In the operation of power-intensive Air Separation Units (ASUs) that produce storable liquid products, optimization opportunities exist at two time scales - week-ahead production scheduling to leverage fluctuations in electricity prices, and real-time decisions that optimize the entire plant operation and capture spot opportunities. In our previous work, we proposed a methodology based on flexibility analysis and robust optimization to ensure feasibility of real-time operational decisions at ASUs for future time periods within a scheduling horizon. In this paper, we build upon the methodology to propose approaches to improve the online performance of a production schedule for ASUs by using the real-time optimization layer. We compare several policies for real-time optimization and our studies on real plant data show interesting trade-offs between week-ahead scheduling and real-time optimization. (C) 2017 Elsevier Ltd. All rights reserved.