit 219. For example, the data generating device 218 may be configured to include in the sets of event data transmitted to the computing system 205 an indication of the location of the power unit 219 with which it is associated. This may allow for the location of the power unit 219 associated with the set of event data to be efficiently determined. In examples, there may be a plurality of sets of grid response data. In examples, determining the grid response data may comprise selecting the grid response data from the plurality of sets of grid response data based at least in part on a location at which the values of the first parameter indicated thereby were measured. This may allow for the value of the electric power flow characteristic to be determined for specific regions and / or allow improved granularity in the specific regions for which the value of the electric power flow characteristic is determined. For example, as illustrated in Figure 7, there may be a set of grid response data for each of the first measurement device 220A, the second measurement device 220C, and the third measurement device 220B. As one example, the set of grid response data from the first measurement device 220A in the first region 702 may be selected. Further, sets of event data associated with the first power unit 219A and the second power unit 219B in the first region 702 may be selected. Accordingly, the determined value of the electric power flow characteristic may be representative specifically of the first region 702. As another example, the set of grid response data from the third measurement device 220B may be selected. Further, sets of event data associated with the first to fourth power units 219A to 219D may be selected. Accordingly, the determined value of the electric power flow characteristic may be representative of the whole grid 700. In examples, each set of grid response data may indicate a location at which the first parameter indicated thereby was measured. For example, each measurement device 220 may be configured to include in the set of grid response data transmitted to the computing system 205 an indication of the location at which the first parameter was measured. This may allow the location associated with the set of grid response data to be efficiently determined. As mentioned, determining the first plurality of sets of event data may comprise selecting the first plurality of sets of event data from a second plurality of sets of event data based on one or more attributes associated with each of one or more of the second plurality of sets of event data. In examples, the one or more attributes may comprise one or more of a magnitude, rate, and duration of the change in power provided to or consumed from the electric power grid indicated by the set of event data. In these examples, determining the first plurality of sets of event data may comprise selecting the first plurality of sets of event data from the second plurality of sets of event data based at least in part on one or more of the respective magnitudes, rates, and durations of the respective changes in power provided to or consumed from the electric power grid indicated thereby. For example, this may allow determination of the value of the electric power flow characteristic to be specific to certain magnitudes, rates, and or durations of the respective power changes. In examples, each set of event data may include values of the magnitude, rate, and / or duration of the power change indicated thereby. In examples, each power change comprises a change of the power consumed and / or provided by the associated power unit 219 from a first value to a second value. In examples, the magnitude may be taken as the absolute value of the change in power, for example the absolute value of the difference between the first value and the second value. In examples, the rate may be taken as the first time derivative of power at the time of the power change. In examples, the rate may be taken as the inverse of the time taken for the power to change from the first value to the second value. In examples, the duration may be taken as the length of time that the power remains at the second value. In some examples, each set of event data may comprise values of the power consumed and / or provided by the associated power unit 219 as a function of time, which represent a change in power consumed and / or provided by the associated power unit 219. In these examples, the sets of event data may be analysed, for example by the computing system 205, to determine the magnitude, rate, and / or duration of the change in power provided to or consumed from the electric power grid by the associated power unit 219. In examples, the first plurality of sets of event data may be selected based at least in part on one or more of the respective magnitudes, rates, and durations of the respective changes in power provided to or consumed from the electric power grid indicated thereby being greater than a threshold magnitude, threshold rate, and threshold duration, respectively. This may help provide for an accurate determination of the value of the electric power flow characteristic. For example, if the magnitude, rate and / or duration of the power change is too small (for example below the threshold magnitude, threshold rate, or threshold duration, respectively) the change this causes in the first parameter may be too small, slow, and / or short, respectively, to be detectable amongst the noise in the first parameter. Accordingly, selecting sets of event data whose power change magnitudes, rates, and / or durations are larger than the respective thresholds may therefore help ensure that power changes whose influence of the first parameter is detectable are included in the determination of the value of the electric power flow characteristic, which may, in turn, help improve the accuracy of the determined value. As another example, selecting the sets of event data based on the magnitudes, rates, and / or durations being larger than the respective thresholds may help ensure that the magnitude, rate and / or duration of the associated power changes are large enough to fit appropriately with assumptions that may be made in the determination of the value of the electric power flow characteristic. For example, in some of the models described above it is assumed that the power change is a step function. If the rate of the power change is too small (for example below the threshold rate), this assumption may no longer be valid, which may negatively impact the accuracy of the determination of the value of the electric power flow characteristic. Accordingly, selecting sets of event data whose power change rates are larger than the threshold may therefore help ensure that the assumption is valid, which may, in turn, help improve the accuracy of the determined value. As another example, in some of the models described above, it is assumed that the unknown power change events occurring in the first time window T can be treated as noise. If the magnitude and / or duration of the power change is too small, the influence of these events on the first parameter may not be detectable in the noise, and hence this assumption may not be valid. Accordingly, selecting sets of event data whose power change magnitudes and / or durations are larger than the respective thresholds may therefore help ensure that the assumption is valid, which may, in turn, help improve the accuracy of the determined value of the electric power flow characteristic. In some of the above examples, the method performed at the computing device 205 involves selecting the first plurality of sets of event data from the second plurality of sets of event data based on one or more of the respective magnitudes, rates, and durations of the respective power changes indicated thereby being greater than a threshold magnitude, threshold rate, and