Differential Evolution for the Offline and Online Optimization of Fed-Batch Fermentation Processes

Abstract

Summary. The optimization of input variables (typically feeding trajectories over time) in fed-batch fermentations has gained special attention, given the economic impact and the complexity of the problem. Evolutionary Computation (EC) has been a source of algorithms that have shown good performance in this task. In this chapter, Differential Evolution (DE) is proposed to tackle this problem and quite promising results are shown. DE is tested in several real world case studies and compared with other EC algorihtms, such as Evolutionary Algorithms and Particle Swarms. Furthermore, DE is also proposed as an alternative to perform online optimization, where the input variables are adjusted while the real fermentation process is ongoing. In this case, a changing landscape is optimized, therefore making the task of the algorithms more difficult. However, that fact does not impair the performance of the DE and confirms its good behaviour.

The use of differential evolution algorithm for solving chemical engineering problems

“…However, when noisier settings were employed, the DE algorithm performance decreased. Using the same case studies and objectives (from Rocha et al 2007), Mendes et al (2008) performed online and offline optimizations of input variables by applying three distinct algorithms, namely DE, a real value EA, and a fully informed particle swarm (FIPS). The results obtained for the offline optimization indicated that DE/Rand/1/ bin has the best results, followed closely by DE/best/2/ bin and FIPS.…”

Real‐time optimization and control for polyhydroxybutyrate fed‐batch production at pilot plant scale

“…important biological variables, such as substrate/nutrients concentrations). 24 Zuo and Wu 25 implemented (via simulation) a method to perform a semireal time optimization for the fed-batch cultivation of Bacillus thuringiensis. A hybrid neural network was used for modelling the process, whereas the genetic algorithm method was used for performing the optimization.…”

“…A hybrid neural network was used for modelling the process, whereas the genetic algorithm method was used for performing the optimization. State variables were measured and the system was re-optimized every 1 h. Mendes et al 24 performed in silico studies on the implementation of on-line optimization to three case studies (fed-batch recombinant E-coli fermentation, fed-batch ethanol production, and a hybridoma reactor). These authors showed that metaheuristic optimization algorithms could solve the problem, even in the presence of simulated uncertainties.…”

“…An alternative to overcome this inconvenience (in order to implement on‐line optimization in a reliable way) is coupling RTO with measurements of relevant state variables ( i.e . important biological variables, such as substrate/nutrients concentrations) 24 . Zuo and Wu 25 implemented ( via simulation) a method to perform a semi‐real time optimization for the fed‐batch cultivation of Bacillus thuringiensis .…”

Evolutionary Hybrid Configuration Applied to a Polymerization Process Modelling

“…The problems tackled are both benchmarks [11][12][13][14][15][16][17] and real-life [18][19][20][21]. Some examples for chemical engineering applications include: oxidation processes [22][23][24], energy, fuels and petrol derivatives [25][26][27][28][29], fermentation [30][31][32].…”