The Polish national gas distribution network provider Polska Spółka Gazownictwa sp. z o.o. PSG for short own and maintain some 185,000 km’s of underground gas distribution pipelines, of which approx. 8000 km are medium elevated and high pressure pipelines. PSG have a legal requirement to monitor these high medium elevated and high pressure pipelines on an annual basis for issues with vegetation and building encroachment.
Like many gas network providers PSG conduct annual visual monitoring via the use of helicopters who fly the length of the high pressure network, but is there a more cost effective, environmentally friendly solution that can pick up on more issues than the human eye?
Spottitt proposed that:
the cost and imagery capture capacity of sub meter resolution satellite imagery rivals that of aerial visual inspection
the use of cloud computing and machine learning means that trained algorithms could be used to search for, and highlight issues pixel by pixel faster and more accurately than the human eye
In June Spottitt conducted a proof of concept exercise on 80 km of PSG high pressure gas distribution pipelines using 50 cm resolution Pleiades imagery from Airbus and algorithms trained to detect:
Vegetation encroachment (trees and tree roots too close to the pipeline)
Potential gas leaks (areas of lower than expected vegetation health close to the pipeline)
So what were the results like?
Even with no-go buffer zones of just a few meters either side of the pipeline, the algorithms developed by Spottitt’s Earth Observation specialist Asterios Papastergios accurately classified and flagged trees, buildings and potential gas leaks along the entire 80km of gas distribution pipeline monitored.
Spottitt’s estimated service provision cost for annual monitoring approx 25% cheaper than traditional aerial visual inspection.
Anticipated increased detection rates of vegetation, building and potential gas leak issue detection via automated analysis of each image pixel versus traditional aerial visual inspection.