threshold duration, respectively. However, in some examples, one or more of the data generating devices 218 may be configured to determine whether a power change of the associated power unit 219 has a magnitude, rate, and / or duration greater than a threshold magnitude, threshold rate, and / or threshold duration, respectively, and, if so, generate and send a set of event data indicating the power change and the time of the change to the computing system 205. On the other hand, if the power change of the associated power unit 219 has a magnitude, rate, and / or duration less than a threshold magnitude, threshold rate, and / or threshold duration, respectively, the data generating device 219 may not generate the set of event data and / or may not send event data indicating the power change to the computing system 205. In this case, only sets of event data that indicate a power change having a magnitude, rate, and / or duration greater than a threshold magnitude, threshold rate, and / or threshold duration, respectively, may be obtained by the computing system 205. Accordingly, the first plurality of sets of event data determined by the computing system may automatically each indicate a power change that has a magnitude, rate, and / or duration that is appropriate for accurately determining the value of the electric power flow characteristic. Referring to Figure 8, there is illustrated a method of providing one or more sets of event data for determining a value of an electric power flow characteristic of an electric power grid 200. The method may be performed by a data generating device 218 associated with a respective power unit 219 (for example at each of one or more generating devices 218, each being associated with a respective power unit). The method comprises, in step 802, determining a change in consumption of electric power from or provision of electric power to the electric power grid 200 by the associated power unit 219. For example, this may be performed in any of the example ways described above. In examples, determining the change in consumption of electric power from or provision of electric power to the electric power grid 200 by the associated power unit 219 may comprises measuring and / or controlling the change in the consumption of electric power from or provision of electric power to the electric power grid by the associated power unit 219. For example, the data generating device 218 may comprise a power measurement device configured to measure or meter electric power consumed from the grid 200 by the associated power unit and / or provided by the associated power unit 219 to the grid 200. The power measurement device may measure the power consumed and / or provided as a function of time, and based on this, the data generating device 218 may determine the power change. As another example, the data generating device 218 may comprise a control device configured to control a change in power consumption and / or provision of an associated power unit 219. For example, the control device may send a control signal to an associated power device 219 to change the power that the power unit 219 consumes from or provides to the grid 200 by a certain amount at a certain time. In these examples, the data generating device 218 may determine the power change based on the control signal. The method comprises, in step 804, determining that one or more of a magnitude, rate, and duration of the change is greater than a threshold magnitude, threshold rate, and threshold duration, respectively. For example, the data generating device 218 may be configured to analyse the power change to determine one or more of a magnitude, rate, and duration of the change. The data generating device 218 may store one or more of the thresholds, for example in a memory. The data generating device 218 may retrieve one or more of the thresholds from the memory, and compare the determined determine one or more of a magnitude, rate, and duration of the change to the appropriate threshold. In examples, the thresholds may bet set during manufacture of the data generating devices 208. In examples, one or more of the thresholds may be changed, for example by manual programming of the data generating device 208, or by remote update, for example via communication over the Internet. The method comprises, in step 806, in response to determining that one or more of a magnitude, rate, and duration of the change is greater than a threshold magnitude, threshold rate, and threshold duration, respectively: generating a set of event data indicating i) the change in consumption of electric power from or provision of electric power to the electric power grid by the associated power unit and ii) a time of the change. In examples, the generated set of event data may be the same as the event data according to any one of the examples described above with reference to Figures 1 to 7. In examples, the data generating device 218 may comprise a clock function. The data indicative of a time of the change may comprise a timestamp, generating by the clock function at the time the power change was determined to have occurred. In examples, the generated set of event data may indicate i) an amplitude of the change in consumption of electric power from or provision of electric power to the electric power grid by the associated power unit and ii) a time of the change. In examples, the generated set of event data may comprise a value indicating the amplitude of the change and a value indicating the time of the change. In examples, the generated set of event data may comprise data representing the power as a function of time. The method comprises, in step 808, sending the set of event data to a computing system over a network. For example, the data generating device 218 may send the set of event data as a data packet to the computing system 205 over a computer network such as the Internet. This may allow that the computing system 205 only receives sets of event data that are appropriate for determining an accurate value of the electric power flow characteristic. This may facilitate the accurate determination of the value of the electric power flow characteristic. Alternatively, or additionally, this may reduce the filtering performed by the computing system 205 of event data in order to select such sets of event data, which in turn may reduce processing performed by the computing system 205. Further, this may reduce the volume of sets of event data that are sent over the computer network, and hence may reduce network traffic. In examples, as described above, the set of event data may indicate a location of the associated power unit. As described above, this may allow the computing system 205 to tailor the determination of the value of the electric power flow characteristic to specific regions of the grid 200. In examples, the determined value of the electric power flow characteristic is representative of the electric power flow characteristic at or during the first time window T. The value of the electric power flow characteristic may change over time. The value of the electric power flow characteristic representative of a particular time can be determined by defining the first time window to include, for example be centred on, the particular time. Accordingly, the value of the electric power flow characteristics for different times can be determined. This may allow for tracking of changes in the value of the electric power flow characteristic over time, which may, in turn, help identify changes occurring in the operation of the grid 200, for example. Referring to Figure 9, there is illustrated a system 900 according to an example. The system 900 comprises data generating device 218A, 218B. Each of the data generating devices 218A, 218B may be the same as or similar to the data generating device 218 according to any one of the examples described above with reference to