Real-Time Distribution State Estimation with Massive µPMU Streaming Data

This paper presents the development of a real-time Deep Neural Network-based non-linear State Estimator (DNN-SE) for distribution networks. The DNN-SE is designed to run on the `Analysis on the Wire' framework, a scalable in-network general-purpose computing architecture capable of processing the high volume data streams originating from micro- Phasor Measurement Units (μPMUs). Simulations are performed on a 200 bus modified distribution network with high penetration of renewable energy resources. Over 4.45 billion data points are obtained from multiple μPMUs each month to train the network. The DNN-SE model is able to accurately estimate bus voltages and angles in a streaming fashion under high variability of wind, solar and demand data.