Large US Electric Utility
A large US electric utility needed to replace its legacy demand forecasting system with a modern, causal ML approach that could handle AMI data at scale, accurately quantify demand response program im…
MAPE Reduction
Customers Covered
Hours Saved / Month
Regulatory Filings Supported
The Challenge
A large US electric utility needed to replace its legacy demand forecasting system with a modern, causal ML approach that could handle AMI data at scale, accurately quantify demand response program impacts, and produce audit-grade confidence intervals for regulatory filings.
Our Approach
EcoMetricx deployed EnerGaze Forecast with a custom causal inference specification that separated baseline demand from program-induced conservation. The team built AWS-native ETL pipelines to process 15-minute AMI data for 1.2 million customers, validated the model against 3 years of historical actuals, and configured automated retraining on a monthly cadence.
Results
The new forecasting system reduced MAPE by 34% compared to the legacy model and produced confidence intervals that satisfied the state PUC's requirements on the first submission. Fully automated model retraining reduced the analytics team's monthly workload by an estimated 120 hours.
Technologies Used
Ready to get results like these?
Talk to our team about your data challenges. We'll scope a solution that delivers measurable outcomes.
Start a Conversation