초록 ▼

Technical solutions are described for forecasting leaks in a pipeline network. An example method includes identifying a subsystem in the pipeline network that includes a first station. The method also includes accessing historical temporal sensor measurements of the stations. The method also include

Technical solutions are described for forecasting leaks in a pipeline network. An example method includes identifying a subsystem in the pipeline network that includes a first station. The method also includes accessing historical temporal sensor measurements of the stations. The method also includes generating a prediction model for the first station that predicts a pressure measurement at the first station based on the historical temporal sensor measurements at each station in the subsystem. The method also includes predicting a series of pressure measurements at the first station based on the historical temporal sensor measurements. The method also includes determining a series of deviations between the series of pressure measurements and historical pressure measurements of the first station and identifying a threshold value from the series of deviations, where a pressure measurement at the first station above or below the threshold value is indicative of a leak in the subsystem.

대표청구항 ▼

1. A computer implemented method for detecting leaks in a pipeline network, the method comprising: identifying a subsystem in the pipeline network, the subsystem comprising a plurality of stations that are topologically connected, the plurality of stations comprising a first station;accessing histor

1. A computer implemented method for detecting leaks in a pipeline network, the method comprising: identifying a subsystem in the pipeline network, the subsystem comprising a plurality of stations that are topologically connected, the plurality of stations comprising a first station;accessing historical temporal sensor measurements of the stations in the subsystem, the historical temporal sensor measurements comprising pressure measurements captured at corresponding timestamps over a predetermined time-span;selecting, from the historical temporal sensor measurements, a first subset of pressure measurements corresponding to timestamps with leak events;generating a prediction model for the first station that predicts a pressure measurement at the first station based on the historical temporal sensor measurements at each station in the subsystem;predicting, according to the prediction model, a series of pressure measurements at the first station based on the historical temporal sensor measurements at each station in the subsystem, wherein the series of pressure measurements is for a series of timestamps in the historical temporal sensor measurements;selecting, from the series of pressure measurements, a second subset of pressure measurements corresponding to the timestamps without leak events;determining a series of deviations between the first subset of pressure measurements and the second subset of pressure measurements of the first station; andidentifying a threshold value for detecting a leak in the subsystem based on the series of deviations, wherein a pressure measurement at the first station being above or below a predicted pressure measurement by the threshold value is indicative of a leak in the subsystem. 2. The computer implemented method of claim 1, wherein the historical temporal sensor measurements are accessed from a data repository of a SCADA system monitoring the pipeline network. 3. The computer implemented method of claim 1, wherein the prediction model predicts the pressure measurement at the first station based on a number of stations in the subsystem, an upstream flow-rate in the subsystem towards the first station, a downstream flow-rate in the subsystem away from the first station, a pressure measurement at an upstream station in the subsystem, and a pressure measurement at a downstream station in the subsystem. 4. The computer implemented method of claim 3, wherein the prediction model predicts the pressure measurement at the first station based further on an air-temperature measurement at the first station, a number of rotations at an upstream compressor in the subsystem, and a number of rotations at a downstream compressor in the subsystem. 5. The computer implemented method of claim 1 further comprising, synchronizing the temporal sensor measurements at the first station and the rest of the stations, the synchronization comprising: identifying a temporal lag between temporal sensor measurements of the first station and temporal sensor measurements of a second station; andcompensating for the temporal lag in the temporal sensor measurements of the second station for use in generating the prediction model. 6. The computer implemented method of claim 5, wherein the temporal lag between the temporal sensor measurements at the first station and the second station is determined by computing maximum correlation value between the temporal sensor measurements at the first station and the second station. 7. The computer implemented method of claim 1, wherein the threshold value is a multiple of standard deviation of the series of deviations. 8. The computer implemented method of claim 1, further comprising, selecting, from the series of deviations, a subset of deviations corresponding to timestamps at which a leak in the pipeline network was reported, and using the selected subset to determine the threshold. 