| 摘 要: 膜污染问题是影响MBR推广应用的主要障碍。针对该问题，首先运用主成分分析法确定影响膜污染的主 要因子为MLSS、操作压力及温度，然后建立基于RBF神经网络的预测模型，再利用遗传算法对预测模型的四个参数进 行优化，使得整个网络模型达到全局最优。最后将预测结果与样本数据对比，结果显示，基于GA-RBF的膜污染模拟 仿真器在收敛速度、预测精度等方面比单纯的RBF网络有较大提高，达到了预期目标。 |
| 关键词: MBR；膜通量；RBF神经网络；遗传算法 |
|
中图分类号: TP389.1
文献标识码: A
|
| 基金项目: 受国家自然科学基金项目（51378350）；国家自然科学基金项目（50808130）资助 |
|
| Research of RBF Neural Network Based on Genetic Algorithm Optimization in MBR Membrane Pollution Simulation |
|
TANG Jia，LI Chunqing
|
( School of Computer Science and Software, Tianjin University of Technology, Tianjin 300387, China)
|
| Abstract: Membrane fouling is the main obstacle to the promotion and application of MBR.To solve this problem, the main factor is the first use of principal component analysis to determine the influence of membrane fouling is MLSS,operating pressure and temperature,and then build the prediction model based on RBF neural network,and then use genetic algorithm four parameters of the prediction model is optimized,which makes the whole network model to achieve global optimum.The predicted results compared with the sample data shows that the membrane fouling of GA-RBF simulator in terms of the convergence speed and prediction accuracy is higher than that of simple RBF neural network based on,to achieve the expected goal. |
| Keywords: MBR;membrane flux;RBF neural network;genetic algorithm |
