| 引用本文: | 张伟,乔俊飞,李凡军.溶解氧浓度的直接自适应动态神经网络控制方法[J].控制理论与应用,2015,32(1):115~121.[点击复制] |
| ZHANG Wei,QIAO Jun-fei,LI Fan-jun.Direct adaptive dynamic neural network control for dissolved oxygen concentration[J].Control Theory and Technology,2015,32(1):115~121.[点击复制] |
|
| 溶解氧浓度的直接自适应动态神经网络控制方法 |
| Direct adaptive dynamic neural network control for dissolved oxygen concentration |
| 摘要点击 2391 全文点击 1536 投稿时间：2014-04-14 修订日期：2014-06-28 |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| DOI编号 10.7641/CTA.2014.40311 |
| 2015,32(1):115-121 |
| 中文关键词 动态神经网络控制器 溶解氧 规则无用率 污水处理过程 |
| 英文关键词 dynamic neural network controller dissolved oxygen useless rate wastewater treatment process |
| 基金项目 国家自然科学基金项目(61034008, 61225016), 北京市自然科学基金项目(4122006), 教育部博士点新教师基金项目(20121103120020)资助. |
|
| 中文摘要 |
| 针对污水处理过程溶解氧浓度的控制问题, 提出一种直接自适应动态神经网络控制方法(direct adaptive dynamic neural network control, DADNNC). 构建的控制系统主要包括神经网络控制器和补偿控制器. 神经网络控制器由自组织模糊神经网络实现系统状态与控制量之间的映射; 提出一种基于规则无用率的结构修剪算法, 并给出结构调整后网络收敛的理论证明. 同时, 为保证系统稳定, 设计补偿控制器减小网络逼近误差, 参数调整由Layapunov 理论给出. 国际基准仿真平台上的实验表明, 与固定结构神经网络控制器、PID和模型预测控制等已有控制方法相比, DADNNC方法具有更高的控制精度和更强的适应能力. |
| 英文摘要 |
| A direct adaptive dynamic neural network control (DADNNC) method is proposed to control the dissolved oxygen concentration in the wastewater treatment process. The established control system mainly includes a neural controller and a compensate controller. The neural controller fulfills the mapping between the system states and control variable using the fuzzy neural network, which can adjust the structure and parameters simultaneously. A novel pruning algorithm is presented based on the useless rate of the rules, and the convergence while adding and pruning neurons is guaranteed theoretically. Further, the compensation controller is designed for decreasing the approximating error introduced by the neural network, and the parameter update law is deduced by the Lyapunov theorem. Finally, the simulation results, based on the international benchmark simulation platform, show that the proposed method can achieve better control accuracy and superior adaptive ability compared with neural network controller with fixed structure, PID controller and model predictive control method. |
|
|
|
|
|