A superstructure optimization approach for optimal refinery water network systems synthesis with membrane-based regenerators

Abstract Water is a key element in the operation of petroleum refineries. As such, there are great interests to incorporate water reuse, regeneration (treatment), and recycle (W3R) approaches in the design of refinery water network systems with the aim of minimizing freshwater consumption and wastewater generation. Hence, this work concerns the optimization of refinery water network systems synthesis comprising water-producing streams (sources), water-using units (sinks or demands), and water-treatment technologies (regenerators). We develop a source-interceptor-sink superstructure representation that embeds as many feasible alternatives as possible for implementing W3R while preserving attractive convexity property and being amenable to tighter model formulation. A mixed-integer nonlinear program (MINLP) is formulated based on the superstructure to determine the optimal retrofit of a water network structure in terms of the continuous variables of total stream flowrates and contaminant concentrations, and the 0-1 variables of stream piping connections. The superstructure and the MINLP explicitly models partitioning regenerators particularly the membrane-based treatment technologies of ultrafiltration and reverse osmosis, with the objective of minimizing the fixed capital costs of installing piping connections and the variable cost of operating all stream connections while reducing the pollutants level to within regulatory limits. The proposed modeling approach is implemented on an industrial case study using the GAMS/BARON platform to obtain a globally optimal water network topology.