Figures 1 to 8. For example, one or more of the data generating devices 218A, 218B may be configured to perform the method according to any one of the examples described above with reference to Figure 8. The system 900 comprises measurement devices 220A, 220B. Each of the measurement devices 220A, 220B may be the same as or similar to the measurement device 210 according to any one of the examples described above with reference to Figures 1 to 8. The system 900 comprises a computing system 205. The computing system 205 may be the same as or similar to the computing system 205 according to any one of the examples described above with reference to Figures 1 to 8. For example, the computing system 205 may be configured to perform the method according to any of the examples described above with reference to Figures 1 to 7. The data generating devices 218A, 218B are in communication with the computing system 205 over a computing network 901, such as the Internet. Specifically, the data generating devices 218A, 218B are configured to send sets of event data to the computing system over the network 901, for example as per any one of the examples described above with reference to Figures 1 to 8. The measurement devices 220A, 220B are in communication with the computing system 205 over the computing network 901. Specifically, the measurement devices 220A, 220B are configured to send grid response data to the computing system over the network 901, for example as per any one of the examples described above with reference to Figures Ito 8. The computing system 205 comprises a first database 902 and a second database 904. One or both of the first database 902 and the second database 904 may be a time series database. The computing system 205 may store the received sets of event data in the first database 902. The computing system 205 may store the received grid response data in the second database 904. The computing system 205 comprises a determination unit 906. The determination unit 906 may be configured to retrieve sets of event data from the first database 902 and retrieve grid response data from the second database 904. The determination 906 unit may be configured to determine the value of the electric power flow characteristic based on the determined first plurality of sets of event data, and based on the determined grid response data, as per any one of the examples described above with reference to Figures 1 to 8. The determination unit 906 may output the determined value of the electric power flow characteristic 908. Referring to Figure 10, there is illustrated an apparatus 1000 according to an example. The apparatus 1000 comprises a processor 1002, a memory 1004, an input interface 1006 and an output interface 1008. In examples, one or more of the data generating devices 218 of any of the examples described above with reference to Figures 1 to 8 may be provided by the apparatus 1000. That is, in examples, the apparatus 1000 may be configured to perform the functions of a data generating device 218 according to any one of the examples described above with reference to Figures 1 to 8. In examples, the apparatus 1000 may be configured to perform the method of any of the examples described above with reference to Figure 8. The memory 1004 may store a computer program which, when executed by the processor 1002 causes the processor to perform the method according to any of the examples described above with reference to Figure 8. In examples, the input interface 1006 may receive power measurements from a power measurement device (not shown in Figure 10), which power measurements may be indicative of a power change by the associated power unit 219. In examples, the input interface 1006 may receive control signals from a control device (not shown in Figure 10), which control signals may be indicative of a power change by the associated power unit 219. In examples, the output interface 1008 may be connected to a computer network such as the internet. In examples, the processor 1002 may send, via the output interface 1008, one or more sets of event data, as per any of the examples described above, to the computing system. In other examples, the computing system 205 according to any of the examples described above with reference to Figures 1 to 8 may be provided by the apparatus 1000. That is, in these other examples, the apparatus 1000 may be configured to perform the functions of the computing system 205 according to any one of the examples described above with reference to Figures 1 to 8. In examples, the apparatus 1000 may be configured to perform the method of any of the examples described above with reference to Figures 1 to 7. In these examples, the memory 1004 may store a computer program which, when executed by the processor 1002 causes the processor to perform the method according to any of the examples described above with reference to Figures 1 to 7. In examples, the memory 1004 may store the sets of event data and / or the grid response data, according to any of the examples described above. In examples, the input interface 1006 may be connected to a computer network such as the internet. In examples, the input interface 1006 may receive sets of event data from data generating devices 218. In examples the input interface 1006 may receive sets of grid response data from measurement devices 220. In examples, the processor 1002 may output the determined value of the electric power flow characteristic via the output interface 1008. In examples, the output interface 1008 may be connected to a computer display such as a computer monitor (not shown in Figure 10). In examples, the processor 1008 may be configured, via the output interface 1008, display the determined value of the electric power flow characteristic on the computer display. In examples, the output interface 1008 may be connected to a further storage (not shown in Figure 10) and the processor 1002 may output the determined value of the electric power flow characteristic to the further storage via the output interface 1008. Referring to Figures 11 to 14, there are illustrated results of a simulation that was conducted to simulate the determination of a value of an electric power flow characteristic according to examples described herein. In this simulation, the first parameter is frequency, and the electric power flow characteristic is inertia. However, it will be appreciated that in other examples, other first parameters and other electric power flow characteristics may be used. In the simulation, power changes were generated and their impact on grid frequency f was modelled using the swing equation (see equation (4) above). In the simulation, the value of the inertia H was set at 100 GWs. Figure 11 illustrates the impact on the deviation Af of grid frequency f from a nominal value (in this example 50Hz) of two power changes, one at 1 second with amplitude of +lkW, and a another at 30 seconds with an amplitude of -5kW. In the simulation, the impact of grid governors was also modelled. Specifically, after a delay Trof 5 seconds, the frequency deviation due to a power change is ramped down linearly to zero over 20 seconds. Accordingly, as can be seen in Figure 11, the power change event at 1 second causes the grid frequency to increase for 5 seconds before the grid frequency deviation is returned, over the next 20 seconds, to zero. Similarly, the power change event at 30 seconds causes the grid frequency to decrease for 5 seconds before the grid frequency deviation is returned, over the next 20 seconds, to zero. Referring to Figure 12, in the simulation, 1 million power changes were generated over a 30 minute first time window. The amplitude of each power change was assigned according to a Weibull distribution. The sign of the amplitude of each power change was randomly assigned. Figure 12 illustrates the probability distribution function of the generated