Research

Background

Our research is concerned with mathematical models and computational tools for future power systems. We define power systems as the combination of physical infrastructure (wires, generators, transformers, solar panels, etc.) with institutions (markets, regulatory frameworks, business models, etc.). These systems are overwhelmingly complex (consider that each outlet of each US home is technically connected through a continuous electrical connection) and the backbone of our economy (what doesn’t depend on electricity?). Every aspect of managing and planning these systems therefore requires a set of reliable decision-support tools. Unfortunately, the properties of current and future power systems are changing and often do not match the underlying assumptions and models of some of the tools that are in use today. Our research addresses this mismatch.

What are the properties of future power systems? When power systems were designed in the past century, their purpose was to burn fuel most efficiently (meaning in large plants) and then transporting the generated electricity to commercial and household loads. This centrally controlled and one-directional paradigm is being replaced by the following four power system properties:

☀️ Renewable: Today wind and solar are economically competitive forms of electricity generation and they align with targets to reduce carbon emissions. However, wind and solar works very differently than fuel-burning generators because their generation is virtually free but depends on uncertain and variable weather-patterns.
🌍 Decarbonized: Economy-wide decarbonization does not only affect the generation side through the deployment of wind and solar, but also the demand side though the electrification of transportation, buildings, and industry processes. This “electrify everything” approach to decarbonization changes load patterns and alters the economics of electricity through its entire supply chain.
📡 Digitalized: New data collection and processing technologies enables new business models and better real-time decision making. But it raises questions on data privacy and cybersecurity.
🧑‍🤝‍🧑 Democratized: Electricity consumption and production will go hand-in-hand. The majority of residential electricity consumers will have roof-top solar panels and batteries. Industrial consumers will be able to make an additional revenue from providing flexibility to the grid.

What are the resulting high-level research questions?

How can power systems be safely operated in the presence of stochastic generation and demand?
How will “electrify everything” shape the future power system and how can we ensure reliable and equitable access?
How can costs and benefits be allocated transparently and fairly to all agents in the system?
What are good models and abstractions for the complex cyber-physical PES of the future?

Ropes Research Overview

Methods, Theory, and Tools

We use methods from electrical engineering (to understand and model physical power flow, grid stability, and aspects of control engineering) and operations research (optimization, machine learning, data science, and economics). Our lab explores the development of new tools and relevant quantitative models for power systems.

Learning + Optimization

Machine learning complements traditional optimization. Classic optimization is often too slow or requires too much approximation for practical purposes. Machine learning models can create computationally efficient model-free surrogates of complex physical behavior that bridge this gap.