燃煤机组过热汽温宽负荷模型前馈控制

Feedforward Control of Superheated Steam Temperature with Wide Load Model for Coal-fired Units

为了对燃煤机组过热汽温宽负荷运行时进行更精确地前馈控制,提出一种基于物理引导神经网络(PGNN)的预测前馈信号模型,并基于间隙度量法确定了多模型的负荷段分配。多模型间隙度量PGNN预测方法采用多模型间隙度量方法对负荷区段进行合理划分,结合过热器机理引导的长短期记忆神经网络,可以对强耦合、大惯性的过热汽温宽负荷前馈信号进行精准预测。结果表明:在机组宽负荷运行时,随着负荷降低控制对象的非线性程度逐渐增强,需要更多的模型数量,采用多模型间隙度量PGNN前馈控制方法可以在不同工况下采用与当前工况相适应的前馈信号,有效提升过热汽温的调节精度和稳定性。

In order to achieve more accurate feedforward control for the superheated steam temperature with wide load operation,a predictive feedforward signal model based on physical guided neural network(PGNN)was proposed,and the load segment allocation of multiple models was determined based on gap measurement.The PGNN prediction method of multi-model gap measurement adopted the multi-model gap measurement method to reasonably divide the load section.Combined with the long short term memory neural network guided by the superheater mechanism,the strong coupling and large inertia superheated steam temperature wide load feedforward signal can be accurately predicted.Results show that when in the wide load of the unit,the nonlinear degree of the control object gradually increases with the load reduction,and more models are needed.The multi-model gap measurement PGNN feedforward control method can adopt feedforward signals suitable for the current working conditions under different working conditions,and improve the adjustment accuracy and stability of superheated steam temperature.