(487c) Data-Driven Optimization for Sustainable Energy Systems Conference: AIChE Annual MeetingYear: 2019Proceeding: 2019 AIChE Annual MeetingGroup: Sustainable Engineering ForumSession: Sustainable Engineering Forum Plenary (Invited Talks) Time: Wednesday, November 13, 2019 - 9:30am-10:00am Authors: Ning, C., Cornell University You, F., Cornell University In recent years, sustainable energy systems have gained tremendous attention from both academia and industry. With growing concerns over climate change and a global energy crisis, the utilization of renewable energy sources is growing rapidly around the globe. Because of their economic incentives and sustainability benefits to our society, it is critical to design and operate sustainable energy systems to meet future energy demands. In these sustainable energy systems, a large amount of data is usually collected and archived, and valuable information embedded within this data can be potentially extracted using machine learning techniques to further support decision making in sustainable energy systems. Data-driven optimization is an emerging optimization paradigm that leverages the organic integration of machine learning and mathematical programming. Therefore, data-driven optimization provides a promising and systematic tool for sustainable energy system design and operations. In the first part of this presentation, we develop a novel data-driven adaptive robust unit commitment (UC) model for integrating wind power into smart grids. By leveraging a Dirichlet process mixture model, a data-driven uncertainty set for wind power forecast errors is constructed as a union of several basic uncertainty sets. Accordingly, the proposed uncertainty set can flexibly capture a compact region of uncertainty in a nonparametric fashion. Based on this uncertainty set and wind power forecasts, a data-driven adaptive robust UC problem is then formulated as a four-level optimization problem. A decomposition-based algorithm is further developed. Compared to conventional robust UC models, the proposed data-driven approach does not presume single mode, symmetry or independence in uncertainty. Moreover, it not only substantially withstands wind power forecast errors, but also significantly mitigates the conservatism issue by reducing operational costs. The effectiveness of the proposed approach is demonstrated with the six-bus and IEEE 118-bus systems. In the second part, we present a data-driven adaptive robust mixed-integer nonlinear fractional programming model for biomass and agricultural waste-to-energy network design. Data-driven uncertainty sets are used to determine the region of feedstock prices and product demands. Latent uncertainties are identified using principal component analysis, and kernel density estimation is employed to accurately extract probability distributions of the projected uncertainty data. Additionally, the return on investment objective function leads to an adaptive robust mixed-integer nonlinear fractional programming model. We apply the data-driven approach to a large-scale case study of a biomass and agricultural waste-to-energy network having 216 technologies and 172 chemicals.