基于参数估计误差的辊道窑温度场优化控制方法

Temperature field optimization control of roller kiln based on parameter estimation error

中文关键词  辊道窑温度场  分布参数系统  策略迭代  参数估计误差  最优控制

英文关键词  temperature field of roller kiln  distributed parameter system  policy iteration  parameter estimation error  optimal control

中文摘要

      辊道窑烧结过程的温度是决定锂离子电池正极材料产品质量的关键. 然而, 根据炉内有限个测温点的温度 建立起描述整个温度场的模型往往非常困难, 导致无法优化控制烧结过程的温度分布; 而控制方法的设计一般需要 进行参数估计, 已有参数估计方法大多依赖于观测器/预测器的状态误差信息, 无法直接反映待估计参数的变化特 征且方法的准确性取决于观测器/预测器的性能. 为此, 本文提出一种基于参数估计误差的温度场自适应动态规划 (adaptive dynamic programming, ADP)优化控制方法. 首先, 基于传热机理建立二维多孔介质能量守恒方程, 构建包 含角系数的边界条件以反映热辐射作用; 考虑到竖直方向温度变化较大, 通过转换边界条件建立起辊道窑一维温 度场模型, 并根据正极材料的特性获得模型参数. 然后, 采用ADP中的策略迭代(policy iteration, PI) 优化设计温度场 控制器, 神经网络(neural network, NN)用于PI中的评价网络以逼近代价函数; 基于权值参数的估计值与真实值之差 构建参数估计误差, 通过将估计误差的信息融入到评价NN参数更新过程, 提出基于参数估计误差的NN权值更新算 法, 以提高参数估计误差的收敛性, 实现有限时间内NN权值的快速收敛. 最后, 通过仿真验证所提建模和控制方法 的有效性.

英文摘要

      The temperature of roller kiln sintering process is the key to determine the quality of lithium-ion battery cathode materials. According to limited measurement points, it is difficult to establish a model describing the temperature field, and leads to the control failure. Parameter estimation is generally required during controller design. Most of existing methods rely on states error information of observer/predictor, which cannot directly reflect the variation of parameters to be estimated. The accuracy of these methods depends on the observer/predictor performance. This paper proposes a temperature field adaptive dynamic programming (ADP) optimal control method based on parameter estimation error. Firstly, the energy conservation equation of two-dimensional porous media is established based on heat transfer mechanism, and the boundary conditions including view actors are constructed to reflect heat radiation. Considering that the vertical direction temperature varies greatly, a one-dimensional temperature field model is established by transforming boundary conditions, and the model parameters are determined according to cathode materials. Then, policy iteration (PI) method of ADP is used to design optimal control method. A neural network (NN) is used in critic network to approximate cost function. Defining parameter estimation errors based on differences between estimated and true values of NN weights, and integrating the error information into updating process, a new weight updating algorithm is proposed to improve the convergence of parameter errors as well as the rapid convergence of NN weights in a finite time. Finally, the effectiveness of the proposed methods is verified by simulation.