摘要
针对污水处理过程的BOD建模问题,提出了一种基于回声状态网络的BOD在线软测量方法;基于梯度下降规则对回声状态网络的在线学习算法进行了研究;为保证学习算法的收敛性,基于Lyapunov理论对学习率范围进行了确定;实验表明,基于回声状态网络的在线BOD预测方法较常规神经网络预测精度提高约两个数量级,模型的适应性也大幅提高。
In order to solve the modelling problem of biochemical oxygen demand （BOD） in wastewater treatment process, this paper proposes an online BOD predictive method based on echo state network （ESN）. The gradient--based rule online algorithm is adopted to train the ESN model. To guarantee the convergence of the online learning algorithm, the range of the learning rate is determined based on Lya- punov theory. The experimental results demonstrate that the BOD prediction precision based on ESN if improved two orders of magnitude than conventional neural networks, and also the flexibility of the model is improved.
出处
《计算机测量与控制》
北大核心
2014年第5期1351-1354,共4页
Computer Measurement &Control
关键词
污水处理
BOD
回声状态网络
收敛性
wastewater treatment
biochemical oxygen demand
echo state network
convergence
