IEEE P3900

New IEEE Standard - Active - Draft.
Focused on federated machine learning for power user electric energy data acquisition and analysis, this document defines a standardized architectural framework and operational guidelines. It provides a blueprint for collaborative model building across heterogeneous grid devices while supporting privacy, security, and regulatory considerations. Included are role mappings, data preprocessing standards, training workflows, and security mechanisms tailored for power metering scenarios.

This guide provides guidance on the application of federated learning and neural network-based models for electric energy consumption data acquisition and analysis. It describes the framework, processes, and methods for utilizing electric energy data in a secure and distributed artificial intelligence (AI) environment. The guide outlines mechanisms for data preprocessing, feature representation learning, model training with neural networks, parameter aggregation strategies, and security enhancement in federated learning scenarios.
The purpose of this guide is to provide a unified reference for the application of federated learning and neural network-based AI methods for electric energy consumption data acquisition and analysis. This guide aims to enable secure, efficient, and interoperable deployment of AI-driven federated learning across heterogeneous metering infrastructures. It intends to support power utilities, device manufacturers, and AI solution providers in improving the value of metering data through distributed neural network training and feature representation learning, thereby enhancing the accuracy of load forecasting, the reliability of anomaly detection, and the effectiveness of demand response and energy management.