Physics-aware distributed intelligence in DER-rich grids
The energy transition is largely led by the installment of GW-scale wind and solar plants to decarbonize our electricity generation. Closer to home (or work) the energy transition is visible from the increasing adoption of rooftop solar, home batteries, electric vehicles, and smart devices. While small compared to solar farms or grid-scale storage, these devices termed distributed energy resources (DERs) are transforming the grids edge. The distribution grid of the past was a connected network of loads being served by a power substation -- on a lab bench, think of a simple circuit with a row of LED lightbulbs being supplied by a single pack of AA batteries. Modern distribution grids that have DERs changes this picture considerably: distributed generation introduces new sources of generation in a grid not designed for power injections and bi-directional power flow. While challenges with grid integration persist, these DERs also uncover new capabilities to realize the smart grid concept by providing distributed sensing, computing, and control throughout the distribution system. DERs are typically equipped with smart inverters, local intelligence, and computational abilities. Many are also integrated into home energy management systems which can coordinate the actions of multiple devices within a single residence or building unit. The widespread adoption of DERs then provides a unique opportunity for grid operators to enhance service quality, improve grid reliability, and reduce operating costs. A key challenge to using DERs towards grid services is that of coordination: the devices are spatially distributed throughout the grid, have different temporal characteristics (ex. intermittent solar generation, EV usage), and are privately owned by different agents.
To address these challenges, we look towards a physics-aware distributed intelligence framework. To make grid services physics-aware and location dependent, we model the underlying distribution grid and device characteristics. The real-time coordination across a large number of these spatially distribution agents is made tractable by taking a distributed approach. Herein, each agent has access to local information and a limited amount of information shared by its neighbours, which it uses to optimize its actions towards meeting local and system-level goals. In the distributed paradigm, DERs can be coordinated to meet system-level goals without the aid of a central authority. Further, the distributed approach respects ownership boundaries by retaining decision making autonomy to the device owner and preserving data privacy.