文献类型：期刊文献

英文题名：Unsupervised Learning-based Robust Optimization under Disjunctive Uncertainties

作者：Du, Chenge[1]; Wang, Zhenlei[1,2]

机构：[1] East China University of Science and Technology, Key Laboratory of Smart Manufacturing in Energy Chemical Process, Ministry of Education, Shanghai, 200237, China; [2] East China University of Science and Technology, Ministry of Education, Engineering Research Center of Process System Engineering, China

年份：2023

外文期刊名：2023 5th International Conference on Industrial Artificial Intelligence, IAI 2023

收录：EI(收录号：20235115240147)

语种：英文

外文关键词：Numerical methods - Optimization - Uncertainty analysis - Unsupervised learning

摘要：In practical industrial processes, uncertain parameters often affect the feasibility and optimality of production decisions. Robust optimization is a classical method for optimization under uncertainty and focuses on the construction of uncertainty sets. However, it is difficult to capture the uncertainty space with disjunctive structures accurately using a single uncertainty set, which will lead to conservative decisions. Therefore, considering the probability distribution information of data, this paper proposes an unsupervised learning-based data-driven robust optimization method under disjunctive uncertainties. Gaussian mixture model (GMM) is applied to extract the underlying Gaussian distributions of uncertainties. Then, support vector clustering (SVC) is employed to construct a union of data-driven uncertainty sets, which can depict disjunctive uncertainties more accurately. The tractable robust counterpart optimization problem is developed on the basis of derived uncertainty sets. A numerical example and an industrial case are presented to demonstrate the effectiveness of the proposed method and the influence of the regularization parameter on the optimization results is discussed. ? 2023 IEEE.