power changes with respect to amplitude and sign. As can be seen in Figure 12, the magnitude of the vast majority of the power changes are between 0 and 4 kW. It is to be noted that in the context of a grid having an inertia of 100 GWs, these are relatively small power changes. It is also to be noted that this range of power changes is typical of the power rating of, for example, domestic appliances, of which there may be many millions connected to a typical national electric power grid such as in the UK. Referring to Figure 13, there is illustrated the frequency signal resulting from the 1 million power changes of Figure 12 over the 30 minute first time period. In the simulation, 10000 of the 1 million power changes were chosen at random. Accordingly, 10000 simulated sets of event data were generated indicating the amplitude and the time of the respective 10000 power changes. The frequency signal of Figure 12 was used as simulated grid response data. A value of grid inertia was determined using the 10000 simulated sets of event data, the simulated grid response data and techniques disclosed herein. Specifically, in this simulation, the approach described above with reference to equations (12) to (16) was used to determine the value of the inertia. This process was repeated 1000 times, each time with a different random set of 10000 power changes. Accordingly, 1000 values of the inertia were determined. Figure 14 illustrates a plot of these 1000 values. The median of these 1000 inertia values was 102.5 GWs. This is relatively close to the modelled value of 100 GWs. This demonstrates that methods disclosed herein may provide a determination of the value of an electric power flow characteristic of the a grid based on sets of event data indicating (relatively small) power changes that may be occurring as part of the typical, day-to-day use of the grid. Further, this demonstrates that this may be the case even where the sets of event data represent only a relatively small proposition of the total power changes occurring in the grid (in the above example, the 10000 sets of event data on which any one of the inertia values is determined represents 1% of the total number of power changes occurring in the grid in the 30 minute first time window). In some of the above examples, the first parameter is grid frequency, and the electric power flow characteristic is inertia H. However, this need not necessarily be the case and other first parameters and / or electric power flow characteristics may be used. The first parameter may be any parameter of electric power flow in an electric power grid whose value is caused to change by a respective change in consumption of electric power from or provision of electric power to the electric power grid by a power unit of the grid. The electric power flow characteristic may be any electric power flow characteristic of an electric power grid which can be determined based on sets of event data and grid response data as described herein. As one specific further example, the first parameter of electric power flow in the grid may be a difference between voltage phase angle at a first grid location and a second grid location (hereinafter ‘voltage phase angle difference’), and the electric power flow characteristic of the electric power grid may be indicative of a current topology or operational state of the electric power grid (that is to say, a topology or operational state of the electrical power grid that is currently or presently occurring). For example, the magnitude of the voltage phase angle difference resulting from a power change event is dependent on the difference in grid impedance between the first and second grid locations relative to the location of the power change event. This, in turn, is indicative of a current topology of the grid (e.g. whether or not a transmission line between the first and second grid locations is operational). Accordingly, a value indicative of a current topology or operational state of the grid can be determined based on the determined sets of event data as described above and grid response data indicating the voltage phase angle difference as a function of time. For example, similar techniques to those described above for frequency and inertia may be used for voltage phase angle difference and a value indicative of the current topology or operational state of the grid. For example, the first plurality of sets of event data may represent power changes by power units in a certain region of the grid in the first time window. The grid response data may represent the voltage phase angle difference as a function of time in the first time window. Using example methods disclosed herein, a value of the voltage phase angle difference for a particular (e.g. unit) power change may be determined. For example, the average absolute power change P(t) of the power changes indicated in the determined sets of event data in the first time window T may be given by equation (12) as above. The average voltage phase angle difference response Wijtt) to the power change may be given by: = ^Zn=lS9n(An)(pi,j^n + 0 (19) where, similar to equation (13) above, the factor <Pij(tn + t) indicates the measured voltage phase angle difference (pij beginning at the time Tn of the nth power change event up to Ni; where i denotes the first grid location, and j denotes the second grid location. As above, the sign compensation is to take into account that the absolute value \An | of the power change is used in equation (12), and hence voltage phase angle differences (ptj that correspond to negative power changes should be accordingly sign compensated before averaging. The value indicative of the voltage phase angle difference for a particular (e.g. unit) power change may be determined by dividing the magnitude of by the magnitude of P(t). Accordingly, this will provide a value of an electric power flow characteristic that is indicative of a current topology or operational state of the grid (e.g. whether or not a transmission line between the first and second grid locations is operational). For example, comparing the value of this characteristic calculated for different first time windows may provide for a dynamic determination of changes in the topology (and hence operational state) of the grid. It will be appreciated that the techniques disclosed herein may be applied to determine yet other electric power flow characteristics, for example using yet other first parameters. In some examples, the power change described above may be a change in active power consumed from the grid 200 by a power unit 219 or provided to the grid 200 by a power unit 2019. Alternatively, or additionally, in some examples, the power change described above may be a change in reactive power consumed from the grid 200 by a power unit 219 or provided to the grid 200 by a power unit 219. In some of the above examples, the data generating devices 218 send the sets of 5 event data to the computing system 205. However, it will be appreciated that this need not necessarily be the case and that in other examples, the data generating devices 218 may send the sets of event data to one or more other computing systems (not shown) for storage. In these examples, the computing system 205 may obtain the sets of event data from the one or more other computing systems (not shown). 10 The above examples are to be understood as illustrative examples of the invention. It is to be understood that any feature described in relation to any one example may be used alone, or in combination with other features described, and may also be used in combination with one or more features of any other of the examples, or any combination of any other of the examples. Furthermore, equivalents and 15 modifications not described above may also be employed without departing from the scope of the invention, which is defined in the accompanying claims.

