Method for detecting anomalies in a distribution network, in particular a water distribution network

Patent metadata

Field	Value
Publication number	US-10156465-B2
Application number	US-201515328515-A
Country	US
Kind code	B2
Filing date	Jul 23, 2015
Priority date	Jul 25, 2014
Publication date	Dec 18, 2018
Grant date	Dec 18, 2018

Abstract

A hydraulic model of the network is used that describes the structural data and the laws governing the distributed flows. The model is fed with operational data relative to primary input parameters and, using the model, theoretical values are obtained for primary output parameters. The theoretical values are compared with measured values. In case of a significant deviation, the corresponding primary output parameter becomes a secondary input parameter in an inverse model including new secondary output parameters, which are added relative to the direct model and which correspond to primary input parameters. The secondary output parameters are, by priority, those to which the primary output parameter exhibiting the abnormal value is particularly sensitive. If necessary, the method is implemented iteratively by progressively restricting the number of the secondary output parameters.

First claim

The invention claimed is:

1. A method for detecting anomalies in a fluid distribution network, the method comprising:
establishing, for the distribution network, a direct model comprising:
structural data of the distribution network relative to the topography of said distribution network and comprising setpoints or status data of equipment of said distribution network,
operational data relative to parameters chosen as primary input parameters describing an operational scenario, and
laws linking said structural data, the primary input parameters, and primary output parameters;

executing the direct model by a hydraulic simulation engine to determine theoretical values of the primary output parameters;
obtaining measurements from data acquisition systems of the primary output parameters;
comparing the theoretical values with the measurements;
when the comparing reveals at least one of the primary output parameters that is abnormal because of at least one significant deviation between at least one of the theoretical values and at least one of the measurements, establishing an inverse model comprising a minimization of a function of a deviation between measurement and simulation at a measurement point, relative to a configurable set of parameters, the inverse model being derived from the direct model by introducing the at least one abnormal primary output parameter as a secondary input parameter of the inverse model, the secondary input parameter taking the measurement as a value, the inverse model comprising at least one secondary output parameter in addition to the primary output parameters which are retained;
executing the inverse model by the hydraulic simulation engine; and
locating an anomaly of the distribution network from the execution of the inverse model.

2. The method according to claim 1 , wherein the at least one secondary output parameter is a parameter added to the inverse model, relative to the direct model.

3. The method according to claim 1 , wherein the at least one secondary output parameter corresponds to at least one deleted datum, relative to the direct model.

4. The method according to claim 1 , wherein the at least one secondary output parameter is selected by applying criteria indicating a likelihood that at least one new secondary output parameter is implicated in the at least one deviation observed between the at least one measurement and the at least one theoretical value.

5. The method according to claim 1 , wherein at least some of the operational data, the theoretical values, and the measurements comprise temporal series that is a series of values each associated with a schedule.

6. The method according to claim 5 , wherein the temporal series formed from values obtained from the measurements are smoothed following a configurable characteristic duration.

7. The method according to claim 1 , wherein, in an initialization step, an inverse model is used to determine certain structural or operational data that are impossible or difficult to obtain directly, the inverse model being executed by omitting values of certain primary output parameters.

8. The method according to claim 1 , wherein, in the execution of the direct model, calculations of sensitivity of at least some of the primary output parameters to the variations of at least some of the other parameters are incorporated, and
for the inverse model, primary input parameters to which the abnormal primary output parameter is more particularly sensitive are chosen as secondary output parameters.

9. The method according to claim 8 , further comprising:
obtaining, by the direct model, a deviation function of the value of the at least one deviation between the at least one theoretical value and the at least one measurement for the abnormal primary output parameter, as a function of the secondary output parameters;
calculating the gradient of the deviation function relative to the parameters according to the sensitivities;
determining, by application of the inverse model, a set of new data, modifying the initial data, and neutralizing positive and negative sensitivities for which the overall gradient of the deviation function exhibits a zero value;
from the parameters relative to which the sensitivities are calculated, selecting a subset of parameters that is formed from the parameters for which the sensitivity is of the same sign;
iteratively restricting the secondary output parameters to the parameters forming part of the subset until the anomaly is located with a single parameter or a small number of parameters.

10. The method according to claim 1 , wherein when a single primary output parameter exhibits an anomaly of the measured value relative to the theoretical value, a measurement error is determined.

11. The method according to claim 1 , wherein the direct model and the comparison of the theoretical and measured values are executed cyclically, and
wherein the execution of the inverse model is automatically activated when at least two primary output parameters exhibit an anomaly of the measured value relative to the theoretical value.

12. The method according to claim 1 , wherein the model elements further comprise indications of uncertainties on the data and the parameters, and
during the comparison, the uncertainties are taken into account to consider an output parameter measurement as being conformal or abnormal.

13. The method according to claim 1 , further comprising improving a performance level of the distribution network.

14. The method according to claim 2 , wherein the at least one secondary output parameter corresponds to at least one deleted datum, relative to the direct model.

15. The method according to claim 2 , wherein the at least one secondary output parameter is selected by applying criteria indicating a likelihood that at least one new secondary output parameter is implicated in the at least one deviation observed between the at least one measurement and the at least one theoretical value.

16. The method according to claim 3 , wherein the at least one secondary output parameter is selected by applying criteria indicating a likelihood that at least one new secondary output parameter is implicated in the at least one deviation observed between the at least one measurement and the at least one theoretical value.

17. The method according to claim 2 , wherein at least some of the operational data, the theoretical values, and the measurements comprise temporal series that is a series of values each associated with a schedule.

18. The method according to claim 3 , wherein at least some of the operational data, the theoretical values, and the measurements comprise temporal series that is a series of values each associated with a schedule.

19. The method according to claim 4 , wherein at least some of the operational data, the theoretical values, and the measurements comprise temporal series that is a series of values each associated with a schedule.

20. The method according to claim 2 , wherein, in an initialization step, an inverse model is used to determine certain structural or operational data that are impossible or difficult to obtain directly, the inverse model being executed by omitting values of certain primary output parameters.

Assignees

• Suez Groupe

Inventors

• Jarrige Pierre-Antoine
• Gancel Guillaume
• Campan Francis

Classifications

• G06F2113/14

  Pipes · CPC title

• G01F15/068 Primary

  with electrical means (G01F15/063 takes precedence) · CPC title

• G06Q50/06 Primary

  Energy or water supply · CPC title

• G06F30/20

  Design optimisation, verification or simulation (optimisation, verification or simulation of circuit designs G06F30/30) · CPC title

• G01L19/083

  electrical · CPC title