Machine Learning Applications in Power Distribution System Operation

This panel focuses on machine learning applications in power distribution grid operation. The panelists are from utilities, national labs, and universities and will present methods and results developed in a number of projects sponsored by utilities and the Department of Energy in the past few years. The panel begins by an introduction of the machine learning needs in advanced data analytics for assisting utilities, especially small utilities that do not have sufficient personnel or expertise to conduct massive data analysis. Next, we will present methods using smart meters and SCADA data as inputs to identify the baseline of demand response programs (e.g. conservation voltage reduction and HVAC load reduction programs) and to identify the behind-the-meter distributed energy resources. We will also introduce the growing needs from the industry on applying advanced data analytics for modeling and control. The presentations will cover the generation of synthetic network topology and load profiles and the use of reinforcement learning to develop adaptive, distributed control algorithms in order to facilitate grid restoration in abnormal operation conditions. Presentations in this panel session: - Machine learning based synthetic data generation (23PESGM3989) - Smart meter data analysis for distribution transformer monitoring and sizing (23PESGM3990) - Machine learning powered residential load profiles analysis and DER capacity forecasting (23PESGM3991) - Optimal Coordination of Distributed Energy Resources Using Deep Deterministic Policy Gradient (23PESGM3992) - Optimal Design of Volt/VAR Control Rules using Deep Learning (23PESGM3993) - Learning to Operate an Electric Vehicle Charging Station Considering Vehicle-grid Integration (23PESGM3994)