Smart Meter Outage Prediction
3AI August 12, 2020
Client: Leading Utility Organization
Problem Statement
Organization wanted to gauge value of its new Smart-Grid initiative for their infrastructure management:
-
- Predicting meter events and outages including likely malfunction
- Predicting spike in consumption to reduce spot-procurement
- Ability to notify customers of their consumption habits / patterns in near-real time.
Analysis Led Approach
- Define population by zip codes, time period, manufacturer & events (power and meter events)
- A pair wise correlation of events and Linear Regression model to predict total event count with variables as events identified in previous events
- Prioritization Algorithms arrived by using cluster analysis for prediction of outages & spikes
Business Approach
- Prediction of last mile consumption spikes with accuracy of 89%
- Increase up to 80% in prediction of consumption leakage
- 10 times decrease costs decrease in network audit
- Reduction in field staff workload that repairs/replaced smart meters at 0.014% from 1% of 1.2 million meters in a 45 day period
Critical Success Factors
- Consumption spikes accurately predicted with an efficiency of 89%
- Consumption leakage controlled with a 80% positive prediction for similar events
- Results seen in 45 days post implementation of the suggestions
