Development of NARX-DNN Model for Coal-fired Boiler Using Real Plant Data

来源 :第26届中国过程控制会议 | 被引量 : 0次 | 上传用户:usernameing
下载到本地 , 更方便阅读
声明 : 本文档内容版权归属内容提供方 , 如果您对本文有版权争议 , 可与客服联系进行内容授权或下架
论文部分内容阅读
  In this paper,a deep learning model referred as NARX-DNN based on Deep Neural Network(DNN)and Nonlinear Auto Regressive with Extension X input(NARX)neural network is developed to model the coal-fired boiler.First,the distributed control system(DCS)data is trained by DNN to extract highly abstract features for NARX.Then the NARX neural network is trained to predict parameters for coal-fired boiler with the output of DNN.Moreover,with the powerful learning ability of NARX-DNN model to deal with the time delay problem,the long-time predicting of operation status can be available.Thanks to the feature learning ability of DNN,the overwhelmingly complex property of coal-fired boiler can be represented more reasonable and effective.Experiments are provided to illustrate the effectiveness of the proposed design approach.Compared with traditional TDNN and ARNN model,our approach achieves the lowest error rate.
其他文献
It has more special significance to design a high-efficient wireless sensor network with energies constrained.A kind of affinity propagation clustering algorithm was used for routing algorithm of hete
CO2 管道运输系统是碳捕集、利用与封存(CCUS)技术的重要环节。优化配置管道入口压力、管道内径、壁厚、中间泵站数量,不仅能够为管道安全运输提供保障,而且可以有效降低系统费用。本文将线性鲁棒优化方法应用于CO2 管道运输系统设计,建立其鲁棒优化模型,解决了管道运输系统不确定性优化问题,使优化结果更为合理。仿真结果表明,与现有优化设计方法相比,运输系统鲁棒优化算法具有较强的适应能力。
A new multiple-model predictive control scheme for the rare earth extraction process is proposed for control of the component content.
The design of a benchmark cogeneration system is a multi-objective optimal problem,it needs a simultaneousconsideration of thermodynamic efficiency and economic objective.On the one hand,we hope the t
This paper investigates the controllability of Boolean control networks(BCN)with time-delays in state by using the semi-tensor product of matrices.The definition of controllability matrices for BCNs w
In this paper,controller performance assessment is considered for a class of switched linear systems.The tensor-based method is proposed for the controller performance assessment for this systems evol
This article discusses MPCA(Multi-way Principal Component Analysis)and MPLS(Multi-way Partial Least Squares)have been used to compress the information into low-dimensional spaces and pinpoint the root
研究了一类δ算子系统的滑模控制问题。首先,研究了带有时滞的δ算子不确定系统,并引入一种能够动态改善系统运动品质的非线性滑模面,利用线性矩阵不等式技术,得出了滑模面存在的充分条件,并给出具体设计方法。其次,基于一种改进的自适应趋近律,设计了相应的滑模控制器,保证系统状态能够在有限时间内到达滑模面,从而将此类非线性系统的滑模控制的设计推广到了现代高速采样条件下的时滞不确定非线性系统。最后,通过一个仿真
This paper focuses on a 4-DOF manipulator with flexible beam frame.Due to the performance requirements of rapidity and stability,the model predictive control algorithm is adopted to make trajectory pl
In the process industries,local linearization of a nonlinear process,which is essentially modeling of the process dynamics at a specific operating point with a simplified linear model,is commonly used