Here are a few projects that I’d like to pursue someday:
Optimal Management of Grid-Scale Battery Energy Storage Systems: Battery Energy Storage Systems have exploded in popularity over the last few years. They’ve been described as a key technology for decarbonizing the grid, and as clean energy’s next trillion-dollar technology. All of this hinges on the ability of batteries to generate revenue, which consists of a kind of portfolio optimization problem in which battery operators must choose how much capacity to allocate to which use at what time. This problem is made more challenging by the fact that battery operations can affect market outcomes. Charging decisions can also affect the performance of the physical battery system over time. And there are often multiple batteries operating in the market at the same time, each myopically pursuing revenue-maximizing behavior. I’m interested in how batteries should make these operational decisions, and whether techniques from Safe Reinforcement Learning or Multi-Agent Reinforcement Learning could be productively applied to this problem.
Natural Hazards, Forecasting, and Risk: Natural disasters are becoming increasingly frequent and severe. For this reason, understanding and quantifying disaster risk is increasingly important to protecting populations. This is a challenging problem. Natural hazards models must contend with random variability and path-dependence in atmospheric and geologic conditions, as well as changes in human behavior and exposure to natural risks. Natural disasters are also rare events, which poses problems for statistical estimation. I’m interested in how ML, particularly physics-based machine learning, can give us better, more decision-relevant hazard models.