(746b) Enabling Early Planning and Development of Eco Industrial Parks for Photovoltaic Circular Economy Using Multi Objective Optimization Techniques

Eco-Industrial Parks (EIPs) are widely studied as a potential solution to waste reutilization as it allows industries to work in synergy for reutilization of waste, thus avoiding landfills and other potential environmental impact. Our previous work has demonstrated the environmental benefits as a result of implementing Life Cycle Symbiosis (LCS), an extension of Industrial Symbiosis (IS) (Mathur et al., 2020). In order to implement such an LCS network, it is necessary to provide additional guiding principles for decision making, which we aim to address through two objectives in this work. The first is the expansion of an already developed hypothetical solar photovoltaic (PV) centric LCS network in a given geographic region. The second is the identification of tools that will assist in the design of such an LCS driven EIPs. Decision-making in the early stages of EIP development is challenging due to conflicting objectives of cost and environmental impact. Hence, the problem can be formulated as a multi-objective optimization (MOO) problem. In this work, we explore the development of an EIP as a combinatorial problem and formulate it as a constrained MOO problem. Given that there exist several different optimization techniques, identification of right method for solving EIP problem at an early stage is an open problem. We address this problem by comparative evaluation of a traditional scalarization optimization (the weighted sum (WS)) method with a heuristic (Genetic Algorithm (GA)) optimization technique. Further, a hybrid optimization technique (using GAs and the ε-constraint method in combination) was used to leverage the strengths of both the scalarization and heuristic optimization methods. It was found that in order to run a MOO problem using the WS method means effectively reducing the problem to a single objective optimization problem based on user preferences and results in the generation of one solution. In contrast, it was found that GAs effectively explore and exploit a given search space, thus resulting in several near-optimal solutions as opposed to a single optimal solution. One challenge in using GAs is the generation of multiple sub-optimal solutions, thus requiring additional decision-making criteria to identify the best solution for the implementation of EIP. Finally, the hybrid optimization method successfully explored the search area and based on subsequent higher-level decision-making to select a single objective would enable the identification of one global optimal solution. We present the results on comparative analysis of these methods for implementing EIP in Arizona, USA, based on emerging PV waste. To conclude, this study looks at enriching our understanding of the suitability or unsuitability of a given MOO method in the context of early design and evaluation of EIPs as opposed to simply comparing the MOOs from a mechanistic point of view for a generic MOO problem.