Experimental Exploration Of Operating Optimization Parameters Of Penicillin Fermentation Process

The fermentation process has the characteristics of multi-process,multi-variable and non-linearity.It is very challenging to control the process variables and estimate system parameters,especially the biological parameters in the fermentation process,which cannot be measured directly or indirectly.Usually,the optimal experiment design(OED)technology is used to estimate the biological parameters of the fermentation process.However,most of the existing optimal experimental design methods focus on maximizing experimental data information to improve the structural accuracy of the system model and the estimation of model parameters,but ignores the ability of the model to describe the actual system under specific production conditions.From the perspective of model identification,the goal of optimal experimental design is only for the model structure and parameters without considering the actual conditions of the system process.The model excitation signal(such as initial input conditions,time-varying control conditions,and etc.)is different with the actual production process.From the perspective of actual production,more attention is paid to the goal maximization of the system process(such as the target product,the substrate conversion rate,and the economic benefit).The difference between the two perspectives leads to the error between the model obtained the traditional optimal experimental design and the actual system.There is a deviation in the actual system.The main content of this article is embodied in the following three aspects:(1)This paper proposes a multi-objective optimal experimental design strategy that combines actual process control and parameter estimation.An integrated multi-objective optimal experimental design is adopted that considers the optimization objective of the system process and the uncertainty of model parameter estimation simultaneously.The framework is used to improve the accuracy of the model,the accuracy of model parameter estimation and the ability to predict the actual production were balanced by building an integrated process control and optimal experimental design framework,and the goal of process optimization is added to the goal of the traditional optimal experimental design to obtain a model that is more in line with the actual production process,so as to better express the dynamic process of the actual system.(2)For the above-mentioned multi-objective optimization problem combining process control and experimental design,it is very important to choose a suitable algorithm to optimize it.Genetic algorithm is more suitable for complex multi-objective optimization problems due to its crossover mutation probability.For the above-mentioned multi-objective optimization problem combining process control and experimental design,the NSGA-? algorithm and fuzzy optimization method were selected to solve it,and the optimal solution in the Pareto optimal solution set was obtained.The crowding distance was improved in the NSGA-? algorithm to optimize the uniformity of the population distribution,and avoid the population falling into local optimum.(3)Taking the continuous fermentation production process in the biochemical process as an example,the relevant criteria of the parameter estimation process and the substrate conversion rate in the process control were selected as the objective function,and the input substrate concentration was selected as the optimization variable to obtain multi-objective optimization problem which combined optimization experimental design and process control.The improved multi-objective genetic algorithm was used for optimization,and the Pareto optimal solution set was obtained,and then the optimal solution was determined by the fuzzy optimization method.The results showed that the method proposed in this paper can significantly improve the substrate conversion rate at the cost of a slight decrease of parameter estimation accuracy.