根据发酵过程中各阶段具有模糊性的特点,采用最小二乘支持向量机(1east square support vector machines,LS-SVM)提取判断发酵阶段的模糊规则,把能够在线测量的物理量如发酵时间、pH、CO_2生成速率和NH_3的添加量作为输入,通过模糊推理机判定发酵过程所处的阶段,然后依据加权平均的方法确定各个阶段葡萄糖的流加策略,实现分段连续流加补料控制。通过实验,验证了该方法的有效性。 According to the fuzziness in each stage of the fermentation process, the fuzzy rules of the fermentation stage were extracted by LS-SVM, and the physical quantities that could be measured on-line such as fermentation time, pH, CO 2 generation rate and NH 3 addition were taken as inputs. The stage of fermentation process was judged by fuzzy inference engine. Then the strategy of adding glucose in each stage was determined according to the weighted average method. Through experiments, the effectiveness of this method is verified.