Claims

1. A method of determining a value of an electric power flow characteristic of an electric power grid, the grid comprising a plurality of power units each configured to consume electric power from and / or provide electric power to the electric power grid, a change in consumption of electric power from or provision of electric power to the electric power grid by any particular one of the power units causing a respective change in a value of a first parameter of electric power flow in the electric power grid, the method comprising, at a computing system:determining a first plurality of sets of event data generated by a respective first plurality of data generating devices, each data generating device being associated with a respective one of a first plurality of the power units, each set of event data indicating i) a change in consumption of electric power from or provision of electric power to the electric power grid by the associated power unit and ii) a time of the change, wherein the time of the change indicated by each of the determined first plurality of sets of event data is within a first time window;determining grid response data, the grid response data indicating values of the first parameter of electric power flow in the grid as a function of time during the first time window; anddetermining, based the determined sets of event data, and based on the determined grid response data, the value of the electric power flow characteristic of the electric power grid.

2. The method according to claim 1,wherein the method comprises, for each of one or more of the first plurality of sets of event data:categorising the set of event data as belonging to a first category, responsive to determining that the change indicated by the set of event data results from a power flow change event that is correlated with one or more further power flow change events, the one or more further power flow change events each resulting in a change in consumption of electric power from orprovision of electric power to the electric power grid by a respective one or more power units that are not of the first plurality of power units; andwherein determining the value of the electric power flow characteristic of the electric power grid is based additionally on the categorisation of the one or more sets of event data as belonging to the first category.

3. The method according to claim 2, wherein the method comprises:excluding the one or more sets of event data categorised as belonging to the first category from being used in the determination of the value of the electric power flow characteristic of the electric power grid.

