Method for determining events in a network

Abstract

A method for determining an event in a network (1) of consumption meters (5), wherein an event is determined by grouping data characterizing the event according to the consumption meters (5) that have detected the characterizing data. If the group of consumption meters matches, then the event is determined, the grouping being performed according to the following criteria: - the temporal coincidence of the events, - the spatial correspondence of the events, - the type consistency of the events. The spatial correspondence is determined by determining the Euclidean distance of each of a plurality of consumption meters (5) from each of the other consumption meters (5) of the plurality and ranking the Euclidean distances, and assigning to the group of consumption meters (5) the n closest consumption meters (5) that have detected the characterizing data.

Description

[0001] The present invention relates to a method for determining events in a network of consumption meters, particularly a network of power meters.

[0002] In current power supply networks used to supply electrical power to homes and industries, the goal is to provide electrical power with the most constant voltage possible. However, in practice, events that cause deviations exist. There are very limited local events and other events involving several consumers. Distinguishing these events, especially distinguishing between significant and insignificant events, is always a problem. To cluster these events, at least information about the start time, duration, event type, and event location is required. This problem is known from US10,809,885B2. In that document, a method is disclosed to use different event reports from various monitoring and reporting systems and identify those related events to create a composite fault map of the power distribution system.

[0003] Generally, checking the timing overlap of events by examining their start and duration, or by checking their start and stop times, is not a problem. Checking the type of event is also not a problem. In power supply networks, there are typically three important events that need to be known: undervoltage, overvoltage, and power outages. However, pinpointing the location of an event can be difficult. If metering, especially power meters, is used to capture events, locating those meters that have detected the event can be problematic. Depending on the density of metering (meters per square kilometer), there may be many meters close together or only a few widely dispersed. This depends on whether the meters are located in a town or a rural area. For example, there may be 20 metering meters in the same apartment building, with many meters less than 50 meters apart.

[0004] From a power perspective, what matters is not the Euclidean distance between meters. Instead, what matters is the connectivity along the cable. Meters connected to the same substation or the same radial line of a substation experience similar voltages and settings. Using Euclidean distance between events is problematic in order to enable algorithms to cluster events that are close together from a power perspective and to avoid clustering distant and unrelated events.

[0005] This problem is addressed by a method for determining events in a network of consumption meters, particularly in a network of power meters. Furthermore, a head-end system according to the invention is specified for performing this method.

[0006] The method for determining events in a network of consumption meters according to the present invention operates as follows: events are determined by grouping data characterizing events and already detected by the consumption meters. If a group of consumption meters matches, it is determined to be an event. Grouping is performed according to the following criteria: a. Time overlap of events b. Spatial consistency of events c. Consistency of event types.

[0007] According to the present invention, spatial consistency is determined by: determining the Euclidean distance between each of a plurality of consumption tables and each of the other consumption tables in the plurality of consumption tables and sorting the Euclidean distances, and assigning the n nearest consumption tables that have detected characterization data to a consumption table group.

[0008] The main idea of ​​this invention is not to use the Euclidean distance between the consumption tables, but to sort the Euclidean distances between each of the multiple consumption tables and each of the other consumption tables in the multiple consumption tables, and then assign the n nearest consumption tables that have detected the characterization data to the consumption table group.

[0009] A major advantage of this method is that it is more or less independent of the Euclidean distance between the consumption tables. This means that the method is suitable for consumption tables that are placed relatively close to each other (since they may be located in cities) as well as consumption tables that are quite far apart (since they may be located in rural areas). Since only the ordering is considered, this is independent of the distance between them. Another major advantage of this method is that the ordering only needs to be performed once, and in practice, it is performed periodically as consumption tables are added to or removed from the network.

[0010] The Euclidean distance between two meter readings is the length of the line segment between them. This distance can be calculated using Pythagorean theory based on the Cartesian coordinates of the meter readings. In practice, this would be the distance calculated from the GPS data of the two meter readings. This means that differences in altitude are not considered.

