Data-driven robust optimization for crude oil blending under uncertainty

Optimization of crude oil blending helps improve the operating efficiency of refineries. However, widespread uncertainties, such as oil properties, bring difficulty in realizing this task. A data-driven robust optimization (DDRO) approach is proposed to optimize crude oil blending under uncertainty. First, the blending effect model is used to extract uncertainties of oil components from production data by recursive least squares method. Second, the uncertainty set is constructed by combining principle component analysis and robust kernel density estimation based on the historical data of blending effects. A novel data-driven robust model for recipe optimization of crude oil blending is developed by utilizing the obtained uncertainty set. The dual transformation is applied to derive the linear counterpart of the DDRO model. A case study is adopted to illustrate the effectiveness of the proposed method. (C) 2019 Elsevier Ltd. All rights reserved.