4. The method according to claim 2,wherein the method comprises parametrising the change indicated by the one or more sets of event data categorised as belonging to the first category with a second parameter, wherein the second parameter represents the proportion, of the total change in consumption or provision of power resulting from the correlated power flow change events, that the change indicated by the one or more sets of event data categorised as belonging to the first category represents; andwherein determining the value of the electric power flow characteristic is based on the parameterised change indicated by the one or more sets of event data categorised as belonging to the first category.

5. The method according to claim 4, wherein the method comprises determining a value of the second parameter concurrently with the determination of the value of the electric power flow characteristic.

6. The method according to any one of claim 2 to claim 5, wherein categorising the one or more of the first plurality of sets of event data as belonging to the first category comprises:analysing the first plurality of sets of event data to identify the one or more sets of event data.

7. The method according to claim 6, wherein analysing the first plurality of sets of event data to identify the one or more sets of event data comprises:for each of a plurality of second time windows within the first time window: determining a sum of the changes indicated by respective sets of eventdata that indicate a time within the second window;in response to a magnitude of the sum being larger than a first threshold value, categorising the sets of event data that indicate a time within the second time window as belonging to the first category.

8. The method according to any one of claim 1 to claim 7, wherein determining the value of the electric power flow characteristic of the electric power grid comprises fitting the first plurality of sets of event data, or first data derived from the first plurality of sets of event data, to the grid response data, or second data derived from the grid response data.

9. The method according to any one of claim 1 to claim 8, wherein determining the value of the electric power flow characteristic comprises:averaging the changes indicated by each of two or more of the first plurality of sets of event data or data derived therefrom;averaging a plurality of portions of the grid response data or data derived therefrom, wherein each portion of the response data or data derived therefrom corresponds to values of the first parameter within a third time window beginning at the time indicated by a respective one of the two or more of the first plurality of sets of event data; anddetermining the value of the electric power flow characteristic based on the averaged changes or data derived therefrom and the averaged portions of response data or data derived therefrom.

10. The method according to claim 9, wherein the grid comprises one or more compensating devices configured to provide power to or consume power from the electric power grid in response to a change in a value of the first parameter in order to compensate the change in the value of the first parameter, and wherein the third timewindow is set to have a duration less than a response time in which the one or more compensating devices provide power to or consume power from the electric power grid in response to the change in the value of the first parameter.

11. The method according to any one of claim 1 to claim 10, wherein determining the first plurality of sets of event data comprises:selecting the first plurality of sets of event data from a second plurality of sets of event data based on one or more attributes associated with each of one or more of the second plurality of sets of event data.

12. The method according to claim 11, wherein the one or more attributes comprise a location of the power unit associated with the set of event data, and wherein determining the first plurality of sets of event data comprises:selecting the first plurality of sets of event data from the second plurality of sets of event data based at least in part on the respective locations associated therewith.

13. The method according to claim 12, wherein the first plurality of sets of event data are selected based at least in part on the respective locations associated therewith each belonging to a common geographical region or region of the electric power grid.

14. The method according to claim 12 or claim 13, wherein each set of event data is indicative of the location of the associated power unit.

15. The method according to any one of claim 11 to claim 14, wherein the one or more attributes comprise one or more of a magnitude, rate, and duration of the change indicated by the set of event data, and wherein determining the first plurality of sets of event data comprises:selecting the first plurality of sets of event data from the second plurality of sets of event data based at least in part on one or more of the respective magnitudes, rates, and durations of the respective changes indicated thereby.

16. The method according to claim 15, wherein the first plurality of sets of event data are selected based at least in part on one or more of the respective magnitudes, rates, and durations of the respective changes indicated thereby being greater than a threshold magnitude, threshold rate, and threshold duration, respectively.

17. The method according to any one of claim 1 to claim 16, wherein determining the grid response data comprises:selecting the grid response data from a plurality of sets of grid response data based at least in part on a location at which the values of the first parameter indicated thereby were measured.

18. The method according to any one of claim 1 to claim 17, wherein the first parameter of electric power flow in the grid is a frequency of electricity flowing in the grid and the electric power flow characteristic of the electric power grid is inertia or wherein the first parameter of electric power flow in the grid is a difference between voltage phase angle at a first grid location and a second grid location and the electric power flow characteristic of the electric power grid is indicative of a current topology or operational state of the electric power grid.

19. The method according to any one of claim 1 to claim 18, wherein the method comprises:receiving the first plurality of sets of event data from the respective data generating devices over a network.

20. A method of providing one or more sets of event data for determining a value of an electric power flow characteristic of an electric power grid, the grid comprising a plurality of power units each configured to consume electric power from and / or provide electric power to the electric power grid, the method comprising, at a data generating device associated with a respective power unit:determining a change in consumption of electric power from or provision of electric power to the electric power grid by the associated power unit;determining that one or more of a magnitude, rate, and duration of the change is greater than a threshold magnitude, threshold rate, and threshold duration, respectively;in response to determining that one or more of a magnitude, rate, and duration of the change is greater than a threshold magnitude, threshold rate, and threshold duration, respectively:generating a set of event data indicating i) the change in consumption of electric power from or provision of electric power to the electric power grid by the associated power unit and ii) a time of the change; andsending the set of event data to a computing system over a network.