[0011] However, according to another embodiment of the invention, a third dimension can be considered. This may be relevant in cities with skyscrapers, where several consumption meters can be found at the same GPS location but at different altitudes.

[0012] Advantageously, grouping is performed by first grouping events with the same event type, then grouping events that match in time, and finally by determining spatial consistency.

[0013] It is beneficial to group by sorting the distances of the n nearest consumption tables to the events that have been detected. Preferably, this number n is between 20 and 100, particularly 50. The number n used may depend on the type of consumption table. This number can be adjusted if necessary.

[0014] Regarding the time overlap of events, it is advantageous to divide the time into the event start time and the event duration, or the event duration itself. Clearly, if time detection is separated, the start and end times of the event, or the start and duration, are required. For completeness, this could also be done using the event duration and the event end time, although this seems less feasible.

[0015] It can calculate the duration of an event, which is the time difference between the end of the event and the beginning of the event.

[0016] To assess whether the start and / or end times of events are considered to belong to the same event, these (start or end) times should not differ by more than a time difference limit, preferably less than 15 seconds. Otherwise, it is highly certain that these events do not belong together.

[0017] For durations considered to be the same event, the durations should not differ by more than a second time difference limit, which is preferably less than 30 seconds.

[0018] The method according to the invention can be used with consumption meters such as water meters, gas meters, and heat meters. However, it is preferred to use the method with power meters, as they are installed in every household. Power meters can be used to detect events in the power supply network. If an event in the power grid needs to be determined, the type of event can be undervoltage, overvoltage, deviation of the mains frequency, or power outage. Of course, the invention is not limited to this type of event, but these events should generally be detected.

[0019] Standards contain limits for undervoltage, overvoltage, and mains frequency deviations, and these limits should be maintained. Generally, identifying events that do not conform to the standard is advantageous. However, sometimes events exist within the range given by the standard, but at the end of the tolerance limits. To identify these events, an adjustable trigger level with undervoltage, overvoltage, and / or mains frequency deviations is advantageous. Using this feature, specific events that should be searched can be specified.

[0020] To determine spatial consistency, it might be sufficient to sort the Euclidean distances between multiple cost tables or between groups of cost tables. However, the best and simplest way would be to determine the Euclidean distance between each of all cost tables in the network and each of all other cost tables in the network. Events in the entire network can then be determined, and if this has been done, the sorting of all cost tables in the network can be used to further determine events in the network without updating that sorting.

[0021] Sometimes, one or more events may be determined by a single consumption table. In this case, the single consumption table should be marked as defective and replaced by another consumption table. That consumption table is no longer used to determine events in the network.

[0022] According to a further development of the invention, it may be advantageous if a first time difference constraint is used to compare events reported from the same consumption scale, while a second time difference constraint is used to compare events reported from different consumption scales, wherein the second time difference is smaller than the first time difference.

[0023] By using two distinct time difference constraints, the filter effect can be customized for reported events originating from the same table and events from different tables. For events from the same table, a larger time difference is accepted; for events from different tables, only a smaller time difference is accepted.

[0024] The method according to the invention can be extended. Once an event detected by the consumption meter has been added to the event group, the reported data from the next most recent consumption meter is evaluated according to criteria (i.e., time overlap and event type consistency), and if the criteria are met, the event is added to the event group.

[0025] This method automatically expands the search space. It checks whether the consumption tables immediately following the sort also meet the criteria.

[0026] To determine events in the network, data from the metering devices is transmitted to a headend system, where this data is collected and evaluated. These headend systems are known to receive data from power meters or heat meters. The data is primarily transmitted wirelessly to a headend system suitable for receiving it. According to the invention, such a headend system, connected to the data of multiple metering devices, is configured to perform the method described above for determining events in the metering device network. Advantageously, this headend system is configured to communicate with the metering devices to, for example, adjust the trigger levels of the metering devices. Therefore, the configuration of the headend system and the adjustment of the trigger levels of the metering device group are feasible.

