Integrated Energy Monitoring and Control IoT System and Validation Results from Neural Network Control Demonstration

Abstract

Increasing use of renewable and distributed power generation creates opportunities for customer resources to support power system operations by adjusting power consumption and generation to address grid needs, based on system-wide and local grid conditions. We present an integrated Energy Internet of Things (E-IoT) testbed including (1) distributed Advanced Realtime Grid Energy Monitor Systems (ARGEMS) with sensing, communication, and control capabilities, (2) distributed smart home sites, including smart home hubs for monitoring and control of physical and simulated Internet of Things (IoT) distributed energy resources (DERs) such as solar systems, home batteries, and smart appliances, and (3) control algorithms based on artificial intelligence and optimization, which manage customer DERs to respond to power grid conditions while serving customer needs. The integration of these three components enables demonstration and assessment of a variety of advanced DER monitoring and control strategies for improved power grid operations and customer benefits. We validate the functionality of this E- IoT testbed by demonstrating control of a simulated home battery by a neural network imitation learning algorithm running on a physical smart home hub, where the controller responds to grid services events triggered by an ARGEMS device based on local power system measurements and simulated bulk power system conditions. The developed neural network controller imitates the performance of a model predictive control optimization algorithm, but requires nearly 20,000 times less computational time and can run on small distributed computers.