Online Modeling Method Based on Dynamic Time Warping and Least Squares Support Vector Machine for Fermentation Process

摘要

A new online local modeling method is proposed for fed-batch fermentation processes based on dynamic time warping (DTW) and least squares support vector machine (LS_SVM). In this method, a set of data within the sliding window is set as a query sequence in the current process, and then search for the most similar sub-sequence from the historical batch database to form the training set. At last, this training set will be used for build online local model based on LS_SVM. A forecast model of penicillin's concentration is constructed based on the proposed method and off-line global modeling method using the data generated by the Pensim fermentation simulation platform. The simulation result shows that this method has a higher forecast accuracy and dynamic adaptability compared with the traditional offline modeling method.