Situation
Meter behavior can often be a black hole. Our client, a large WaterUtility, was dealing with many meter related issues:
- couldn’t tie asset issues to revenue impact
- couldn’t find improperly installed meters or meters at customer types with typically high water usage.
- estimated 60-65% of meter issues were false positives.
As a result, resources were being wasted with unnecessary service calls.
Solution
We deployed a meter operations app on the App Orchid platform to improve the accuracy of detecting faulty meters and to help prioritize which service calls were the most urgent in terms of both risk and revenue loss.
AI/ML Enablers
- “Listened” to meter data on a daily basis, integrating all data coming from meters – such as meter alarms
- Natural Language processing - Integrated employee “tribal knowledge” such as knowledge of specific location information that effects meter readings. Processed data to find patterns in service notes and surface this as knowledge.
- Knowledge Graphs - Integrated knowledge of customer type and bills into meter health and issue identification such as “a carwash would use lots of water”
- Utilized deep learning to continuously improve knowledge about meter issues and meter health
Outcomes
Meter Operations personnel immediately started to address the highest risk meters and document the accuracy of insights via the easy-to-use interface
- Detected faulty and problematic meters and assets
- Measure meter performance, meter life and probable outcomesDetect what causes meter failure
- Proactively repair or replace meters and prevent revenue leaks.