9. A system for detecting leaks in a pipeline network, the system comprising: a memory; anda processor configured to: identify a subsystem in the pipeline network, the subsystem comprising a plurality of stations that are topologically connected, the plurality of stations comprising a first station;access historical temporal sensor measurements of the stations in the subsystem, the historical temporal sensor measurements comprising pressure measurements captured at corresponding timestamps over a predetermined time-span;select, from the historical temporal sensor measurements, a first subset of pressure measurements corresponding to timestamps with leak events;generate a prediction model for the first station that predicts a pressure measurement at the first station based on the historical temporal sensor measurements at each station in the subsystem;predict, according to the prediction model, a series of pressure measurements at the first station based on the historical temporal sensor measurements at each station in the subsystem, wherein the series of pressure measurements is for a series of timestamps in the historical temporal sensor measurements;select, from the series of pressure measurements, a second subset of pressure measurements corresponding to the timestamps without leak events;determine a series of deviations between the first subset of pressure measurements and second subset of pressure measurements of the first station; andidentify a threshold value to detect a leak in the subsystem based on the series of deviations, wherein a pressure measurement at the first station above or below a predicted pressure measurement by the threshold value is indicative of a leak in the subsystem. 10. The system of claim 9, wherein the threshold value is a multiple of standard deviation of the series of deviations. 11. The system of claim 9, wherein the prediction model predicts the pressure measurement at the first station based on a number of stations in the subsystem, an upstream flow-rate in the subsystem towards the first station, a downstream flow-rate in the subsystem away from the first station, a pressure measurement at an upstream station in the subsystem, and a pressure measurement at a downstream station in the subsystem. 12. The system of claim 11, wherein the prediction model predicts the pressure measurement at the first station based further on an air-temperature measurement at the first station, a number of rotations at an upstream compressor in the subsystem, and a number of rotations at a downstream compressor in the subsystem. 13. The system of claim 9, wherein the processor is further configured to synchronize the historical temporal sensor measurements at the first station and at the rest of the stations, the synchronization comprising: identification of a temporal lag between temporal sensor measurements of the first station and temporal sensor measurements of a second station; andcompensation for the temporal lag in the temporal sensor measurements of the second station for use in generating the prediction model. 14. The system of claim 13, wherein the temporal lag between the temporal sensor measurements at the first station and the second station is determined by computing maximum correlation value between the temporal sensor measurements at the first station and the second station. 15. A computer program product for detecting leaks in a pipeline network, the computer program product comprising a computer readable storage media, and the computer readable storage media comprising instructions to: identify a subsystem in the pipeline network, the subsystem comprising a plurality of stations that are topologically connected, the plurality of stations comprising a first station;access historical temporal sensor measurements of the stations in the subsystem, the historical temporal sensor measurements comprising pressure measurements captured at corresponding timestamps over a predetermined time-span;select, from the historical temporal sensor measurements, a first subset of pressure measurements corresponding to timestamps with leak events;generate a prediction model for the first station that predicts a pressure measurement at the first station based on the historical temporal sensor measurements at each station in the subsystem;predict, according to the prediction model, a series of pressure measurements at the first station based on the historical temporal sensor measurements at each station in the subsystem, wherein the series of pressure measurements is for a series of timestamps in the historical temporal sensor measurements; select, from the series of pressure measurements, a second subset of pressure measurements corresponding to the timestamps without leak events;determine a series of deviations between the first subset of pressure measurements and second subset of pressure measurements of the first station;andidentify a threshold value to detect a leak in the subsystem based on the series of deviations, wherein a pressure measurement at the first station above or below a predicted pressure measurement by the threshold value is indicative of a leak in the subsystem. 16. The computer program product of claim 15, wherein the threshold value is a multiple of the standard deviation of the series of deviations. 17. The computer program product of claim 15, wherein the prediction model predicts the pressure measurement at the first station based on a number of stations in the subsystem, an upstream flow-rate in the subsystem towards the first station, a downstream flow-rate in the subsystem away from the first station, a pressure measurement at an upstream station in the subsystem, and a pressure measurement at a downstream station in the subsystem. 18. The computer program product of claim 17, wherein the prediction model predicts the pressure measurement at the first station based further on an air-temperature measurement at the first station, a number of rotations at an upstream compressor in the subsystem, a number of rotations at a downstream compressor in the subsystem, and compressor energy usage. 19. The computer program product of claim 15, wherein the computer readable storage media further comprises instructions to synchronize the historical temporal sensor measurements at the first station and at the rest of the stations, the synchronization comprising: identification of a temporal lag between temporal sensor measurements of the first station and temporal sensor measurements of a second station; andcompensation for the temporal lag in the temporal sensor measurements of the second station for use in generating the prediction model. 20. The computer program product of claim 19, wherein the temporal lag between the temporal sensor measurements at the first station and the second station is determined by computing maximum correlation value between the temporal sensor measurements at the first station and the second station.