Soft Sensing And Feed Predictive Control Of Marine Lysozyme Fermentation Process Based On IFA-MLSSVM

Marine lysozyme as a new source of high-basic protein,due to the diversity of marine microorganisms and the special high-salt,high-pressure,low-temperature growth environment,it has a higher quality of biologically active substances than traditional land-based lysozyme.It has no side effects,no pollution,and has high-efficiency sterilization and anti-viral effects.It has been used in the food industry,pharmaceutical clinical,bioengineering,livestock and poultry breeding industries.In actual marine lysozyme fermentation production,so as to obtain higher-yield and efficient products,save materials and costs,avoid environmental pollution and achieve large-scale production,it is necessary to effectively control the marine lysozyme fermentation process.However,the fermentation of marine lysozyme involves the growth and metabolism of bacteria,which is highly time-varying and uncertain.The whole fermentation process is affected by various parameters,showing strong nonlinearity and coupling.It is difficult to control the fermentation process by traditional control methods.At the same time,many key parameters(cell concentration,matrix concentration,relative enzyme activity)that reflect the fermentation quality are difficult to be measured directly online,and the offline detection methods commonly used have a long time lag and risk of infection.These problems severely restrict the large-scale industrial production of marine lysozyme.In view of the above problems,this paper focuses on real-time online measurement and feed prediction control of key parameters of the marine lysozyme fermentation process.The specific research contents are as follows:Firstly,through a brief introduction of the basic principles of LS-SVM,aiming at the limitations of traditional LS-SVM kernel function,a mixed kernel function LS-SVM modeling is proposed to improve the modeling accuracy.Aiming at the problem of randomness and empirical dependence of MLS-SVM parameter selection,the firefly algorithm with good performance is used to optimize the model parameters,and an improved firefly algorithm is proposed.The behavior optimization,search stepoptimization and initialization distribution optimization of FA are carried out to solve the problems of slow search speed and poor search accuracy of standard firefly algorithm.The IFA-MLSSVM model is constructed,and its forecasting effect is preliminary feasibility verified,which provides theoretical support for the construction of the soft sensing and feed prediction model in following chapters.Secondly,a soft sensing model based on IFA-MLSSVM is constructed and applied to the online measurement of the key parameters of the multi-input and multi-output marine lysozyme fermentation process.Based on the analysis of fermentation technology,the coupling and influence of various parameters in fermentation process,a dynamic model related to the dominant variables is established,and the auxiliary variables closely related to the dominant variables are selected by the method of consistent correlation degree.A soft sensing model of marine lysozyme fermentation process based on IFA-MLSSVM is established.In order to improve the prediction accuracy of each dominant variable,according to the size of the prediction error of the soft sensing model established for different dominant variables,the most appropriate mixed kernel function is selected.Finally,aiming at the nonlinear and strong coupling characteristics of the marine lysozyme fermentation process,an IFA-MLSSVM-based nonlinear multi-step predictive control model is established.Aiming at the local optimal phenomenon of solving the control quantity problem of objective function in rolling optimization,IFA is used to improve it,and a closed-loop feeding control system of fermentation process is constructed based on soft sensing and feed predictive control method.From the control effect of the dominant variables,the tracking effect in the presence of system interference and the simulation results of the tracking error,it can be seen the effectiveness of this method in the feeding control of marine lysozyme fermentation process,and it has a good tracking effect on the actual curve of marine lysozyme fermentation experiment.