[0027] It would be further advantageous to configure the headend system to calculate and sort the Euclidean distance between each consumption table and every consumption table connected to the headend system before determining multiple events. If new consumption tables have been added and old consumption tables have been removed or replaced, this sorting may need to be updated from time to time.

[0028] While power meters typically have enough power to continuously send data, it would be advantageous to send event data to the headend system at time intervals. This would be even more important for all battery-powered meters not connected to electricity. Event data could potentially be sent in batches every six hours, or even daily.

[0029] The Euclidean distance between tables is determined most simply by using their GPS data. When this method / algorithm is used for the first time, the Euclidean distance sorting only needs to be performed once at the beginning. Distance sorting is only required again if the number of tables changes or if a table is physically moved. This means savings in time and computational power, as sorting is not necessary every time new event data arrives. This sorting is highly advantageous if the consumption tables are close to each other (e.g., in urban areas) or if they are farther apart (e.g., in rural areas), as it is independent of the actual distance between the tables. This sorting allows for determining the spatial relevance of events independently of distance.

[0030] The present invention is described in detail with reference to the following figures: Figure 1 A schematic diagram of the power supply network. Figure 2a A first example for sorting distances. Figure 2b A second example for sorting by distance relative to another table. Figure 3 Three events of the same type with a time difference, and Figure 4 A flowchart illustrating the methods used to determine events in a network.

[0031] exist Figure 1 In the diagram, power supply network 1 is represented by a symbol, which shows a large number of power consumers scattered throughout various houses 2 in towns and villages. Each house includes at least one power meter 5 configured to detect electrical events in network 1. Figure 1 (Not visible in the image). Events are represented by the symbol 3, which is shown as a solid line, dotted line, or dashed line, and uses symbols to represent different detected events. These events may differ in time, type, and location.

[0032] These detected events are detected by power meters 5. These power meters 5 wirelessly transmit signals corresponding to the detected events to a headend system 4, which is configured to evaluate these signals for grouping, spatially, temporally, and by type, assigning the events. Since this is power network 1, events may differ in type, such as overvoltage, undervoltage, mains frequency, or power outage. In all meters 5 installed in house 2, the detection level for undervoltage and overvoltage is the same. This detection level corresponds to a standard, so if the voltage deviates from the rated voltage by more than 15%, it is determined to be overvoltage / undervoltage. If the voltage measured by the meters 5 installed in house 2 is less than 85% of the nominal voltage, it is determined to be undervoltage. For events to be grouped, they must be matched in both type and time. Time is divided into the event start time and the event duration. There is a tolerance of approximately 60 seconds for the start time and approximately 120 seconds for the event duration.

[0033] In head-end system 4, time tolerances and event types are evaluated. However, head-end system 4 is configured to communicate with table 5 so as to not only receive but also transmit data. Therefore, head-end system 4 can adjust the undervoltage and overvoltage detection levels of the group of tables or all tables 5 connected to it. This can be used to specify grouping for specific events.

[0034] exist Figure 3 The diagram illustrates three events A, B, and C of the same type, such as undervoltage events with different start times. A first meter detects event A starting at time t1. Another meter detects event B starting at time t2. It can be seen that events A and B have the same duration but different start times t1 and t2. The start times t1 and t2 are sent by the meter to a headend system, which calculates the time difference between t2 and t1. The headend system then limits the time difference of the start times to tdl. s Compare with the time difference between t1 and t2, and when that time difference is less than tdl s Events A and B are grouped together, meaning they are considered to be time-overlapping events. Event C of the same type is detected from the third consumption table, and its start time t3 and duration are sent to the headend system. The headend system calculates the difference between start time t3 and t1, t2, and t3. This difference is less than tdl. s However, the duration of event C is much longer than the durations of events A and B (which are the same). The duration of event C is too long, therefore event C is not grouped with events A and B. The duration of event C exceeds the time difference limit (tdl). dIn practice, there are two different time difference constraints for events: one for the start time and the other for the time difference constraint tdl. s and time difference constraints for duration (TDL) d Because there may be time fluctuations between the start and stop times, the time difference limit for duration is usually greater than the time difference limit for start time.

