Multi-objective genetic algorithm for sustainable optimalization

Abstract

The long term, large scale, hybrid, multidisciplinary models of Computational Sustainability requires new optimization methodologies. In achieving optimal process design and control we have to choose the „best” from various structures and parameters. The usual objectives are minimal cost or maximal profit. One of the accepted approaches is to find the exact optimal solution for a simplified model formulated by sophisticated mathematical constructs like MINLP. Another approach is based on the qualitative knowledge of engineers and described by heuristic rules and rule-based decision algorithms. Optimization for sustainable development cannot often controlled by a single, aggregated objective. We have to consider multiple objectives according to short, middle, or long time horizons. Besides the economic goal function, we have to consider the environmental impacts (e.g. the necessary recycling, etc.). This needs detailed model-based multi-objective process development. Accordingly, in our work we use an engineering approach that focuses on the search for „good enough” solutions, based on the most detailed models. In the solution of practical problems, priority ranking of the constraints and evaluations combined with a new grid method helps to focus on the very part of the Pareto-front where the good solutions are found. The elaborated, multi-objective genetic algorithm supports effective coding and the multi-criteria evaluation of sustainable processes. Keywords: meta-heuristic, multi-criteria, development, modeling, optimization