论文部分内容阅读
针对污水处理过程中曝气池溶解氧浓度无法精确在线测量的问题,本文采用BP神经网络建立了溶解氧浓度预测的软测量模型。将进水参数氨和铵根离子态的氮Snh、快速可生物降解有机物Ss、异养菌生物量Xbh、颗粒性不可生物降解有机物Xi、慢速可生物降解有机物Xs以及进水流量Q作为BP神经网络软测量模型的输入变量,采用遗传算法对BP神经网络的初始连接权值和阈值进行优化。对预测结果的准确性及遗传算法优化BP神经网络的泛化能力进行了分析,讨论了数据归一化对软测量模型预测结果的影响。仿真结果表明,采用遗传算法优化BP神经网络的权值和阈值以及对训练数据归一化处理,有效地解决了溶解氧浓度BP软测量模型精度差的问题,使溶解氧软测量模型的测量精度明显增强。
In order to solve the problem that the concentration of dissolved oxygen in aeration tank can not be accurately measured online, a soft sensor model of dissolved oxygen concentration prediction is established by BP neural network. Nitrogen Snh, fast biodegradable organic matter Ss, heterotrophic bacteria Xbh, granular non-biodegradable organic matter Xi, slow biodegradable organic matter Xs and influent flow rate Q of ammonia and ammonium ion as the influent parameters were selected as BP The input variables of the neural network soft-sensing model are optimized by genetic algorithm to the initial connection weight and threshold of the BP neural network. The accuracy of the prediction results and the generalization ability of BP neural network optimized by genetic algorithm were analyzed. The influence of data normalization on the prediction results of soft-sensing model was discussed. The simulation results show that using genetic algorithm to optimize the weights and thresholds of BP neural network and normalizing the training data can effectively solve the problem of poor precision of dissolved oxygen concentration BP soft sensing model and make the measurement accuracy of dissolved oxygen soft sensing model Significantly enhanced.