21. The method according to claim 20, wherein the set of event data further indicates a location of the associated power unit.

22. The method according to claim 20 or claim 21, wherein determining the change in consumption of electric power from or provision of electric power to the electric power grid by the associated power unit comprises:measuring and / or controlling the change in the consumption of electric power from or provision of electric power to the electric power grid by the associated power unit.

23. Apparatus configured to perform the method according to any one of claim 1 to claim 19, or the method according to any one of claim 20 to claim 22.

24. A computer program comprising instructions which, when executed by a computer, cause the computer to perform the method according to any one of claim 1 to claim 19, or the method according to any one of claim 20 to claim 22.

25. A system comprising an apparatus configured to perform the method according to any one of claim 1 to claim 19 and the data generating devices; an apparatus configured to perform the method according to any one of claim 20 to 22 and the computing system; or an apparatus configured to perform the method according to anyone of claim 1 to claim 19 and an apparatus configured to perform the method according to any one of claim 20 to claim 22.AMENDMENTS TO THE CLAIMS HAVE BEEN FILED AS FOLLOWS:-09 04 25CLAIMS1. A method of determining a value of an electric power flow characteristic of an electric power grid, the grid comprising a plurality of power units each configured to5 consume electric power from and / or provide electric power to the electric power grid, a change in consumption of electric power from or provision of electric power to the electric power grid by any particular one of the power units causing a respective change in a value of a first parameter of electric power flow in the electric power grid, the method comprising, at a computing system:10 determining a first plurality of sets of event data generated by a respective firstplurality of data generating devices, each data generating device being associated with a respective one of a first plurality of the power units, each set of event data indicating i) an amplitude of a change in consumption of electric power from or provision of electric power to the electric power grid by the associated power unit and ii) a time of15 the change, wherein the time of the change indicated by each of the determined first plurality of sets of event data is within a first time window;determining grid response data, the grid response data indicating values of the first parameter of electric power flow in the grid as a function of time during the first time window; and20 determining, based the determined sets of event data, and based on thedetermined grid response data, the value of the electric power flow characteristic of the electric power grid.

2. The method according to claim 1,25 wherein the method comprises, for each of one or more of the first plurality ofsets of event data:categorising the set of event data as belonging to a first category, responsive to determining that the change indicated by the set of event data results from a power flow change event that is correlated with one or more30 further power flow change events, the one or more further power flow changeevents each resulting in a change in consumption of electric power from or09 04 25provision of electric power to the electric power grid by a respective one or more power units that are not of the first plurality of power units; andwherein determining the value of the electric power flow characteristic of the electric power grid is based additionally on the categorisation of the one or more sets 5 of event data as belonging to the first category.

3. The method according to claim 2, wherein the method comprises:excluding the one or more sets of event data categorised as belonging to the first category from being used in the determination of the value of the electric power flow 10 characteristic of the electric power grid.

4. The method according to claim 2,wherein the method comprises parametrising the change indicated by the one or more sets of event data categorised as belonging to the first category with a second 15 parameter, wherein the second parameter represents the proportion, of the total change in consumption or provision of power resulting from the correlated power flow change events, that the change indicated by the one or more sets of event data categorised as belonging to the first category represents; andwherein determining the value of the electric power flow characteristic is based 20 on the parameterised change indicated by the one or more sets of event data categorised as belonging to the first category.

5. The method according to claim 4, wherein the method comprises determining a value of the second parameter concurrently with the determination of the value of the25 el ectri c power fl ow characteri sti c.

6. The method according to any one of claim 2 to claim 5, wherein categorising the one or more of the first plurality of sets of event data as belonging to the first category comprises:30 analysing the first plurality of sets of event data to identify the one or more setsof event data.09 04 257. The method according to claim 6, wherein analysing the first plurality of sets ofevent data to identify the one or more sets of event data comprises:for each of a plurality of second time windows within the first time window: determining a sum of the changes indicated by respective sets of event5 data that indicate a time within the second window;in response to a magnitude of the sum being larger than a first threshold value, categorising the sets of event data that indicate a time within the second time window as belonging to the first category.10 8. The method according to any one of claim 1 to claim 7, wherein determiningthe value of the electric power flow characteristic of the electric power grid comprises fitting the first plurality of sets of event data, or first data derived from the first plurality of sets of event data, to the grid response data, or second data derived from the grid response data.