[0035] However, a very important factor in grouping is spatial consistency. This is accomplished by sorting a table by Euclidean distance to all other tables. This sorting must be performed for each table.

[0036] In practice, we first check the consistency of event types, then check the temporal overlap of events, and finally check the spatial consistency of events.

[0037] exist Figure 2a and Figure 2b In Table 5, eight tables arranged at different distances from each other are represented by small circles. Figure 2a In the table, the distance sorting for the table represented by the black circle symbol is performed relative to the other seven tables represented by the circle symbol. From Figure 2a As can be seen, the table represented by the black circle symbol has no distance from itself, so it is sorted as 0. The nearest table 5 is sorted as 1, the second nearest table is sorted as 2, the third nearest table is sorted as 3, and so on, until the table with the longest distance from the (black) table is sorted as 7.

[0038] This sorting must be performed on each of Table 5. Figure 2b In, with Figure 2a The sorting is performed in the same way as described in the figure; however, this sorting is performed relative to another table, Table 5, on the right side of the figure. Figure 2b Table 5, indicated by the black circle symbol, must be sorted relative to other Table 5s. The nearest neighbor table is sorted as 1, and so on, until the furthest table is sorted as 7. For example... Figure 2a neutralization Figure 2b For the two different Table 5, it is necessary to specify the installation location. Figure 1 This sorting is performed on each table 5 in house 2 of network 1. The sorted distances for each table can be set in a distance matrix (such as matrix 1 below). For simplicity, this matrix only involves the distances of four tables. These tables are numbered 1, 2, 3, and 4. The distance between table 1 and table 2 is 10 meters. The distance between table 1 and table 3 is 15 meters. The distance between table 1 and table 4 is 20 meters. The distance between table 2 and table 3 is 6 meters, and the distance between table 3 and table 4 is 7 meters. Matrix 1:

[0039] According to the invention, the Euclidean distance is replaced by a sorting of distances, as can be shown in matrix 2 below. Such a matrix must be created, including all tables 5 in network 1. This matrix 2 is used to determine the spatial consistency of events, and the sorting of distances between the n nearest tables 5 to which events have been detected is used for grouping. This n is, for example, 50, and can vary depending on the type of tables and network. Matrix 2:

[0040] from Figure 2a and Figure 2b As can be seen, this ranking is independent of the true Euclidean distance. For example, in Figure 2a In this context, n is set to five (five tables must be sorted), and the distance between the tables in region 6 is much smaller than 1. Figure 2b The distances in region 7 of the data are ordered the same.

[0041] In practice, there may be a million or more consumption tables that need to be sorted. Even with a high number of tables, the storage space and computing power required for this sorting are relatively low.

[0042] Figure 4 A flowchart illustrating this grouping method is shown. To begin this method for determining events, the first step 10 is to determine the Euclidean distance between each meter 5 installed in the house 2 connected to network 1. There may be 100 meters or 1,000 meters or more; the distances between these meters must be determined and sorted. The Euclidean distances between these power meters 5 are typically unknown, and the easiest way is to obtain the addresses of the houses 2 where the meters 5 are installed and use the GPS data from those addresses to determine the distances between the meters 5.

[0043] After determining the distances between all tables in Table 5, the next step in step 11 is to sort these distances. This has already been done based on... Figure 2a and Figure 2b The example uses sorting matrices 1 and 2 to show how to perform the sorting.

