System and method for identifying related events in a resource network monitoring system

• US 8,341,106 B1
• Filed: 12/07/2011
• Issued: 12/25/2012
• Est. Priority Date: 12/07/2011
• Status: Active Grant

• Current Assignee: TaKaDu Ltd.
• Original Assignee: TaKaDu Ltd.
• Inventors: Scolnicov, Haggai, Armon, Amitai, Linhart, Chaim, Bien, Lilach, Petrank, Noam
• Primary Examiner(s): FERNANDEZ RIVAS, OMAR F
• Application Number:
  US13/313,261
• Time in Patent Office: 384 Days
• Field of Search: None
• US Class Current: 706/47
• CPC Class Codes: G01F 1/50 Correcting or compensating means
• CPC Class Codes: G01F 1/74 Devices for measuring flow of a fluid or flow of a fluent solid material in suspension in another fluid
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks
• CPC Class Codes: G06Q 50/06 Energy or water supply

Abstract

A computerized method for detecting related events in a water monitoring system. The method includes receiving event data from a plurality of sensors in a water network. The method also includes identifying at least two candidate events from the event data as being candidate events which are likely to constitute part of a common anomaly event, the candidate events each being substantially anomalous. An event combination rule is selected. One or more event characteristics are compared between the at least two candidate events based on the event combination rule. The at least two candidate events are determined whether they are related and associated with a common event causing the at least two candidate events on the basis of the comparison. The determination of the at least two candidate events are related and associated with a common event is reported to a user via a user interface.

Claims

1. A computerized method performed using a processor for identifying related events in a water network monitoring system, the events represented by stored event data derived from data received from one or more sources including sensor data received from a plurality of sensors in the water network, each event data comprising data identifying an event and one or more characteristics of the event, the method comprising:

retrieving event data from an event database, the event data representing a plurality of events which have been processed from anomalies received from the water network monitoring system and which each contain a plurality of data fields including event start time and event magnitude;

identifying at least two events from the event data as being candidate events that are likely to constitute parts of a composite event, the identified candidate events each being determined by the water network monitoring system not to be statistically significant enough to be reported as events by themselves;

selecting an event combination rule, the event combination rule including one or more tests to determine whether the candidate events are related and can be combined to form a composite event of a given event type;

comparing one or more event characteristics between the at least two candidate events based on the event combination rule;

determining, based at least in part on the comparison, that at least two candidate events are related to one another and are processible as a single event of the given event type; and

reporting the determination of the at least two candidate events as being related and as the single event of the given event type to a user via a user interface.

2. The method of claim 1 wherein the one or more event characteristics include at least one of flow rate, pressure, reservoir level, turbidity, chlorine level, and pH.

3. The method of claim 1 comprising detecting a continued or frequent occurrence of an anomaly from which an event is derived over multiple time periods.

4. The method of claim 3 wherein the one or more tests includes a test for determining anomalies to be significant based on the detected continued or frequent occurrence.

5. The method of claim 1 further comprising associating a score with the anomaly, indicating the probability of such an anomaly occurring randomly, without any unusual event in the system being monitored.

6. The method of claim 1, wherein determining the at least two candidate events are related and associated with a composite event comprises determining a related location corresponding to the at least two candidate events.

7. The method of claim 1 further comprising determining a start time of the at least two candidate events.

8. The method of claim 7 wherein determining a start time of the at least two candidate events comprises determining a distinct deviation between a sensor'"'"'s data before the start time and the sensor'"'"'s data after the start time.

9. The method of claim 8 wherein determining a start time of each of the at least two candidate events comprises:

calculating a score for a pair of intervals around each of a plurality of possible start times based on the formula;

|mean(interval1)/var(interval1)−

mean(interval2)/var(interval2)|; and

finding a start time which maximizes the score calculated for the pair of intervals around each possible start time based on the formula.

10. The method of claim 8 further comprising determining a clear start time including:

comparing whether a first percentage of the data in a first interval before a start time is greater or smaller than a second percentage of the data in a second interval after the start time.

11. The method of claim 10 wherein the first and second percentages are 100%.

12. The method of claim 7 wherein determining the at least two candidate events are related and associated with a composite event comprises determining the at least two candidate events having near start times.

13. The method of claim 12 wherein determining the at least two candidate events having near start times comprises determining an amount of data samples between the at least two candidate events based on parameters from the event combination rule.

14. The method of claim 1 wherein determining the at least two candidate events are related and associated with a composite event comprises determining the at least two candidate events having matching anomaly directions.

15. The method of claim 14 wherein matching anomaly directions includes the at least two candidate events associated with measurement increases.

16. The method of claim 14 wherein matching anomaly directions includes the at least two candidate events associated with measurement decreases.

17. The method of claim 14 wherein matching anomaly directions includes one of exactly two candidates associated with a measurement increase and the other associated with a measurement decrease.

