Multi-objective dynamic optimization study of fed-batch bio-reactor

Evolutionary algorithms are widely used for dynamic optimization problems of fed-batch bio-reactors for productivity-yield maximization by optimizing the substrate feed recipe. However, this is usually done for a fixed fed-batch time. Conventionally, the optimum fed-batch time is computed by solving several single objective dynamic optimization problems for different fed-batch time. Since this approach is computationally quite expensive, we propose a Multi-Objective Optimization (MOO) problem formulation to find the optimum fed-batch time for maximizing productivity and/or yield. Such an MOO approach is expected to save significant computational efforts. To demonstrate the proposed MOO implementations for dynamic optimization of fed-batch bio-reactors, secreted protein production is considered as a case study. Specifically, four distinct objectives, namely productivity, yield, fed-batch time, and endpoint substrate concentration are considered in this work. An evolutionary multi-objective differential evolution algorithm is used for solving the MOO problems.