Study on optimal scheduling strategy of CTC fermentation

Considering these online unmeasurable parameters affecting the optimized control, an optimal scheduling strategy was proposed on the basis of data field clustering, fuzzy neural networks and rolling-learning prediction. With the eigenvalues of data field clustering information for input variables initializing model parameters of the fuzzy neural networks, this solution eliminated the stochastic error of artificial parameter selection. In addition, the algorithm of correcting the model making use of offline data was added into the prediction model. Therefore, the proposed strategy improved the estimation accuracy of these unmeasurable parameters of CTC fed-batch process and enhanced the robustness of prediction models. The experimental results show that the scheme increases the production efficiency of the CTC fermentation and takes on a promising practical application.