Abstract for: Balancing-as-a-Service : New Methods for Managing Instability on the Grid

Electricity grids are among the largest man-made systems to require real-time stabilizing feedback loops. These are provided through an ecosystem of independent actors coordinated by the invisible hands of a multitude of markets. Balancing Service Providers operate at the heart of this system. Here, we present new System Dynamics methods developed at one such company. Our work spans several model classes, ranging from the conceptual to the educational, with the main thrust in economic models used to feed business cases for industrial customers. We gradually developed new methods to push performance whilst keeping a lid on model complexity, with the ultimate goal of simulating multiple assets on multiple markets simultaneously. The framework of SD proved helpful for describing key elements of the balancing service business. Simple, high-level models effectively reproduce observed grid behavior, providing meaningful insights. The methodology is also well-suited for modeling industrial sites, but representing market interactions with high fidelity has proved more challenging. These challenges, in turn, provided fertile ground for developing new algorithms, including the discovery of a surprising alternative to Linear Programming for optimizing inventory. Adopting SD for economic modeling is an exception in our sector, dominated by acausal models. However, SD fits perfectly with the physical modeling undertaken by grid operators. There is bound to be more value to be discovered in this area. Our algorithms for constrained stock/flow optimization contribute to the body of molecules and methods at the intersection of System Dynamics and Operations Research. They might find applications beyond grid-scale battery optimization.