OpenMeter: Privacy-Preserving Natural Language Access to Smart Meter Data
The paper introduces OpenMeter, a schema-agnostic multi-agent system that translates natural-language queries over industrial smart meter data into executable code and human-readable insights. It targets utility staff without SQL expertise while keeping sensitive consumer data on local hardware rather than cloud-based LLMs. Multi-scenario ablation studies report the largest accuracy gain from full multi-agent orchestration on resource-constrained local hardware, at +6.1 pp EX-R over the strongest single-agent configuration. At frontier scale, the components remain useful but should be enabled selectively.
It offers a concrete architecture and ablation evidence for deploying privacy-preserving natural-language analytics on local infrastructure in energy operations.
Comments & Questions (0)
No comments yet
Be the first to comment!