中文摘要：

本文采用多参数敏感性分析方法对影响污泥水解液合成PHA产量的参数进行分析.在实验数据的基础上，利用BP神经网络建立了PHA的产量预测模型.通过与真实试验结果的对比，验证了预测模型的精确度.根据训练完成的神经网络模型中的各参数变量到目标的权值和阈值，利用Garson算法定量得到各参数变量对于目标的参数敏感性系数数值.结论表明：基于BP神经网络技术建立的预测模型具有较高的可信度，多参数敏感性分析方法可评估多因素同时变化对PHA产量的影响，具有较高的实用价值.

英文摘要：

Multi-parameter sensitivity analysis method was proposed to analyze the parameters affecting the yield of Polyhydroxyalkanoate （PHA） utilizing sludge hydrolysis liquid. Based on experiment dates, a model based on back propagation neural network （BPNN） used for predicting the yield of PHA was set up. The accuracy of this predicted model was verified by contrastive analysis between theoretical and laboratorial data. On the basis of weights and threshold value of each variable parameter gained in the trained BPNN, Garson algorithm was used for calculating the parameter sensitivity coefficient. Results show the prediction model built by BPNN has high credibility, and multi-parameters sensitivity analysis method can evaluate the impact of multi-factor on PHA production yield therefore has greater practical value.