Improved management of water resources in process industry by accounting for fluctuations of water content in feed streams and products

• Published in: Journal of water process engineering Vol. 39; p. 101870
• Main Authors: Ferrucci, Aurelio, Vocciante, Marco
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.02.2021
• Subjects: Circular economy; Data reconciliation; Food processing; Mathematical optimization; Wastewater reduction; Water reuse; NP-hard; SMINLP; LP; DAE; MINLP; WPA; Food processing; BARON; NLP; Circular economy; BAT; Water reuse; COUENNE; MEN; Data reconciliation; SNLP; Mathematical optimization; Wastewater reduction
• ISSN: 2214-7144; 2214-7144
• DOI: 10.1016/j.jwpe.2020.101870

•Fluctuating waters are of primary importance for several water-consuming industries.•Common strategies not consider fluctuating water, limiting the overall recovery.•Fluctuating waters could prohibitively increase complexity and computational burden.•Statistical information can allow counting fluctuation in reconciliation procedures.•A proper strategy using ordinal optimization can overcome the resulting complexity. In this paper, a general strategy is presented for optimizing water resources in industrial activities taking into account fluctuations of water flow-streams and water contents in feed and products as well as in the utility system. Indeed, considering fluctuations can be of primary importance for several water intensive industries such as the food or the paper&pulp industry. Furthermore, whenever the water balance includes the water content of products (as is typically the case for the food industry) fluctuations in the composition of products can considerably affect the overall balance due to water being a major component of many foods. Therefore, the currently used optimization strategies that disregard water fluctuations, can lead to a severe bias in the water balance and thus overlook the potential of water recovery and reuse. Including this aspect in widely used algorithms for process monitoring (data reconciliation) and optimization might lead to a possibly prohibitive increase in the computational burden. In particular, it is shown in this paper that the resulting optimization algorithm should solve a dynamic non-convex stochastic Mixed Integer Nonlinear Programming problem. To tackle this (typically NP-hard) problem, a strategy has been developed that combines ordinal optimization for dealing with statistical information and nonlinear, non-convex deterministic algorithms. The outlined strategy has been applied to a complex process in the food industry (production of starch and starch-based products). The results seem to confirm the general validity of the algorithm developed.