通信故障下蒸汽发生器水位MPC/DDPG切换控制

侯百乐; 齐义文; 李献领; 李鑫; 唐意雯; 郭诗彤

doi:10.3969/j.issn.2095-1248.2023.05.009

沈阳航空航天大学学报 ›› 2023, Vol. 40 ›› Issue (5) : 66 -74. DOI: 10.3969/j.issn.2095-1248.2023.05.009

信息科学与工程

通信故障下蒸汽发生器水位MPC/DDPG切换控制

侯百乐 ¹ ,
齐义文 ¹ ,
李献领 ² ,
李鑫 ¹ ,
唐意雯 ¹ ,
郭诗彤 ¹

作者信息 +

MPC/DDPG switching control for steam generator water level under communication failures

Baile HOU ¹ ,
Yiwen QI ¹ ,
Xianling LI ² ,
Xin LI ¹ ,
Yiwen TANG ¹ ,
Shitong GUO ¹

Author information +

文章历史 +

PDF (1133K)

摘要

针对核动力蒸汽发生器系统存在未知通信故障、多变工况等复杂因素，提出蒸汽发生器水位模型预测/数据驱动切换控制方法。首先，进行深度强化学习（deep deterministic policy gradient，DDPG）数据驱动控制器设计，包括状态参数选取、动作参数设计、奖励函数设计、isdone条件设计与网络设计等。在蒸汽发生器控制系统没有发生通信故障时，DDPG控制器正常运行。在此基础上，设计模型预测（model predictive control，MPC）补偿控制器来抵御通信故障，在发生通信故障时，系统切换到MPC控制器来补偿丢失的数据。仿真结果表明：在典型外部输入下，蒸汽发生器水位MPC/DDPG切换控制器均能获得良好的控制性能，并且针对未知故障表现出了较好的稳定性和安全性。

Abstract

Considering the complex factors of nuclear power steam generator system，such as unknown communication faults and variable working conditions，a model prediction/data-driven switching control method for steam generator water level was proposed. First，a deep deterministic policy gradient （DDPG） controller was designed，which includes state parameter selection，action parameter design，reward function design，was done condition design and network design，etc. The DDPG controller operates normally when there were no communication faults in the steam generator control system. On this basis，a model predictive control （MPC） predictive compensation controller was designed to resist communication failures. The system was switched to the MPC controller to compensate for the lost data when communication failures occurred. The simulation results show that the MPC/DDPG switching controller of the steam generator water level can obtain a good control performance under the typical external inputs. They also show good stability and security under unknown failures.