Abstract: Support vector machines(SVM) based on structural risk minimization（SRM）is used to model batch fermentation process by considering the complexity and the problem of severe non-linearity in fermentation process.The model based on SVM was featured by less complexity and better prediction ability than that based on artificial neural networks(ANN).The hybrid support vector machines(HSVM) model was proposed by combing SVM with fermentation kinetics mechanism in serial and serial-parallel approaches and used to predict the state of biomass in fermentation.The mechanism analysis and the simulation results showed higher prediction accuracy, more powerful generalization ability and better performance of these two kinds of HSVM in modeling than those based on black-box SVM, simple and approximate kinetics model and the corresponding hybrid ANN.

摘要： 针对生物发酵过程机理复杂、高度非线性的特点，采用基于结构风险最小的支持向量机为发酵过程建模，其算法规范，建模复杂度低于神经网络方法，所建模型的预测效果更好.还将生化过程的动力学机理与支持向量机相结合，采用串联和串并联结构，提出与机理杂交的支持向量机建模方法，并为间歇式酒精发酵过程中酵母菌体浓度变化建立了预测模型.原理分析与试验结果表明与机理杂交的支持向量机建模方法，相比于单一近似的动力学模型、单一的支持向量机模型，以及机理杂交的神经网络模型，它的预测精度高，泛化能力强，性能更为优越.