I-Corps: Data-Driven Robust Optimization Technology for Battery Storage System Management

点击翻译按钮获取中文摘要

The broader impact/commercial potential of this I-Corps project is the development of a data-driven, distributionally robust optimization (DRO) methodology to address real-life decision problems under uncertainty arising in energy systems with renewable integration. Specifically, the proposed technology implements the DRO model to optimize battery storage unit operations in residential areas equipped with rooftop solar photovoltaic (PV) systems. The model is designed to minimize the long-run electricity costs under uncertain electricity usage, photovoltaic engery generation, and electricity prices. The technology will be integrated into an intelligent home system that automatically controls the optimal battery storage unit operation and electricity purchase decisions, and will be implemented in the embedded controller unit to enable real-time decisions. The aim of the project is to establish the feasibility and verify the real-world performance of the proposed DRO model, particularly in the field of energy systems. This I-Corps project is based on the development of new models and algorithms for residential photovoltaic (PV)-battery system operations using the distributionally robust optimization (DRO) paradigm. The proposed scheme automatically controls the operation of the battery storage unit to optimally determine when to store the excess amount of PV generation or to discharge the stored amount to satisfy the household energy demand. Prior systems unrealistically assumed a known probabilistic description for the uncertain electricity prices, PV generation, and energy consumption. In most real-life situations, this description is never available. The decision-makers only have access to historical data that may be used to infer the underlying probabilistic description. The DRO scheme addresses this fundamental shortcoming by first constructing a set of plausible distributions consistent with the available information and then optimizes for battery storage operations that perform best for all distributions in the set, safely anticipating potentially adverse outcomes. The proposed scheme may mitigate overfitting to the data and yield high-quality battery operations in out-of-sample circumstances.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.