基于神经网络解耦的换热站IGPC控制策略

IGPC Control Strategy of Heat Exchange Station Based on Neural Network Decoupling

针对供热系统中质、量调节回路耦合而导致控制效果不佳的问题,提出一种基于换热站质、量分离调节的神经网络解耦隐式广义预测控制(IGPC)策略。首先结合换热站工艺设计基于IGPC和PID的双入双出控制系统;其次通过时滞递归神经网络对预测模型进行解耦,利用滚动优化和在线校正环节克服质调节通道的时变、时滞影响。仿真结果表明:IGPC解耦控制与常规动态矩阵控制(DMC)和IGPC未解耦控制策略相比,在同等扰动的情况下质调节通道稳定时间最多减少95s,超调量最多减少11.3%。该策略的自适应性和鲁棒性可满足实际热工现场需求,同时为类似工业领域控制系统的研究提供了一定的参考和指导。

Aiming at the problem of poor control effect caused by the coupling of quality and quantity control loops in the heating system, a neural network decoupling implicit generalized predictive control(IGPC) strategy based on the separation and regulation of quality and quantity of heat exchange stations is proposed in this paper.First,design a dual-input and dual-output control system based on IGPC and PID in combination with the heat exchange station process.Secondly,the predictive model is decoupled through the time-delay recurrent neural network,and the rolling optimization and online correction links are used to overcome the time-varying and time-lag of the quality adjustment channel.Influence.The simulation results show that compared with the conventional dynamic matrix control(DMC) and IGPC undecoupling control strategies,the IGPC decoupling control can reduce the stability time of the quality control channel by up to 95 s under the same disturbance;the overshoot can be reduced by up to 11.3%.The adaptability and robustness of the strategy in this paper can meet the actual thermal site requirements.