| 摘要: |
| 针对污水处理过程溶解氧质量浓度控制问题,提出一种基于神经网络动态规划的控制器设计方法。该方法不需要建立污水处理过程的非线性动力学模型,控制器的设计只需要系统的输入、输出观测信息。控制器设计采用评价—行动的思想,策略的评价值及最优行动分别采用两个回声状态网络逼近,给出评价网络的收敛条件。对污水处理过程溶解氧的控制试验结果表明,与常规PID控制相比,神经网络动态规划控制器能够有效提高控制精度,抑制干扰能力也明显增强。 |
| 关键词: 人工神经网络 自适应动态规划 污水处理 溶解氧 过程控制 收敛性 |
| DOI：10.3969/j.issn.1673-5005.2013.01.031 |
| 分类号::TP 273 |
| 基金项目:国家自然科学基金重点项目(61034008) |
|
| Application of neural dynamic programming to dissolved oxygen control |
|
BO Ying-chun1, LI Lai-hong2, MA Shan-peng3, XIA Bo-kai1
|
|
(1.College of Information and Control Engineering in China University of Petroleum, Qingdao 266580, China;2.Hekou Power Supply Corporation, Shengli Oilfield, Dongying 257200, China;3.Shandong Shtar Science & Technology Group Company Limited, Dongying 257062, China)
|
| Abstract: |
| An controller design method based on neural dynamical programming controller(NDPC) was proposed to solve the problem of dissolved oxygen control in wastewater treatment plant. Without considering any mechanism model of the activated sludge wastewater treatment plant (WWTP), the controller was designed by using the input-output observed data. The controller adopted actor-critic idea. Two echo state networks were adopted to approximate the value function and the optimal control policy, respectively. The convergence condition of the critic network was given. The experimental results show that the NDPC achieves higher control precision and better robustness than PID control strategy. |
| Key words: artificial neural networks adaptive dynamical programming wastewater treatment dissolved oxygen process control convergence |