159. The method according to any one of claim 1 to claim 8, wherein determining the value of the electric power flow characteristic comprises:averaging the changes indicated by each of two or more of the first plurality of sets of event data or data derived therefrom;20 averaging a plurality of portions of the grid response data or data derivedtherefrom, wherein each portion of the response data or data derived therefrom corresponds to values of the first parameter within a third time window beginning at the time indicated by a respective one of the two or more of the first plurality of sets of event data; and25 determining the value of the electric power flow characteristic based on theaveraged changes or data derived therefrom and the averaged portions of response data or data derived therefrom.

10. The method according to claim 9, wherein the grid comprises one or more 30 compensating devices configured to provide power to or consume power from the electric power grid in response to a change in a value of the first parameter in order to compensate the change in the value of the first parameter, and wherein the third time09 04 25window is set to have a duration less than a response time in which the one or more compensating devices provide power to or consume power from the electric power grid in response to the change in the value of the first parameter.5 11. The method according to any one of claim 1 to claim 10, wherein determiningthe first plurality of sets of event data comprises:selecting the first plurality of sets of event data from a second plurality of sets of event data based on one or more attributes associated with each of one or more of the second plurality of sets of event data.1012. The method according to claim 11, wherein the one or more attributes comprise a location of the power unit associated with the set of event data, and wherein determining the first plurality of sets of event data comprises:selecting the first plurality of sets of event data from the second plurality of sets 15 of event data based at least in part on the respective locations associated therewith.

13. The method according to claim 12, wherein the first plurality of sets of event data are selected based at least in part on the respective locations associated therewith each belonging to a common geographical region or region of the electric power grid.2014. The method according to claim 12 or claim 13, wherein each set of event data is indicative of the location of the associated power unit.

15. The method according to any one of claim 11 to claim 14, wherein the one or 25 more attributes comprise one or more of a magnitude, rate, and duration of the change indicated by the set of event data, and wherein determining the first plurality of sets of event data comprises:selecting the first plurality of sets of event data from the second plurality of sets of event data based at least in part on one or more of the respective magnitudes, rates, 30 and durations of the respective changes indicated thereby.09 04 2516. The method according to claim 15, wherein the first plurality of sets of event data are selected based at least in part on one or more of the respective magnitudes, rates, and durations of the respective changes indicated thereby being greater than a threshold magnitude, threshold rate, and threshold duration, respectively.

517. The method according to any one of claim 1 to claim 16, wherein determining the grid response data comprises:selecting the grid response data from a plurality of sets of grid response data based at least in part on a location at which the values of the first parameter indicated 10 thereby were measured.

18. The method according to any one of claim 1 to claim 17, wherein the first parameter of electric power flow in the grid is a frequency of electricity flowing in the grid and the electric power flow characteristic of the electric power grid is inertia or 15 wherein the first parameter of electric power flow in the grid is a difference between voltage phase angle at a first grid location and a second grid location and the electric power flow characteristic of the electric power grid is indicative of a current topology or operational state of the electric power grid.20 19. The method according to any one of claim 1 to claim 18, wherein the methodcomprises:receiving the first plurality of sets of event data from the respective data generating devices over a network.25 20. A method of providing one or more sets of event data for determining a valueof an electric power flow characteristic of an electric power grid, the grid comprising a plurality of power units each configured to consume electric power from and / or provide electric power to the electric power grid, the method comprising, at a data generating device associated with a respective power unit:30 determining a change in consumption of electric power from or provision ofelectric power to the electric power grid by the associated power unit;09 04 25determining that one or more of a magnitude, rate, and duration of the change is greater than a threshold magnitude, threshold rate, and threshold duration, respectively;in response to determining that one or more of a magnitude, rate, and duration 5 of the change is greater than a threshold magnitude, threshold rate, and threshold duration, respectively:generating a set of event data indicating i) an amplitude of the change in consumption of electric power from or provision of electric power to the electric power grid by the associated power unit and ii) a time of the change; and10 sending the set of event data to a computing system over a network.

21. The method according to claim 20, wherein the set of event data further indicates a location of the associated power unit.15 22. The method according to claim 20 or claim 21, wherein determining the changein consumption of electric power from or provision of electric power to the electric power grid by the associated power unit comprises:measuring and / or controlling the change in the consumption of electric power from or provision of electric power to the electric power grid by the associated power 20 unit.

23. Apparatus configured to perform the method according to any one of claim 1 to claim 19, or the method according to any one of claim 20 to claim 22.25 24. A computer program comprising instructions which, when executed by acomputer, cause the computer to perform the method according to any one of claim 1 to claim 19, or the method according to any one of claim 20 to claim 22.

25. A system comprising an apparatus configured to perform the method according 30 to any one of claim 1 to claim 19 and the data generating devices; an apparatus configured to perform the method according to any one of claim 20 to 22 and the computing system; or an apparatus configured to perform the method according to anyone of claim 1 to claim 19 and an apparatus configured to perform the method according to any one of claim 20 to claim 22.LDCM

Images