[0044] After sorting the distances in Table 5 in step 11, this sorting of all tables is stored in headend system 4 (step 12). Then, in step 13, event detection begins by receiving signals of detected events from Table 5. These signals are sent to headend system 4, where they are stored in a central server, which may be part of headend system 4 or located at a distance from headend system 4, and where event grouping is calculated.

[0045] After receiving and storing this event data, it must be evaluated. Based on... Figure 4 In this embodiment, the event data received from the table and stored in the central server in step 13 is calculated in terms of the duration of the event (step 14). These event data are then stored in event queue 15, and an identification number is assigned to each (group of) event data, including the ID number identifying the meter. This ID number is fixed at installation time and is included in the wireless data telegram when the meter sends data.

[0046] The event data in event queue 15 are processed one by one, starting with the first event data, which is sorted by number 1 and identified by type and time. This event data is stored in event cluster 16. Then, the second event data in event queue 15 is processed. In step 17, it is checked whether the event data comes from the same table as the first event data or from a different table. If the event comes from a different table than the first event data, then in step 18, the distance sorting relative to the first event data is checked. In step 19, it is further checked whether the second event data is within a time threshold (i.e., time difference limit) relative to the first event data. Finally, in step 20, it is checked whether these event data are of the same type. If all three steps 18, 19, and 20 have a positive result, the event data is added to the corresponding cluster 21. If any of these steps 18 to 20 are not positive, the event data is added to event queue 15. Event cluster 16 is the result of grouping the event data, and each cluster involves the same event that has already been detected in the network of the consumption table.

[0047] Therefore, all event data in event queue 15 is processed in this manner, and these event data are grouped into clusters arranged in event cluster 16. This processing of the event data compares each event data with every other event data already grouped in event cluster 16. This fact is important for step 18. Since each event data is compared with every event data in event cluster 16, it is not necessary for the distance sort to fit every event data in event cluster 16, but only for the distance sort set to fit one of the event data in event cluster 16. This processing produces larger event clusters and ensures that all event data originating from the same event are grouped into the same event cluster.

[0048] If the event data is identified as originating from the same table in step 17, a further check is performed to determine if it is the same type of event (step 22), and a further check is performed in step 23 to determine if the event data corresponds to a time constraint. If both results of steps 22 and 23 are positive, the event data is considered to be the same event data, and the table containing this event data has been identified as incorrectly functioning. If one or both results of steps 22 and 23 are negative, event data belonging to another cluster within event cluster 16 is checked in steps 18, 19, and 20. These check boxes 18, 19, and 20 are performed for each event data in event cluster 16. If the event data does not belong to any cluster in event cluster 16, it is removed from the event queue 15.

[0049] According to Figure 4 The flowchart discloses further details of the method for determining events in a network, which details these details in detail: 24 Considered a single event; 25 Considered separate events, remain in the event queue; 26 View events from the queue; 27 View events from clusters; 28 Any events remain in the selection?; 29 Events were searched from the built clusters?; 30 Event does not belong to a cluster. Remove from the event queue; 31 No new events to add to clusters; 32 Uninvestigated events remain in the clusters?; 33 Investigate events in the clusters to see if their neighboring events also belong to the clusters; 34 Clustering complete; 35 Events remain in the event queue?; 36 All events have been scanned and all clusters have been built. The program ends until a new event is received. Tag list 1 Network 2. Houses with tables inside 3 circles 4-head system Table 5 6 regions 7 regions TDL time difference limit tdl s Start time time difference limit tdl d Duration time difference limit 10. First step: Calculation 11. Second step: Sorting 12 steps, storage 13 steps: Receive event data and store it. 14 steps to calculate the duration 15 Event Queues 16 clustering queues 17 steps to check where the event originated. 18 steps to check the sorting 19 steps, check time 20 steps to check the type Step 21: Add the event to the cluster. 22 steps, check type 23 steps, check time The 24 steps are considered a single event. Step 25 is considered a separate event and remains in the event queue. 26 steps to view events from the queue 27 steps to view events from clustering 28 steps, any event left in the selection Step 29: Are events searched from the established clusters? Step 30: The event does not belong to a cluster. Remove it from the event queue. Step 31: No new events to add to the cluster. 32 steps, uninvestigated events remain in the cluster. Step 33: Investigate the events in the cluster to see if their neighboring events also belong to the cluster. 34 steps to complete clustering Step 35: The event remains in the event queue. The program completed 36 steps, scanning all events and building all clusters. It then terminated until a new event was received.