18. The method of claim 1 wherein determining the at least two candidate events are related and associated with a composite event comprises determining the at least two candidate events having matching anomaly magnitudes.

19. The method of claim 18 wherein matching anomaly magnitudes includes the at least two candidate events associated with measurement increases of at least substantially similar magnitude.

20. The method of claim 18 wherein matching anomaly magnitudes includes the at least two candidate events associated with measurement decreases of at least substantially similar magnitude.

21. The method of claim 18 wherein matching anomaly magnitudes includes the at least two candidate events being associated with measurement increases and decreases which sum to approximately zero.

22. The method of claim 1 wherein the composite events of the given event type is one of at least DMA breach, leak, burst, and pressure drop.

23. A computerized method performed using a processor for identifying related events in a network monitoring system, the events being derived from anomalies detected as occurring in the network being monitored, each event data comprising data identifying a type of event and one or more characteristics of the event, the method comprising:

retrieving event data from an event database, the event data representing a plurality of events which have been processed from anomalies received from the network monitoring system and which each contain a plurality of data fields including event start time and event magnitude;

retrieving a first event combination rule from a stored collection of event combination rules, the event combination rules each comprising one or more parameters for identifying candidate events that are likely to constitute parts of a composite event and one or more tests to determine whether the candidate events are related and can be combined to form a composite event of a given event type, wherein the identified candidate events are determined by the network monitoring system not to be statistically significant enough to be reported as events by themselves, the one or more tests including a first test for determining whether two or more candidate events occurred at related locations in the network, a second test for determining whether two or more candidate events occurred at near start times to one another, and a third rule for determining a permitted set of magnitudes or directions for the anomalies from which the candidate events were determined;

applying the first event combination rule to identify at least two events from the event data as being candidate events;

comparing one or more event characteristics between the at least two candidate events based on one or more of the tests included in the first event combination rule;

determining, based at least in part on the comparison, that at least two candidate events are related to one another and are processible as a single event of the given event type; and

reporting the composite event as the single event of the given event type to a user via a user interface.

24. The method of claim 23 wherein the second test for determining near start times of the at least two candidate events includes one or more parameters for identifying possible start times for the at least two candidate events, determining for each candidate event whether the most likely start time is a clear anomaly start time, and determining a difference between the clear start times for the candidate events.

25. The method of claim 23, wherein the first event combination rule includes one or more parameters for selecting an anomaly detector in the network management system, and wherein applying the first event combination rule to identify two or more candidate events comprises selecting the anomaly detector to be used to detect the anomalies from which the candidate events are determined, and generating the candidate events.

26. The method of claim 25, wherein the first event combination rule includes one or more parameters for selecting events as candidate events, and wherein applying the first event combination rule to identify two or more candidate events comprises selecting the candidate events from a set of events stored in an event database.

27. The method of claim 23, wherein the network is a water delivery network.

28. The method of claim 23, wherein comparing one or more event characteristics between the at least two candidate events comprises comparing the characteristics based on all the tests included in the first event combination rule.

29. A computerized method performed using a processor for identifying related events in a water network monitoring system, the events represented by stored event data derived from data received from one or more sources including sensor data received from a plurality of sensors in the water network, each event data comprising data identifying an event and one or more characteristics of the event, the method comprising:

retrieving event data from an event database, the event data representing a plurality of events which have been processed from anomalies received from the water network monitoring system and which each contain a plurality of data fields including event start time and event magnitude;

selecting an event combination rule, the combination rule including one or more tests to determine whether the plurality of events are related and can be combined to form a composite event of a given event type, the candidate events determined by the water network monitoring system not to be statistically significant enough to be reported as events by themselves;

generating a plurality of candidate events based on the event combination rule;

identifying, based at least in part on the comparison, a subset of the candidate events are related to one another and are processible as a single event of the given event type; and

reporting the identification of the subset of candidate events as being related and as the single event of the given event type to a user via a user interface.

30. A system for identifying related events in a water network monitoring system, the system comprising:

one or more sensors in the water network;

an event data store storing a plurality of event data derived from data received from the one or more sensors, the plurality of event data representing a plurality of events which have been processed from anomalies received from the water network monitoring system and which each contain a plurality of data fields including event start time and event magnitude; and

a composite event subsystem operative to;

retrieve the plurality of event data from the event data store;

identify at least two events from the plurality of event data as candidate events likely to constitute parts of a composite event, the candidate events determined by the water network monitoring system not to be statistically significant enough to be reported as events by themselves;

select an event combination rule, the combination rule including parameters for generating candidate events and one or more tests to determine whether the candidate events are related and can be combined to form a composite event of a given event type;

compare one or more event characteristics between the at least two candidate events based on the event combination rule;

determine, based at least in part on the comparison, that at least two candidate events are related to one another and are processible as a single event of the given event type; and

report the determination of the at least two candidate events as being related and as the single event of the given event type to a user via a user interface.