Events

Speaker: Bolun Xu
Faculty Host: Prof. Matthias Preindl, Prof. Dan Steingart

Abstract: Future energy systems must maintain instantaneous power balance against large penetration of intermittent wind and solar generations, posing a series of new challenges and opportunities on system and market design. Energy storage is an ideal candidate for providing carbon-free operation flexibility that compensates renewable fluctuations, but how to control storage devices optimally in a computation tractable way remains an open question. The optimal scheme must address unique operation characteristics of storage devices such as state-of-charge constraints and electrochemical degradation, while the computation must be sufficiently fast in order to fully utilize the fast response speed of storage devices.

This talk follows a bottom-up design process and introduces two novel optimal control designs for incorporating storage into the power system: a non-anticipatory control policy for mitigating instantaneous system errors, and a predictive control policy for scheduling storage dispatch. Both policies make instant optimal decisions subject to storage operation constraints and the electrochemical degradation mechanism, and their performances are compared with state-of-the-art approaches using real system data. The non-anticipatory policy achieves near-zero regret for maximizing participants’ market profit in frequency regulation without needing to forecast future frequency realizations. The predictive control policy connects the optimal control with the Lagrangian multiplier associated with the state-of-charge constraint, and uses a binary search algorithm for finding the Lagrangian multiplier value according to the Karush-Kuhn-Tucker (KKT) condition, achieving constant space complexity and worst-case log-linear time complexity, and demonstrated up-to 100,000 time speed-up compared to the state-of-the-art solver Gurobi.

Based on the proposed control designs, this talk demonstrates two case studies of storage participation in current electricity markets including market revenue and battery lifetime expectancy. In the end, this talk summarizes a roadmap towards a reliable and economic decarbonization in the power sector with future design proposals at the market, system, and device level.

Bio: Bolun Xu is a postdoctoral associate at MIT, his position is jointly affiliated with MIT Energy Initiative and Lab for Information and Decision Systems. He received his PhD from University of Washington in 2018, MS from ETH Zurich in 2014, and BS from Shanghai Jiaotong University in 2011, all in Electrical Engineering. He was awarded the 2018 Scientific Achievement Award by the Clean Energy Institute in University of Washington.