Claims

1. A method for determining events in a network (1) of a consumption table (5), wherein events are determined by grouping data that has been detected by the consumption table (5) and characterizes events, and if a consumption table group matches, it is determined to be an event, said grouping being performed according to the following criteria: - Time overlap of events - Spatial consistency of events - Consistency of event types Its features are, The spatial consistency is determined by: determining the Euclidean distance between each of the plurality of consumption tables (5) and each of the other consumption tables in the plurality of consumption tables (5) and sorting the Euclidean distances, and assigning the n nearest consumption tables to the consumption table (5) group that have already acquired the data.

2. The method according to claim 1, characterized in that, The grouping is performed by first grouping events with the same event type, then grouping events that match in time, and finally determining the spatial consistency.

3. The method according to claim 1 or 2, characterized in that, Grouping is performed using the sorting of the distances between the n nearest consumption tables (5) of the detected events, where n is between 20 and 100.

4. The method according to claim 1 or 2, characterized in that, Time detection is divided into the start time of the event and the duration or end time of the event.

5. The method according to claim 1 or 2, characterized in that, If the start time and / or end time of an event do not differ by more than the first time difference limit (tdl) s If the start time and / or end time of the event are the same, then the first time difference is limited to less than 15 seconds.

6. The method according to claim 5, characterized in that, If the durations of the events do not differ by more than the second time difference limit (tdl) d If the duration of the events is considered to be the same event, then the second time difference limit is less than 30 seconds.

7. The method according to claim 1 or 2, characterized in that, The consumption meter (5) is an electrical power meter.

8. The method according to claim 1 or 2, characterized in that, The types of events are undervoltage, overvoltage, mains frequency deviation, or power outage.

9. The method according to claim 1 or 2, characterized in that, The trigger level for undervoltage, overvoltage, and / or mains frequency deviation is adjustable.

10. The method according to claim 1 or 2, characterized in that, The Euclidean distance between each of all consumption tables (5) of the network (1) and each of the other consumption tables (5) of the network (1) is determined by using the geographical or address data of the tables via the Internet.

11. The method according to claim 1 or 2, characterized in that, If one or more events are identified in a consumption table (5), then only the consumption table (5) is marked as defective.

12. The method according to claim 6, characterized in that, First time difference limit (tdl) s ) is used to compare events reported from different consumption scales, while the second time difference limit (tdl) d () is a time difference limit for the duration of an event, where the first time difference limit is less than the second time difference limit.

13. The method according to claim 1 or 2, characterized in that, Once an event detected by the consumption scale has been added to an event group, the reported data from the next most recent consumption scale is evaluated according to the time overlap criterion and the event type consistency criterion, and if the criteria are met, the event is added to the event group.

14. The method according to claim 1 or 2, characterized in that, Grouping is performed using the distance sorted from the n nearest consumption tables (5) of the events that have been detected, where n is 50.

15. A head-end system, the head-end system being data-connected to a plurality of consumption meters (5), the head-end system being configured to perform the method according to any one of the preceding claims.

16. The head-end system according to claim 15, characterized in that, The head-end system is configured to adjust the trigger level of the consumption meter (5).

17. The head-end system according to claim 15 or claim 16, characterized in that, The head-end system is configured to periodically calculate and sort the Euclidean distance between each consumption table (5) and each consumption table (5) connected to the head-end system (4) before and / or after the multiple events have been determined.

Images