MPC relevant identification––tuning the noise model

The philosophy of identification by minimizing an objective function that is commensurate with the control objective function is called control relevant identification. The control relevant method studied in this paper minimizes a multistep ahead prediction error objective function, suitable for model predictive controllers, to obtain an optimal multistep ahead predictor. It is shown that the method described in this paper provides better designed performance of the controller. A number of properties of this method in the context of FIR models are presented in this paper. The noise model plays a pivotal role in determining the performance of multistep ahead prediction errors. A method for tuning the noise model using the proposed control relevant method is presented in this paper.     

一种无约束多步递归神经网络预测控制器

本文针对具有强非线性、多工作点特性的控制系统, 提出了一种基于递归BP神经网络的多步预测模型; 通过分析预测模型的内在数学关系, 选择了二次型函数作为预测控制器的目标函数, 并给出了目标函数关于控制序列的雅可比矩阵和赫森矩阵的计算方法; 最后使用Newton-Rhapson算法设计出了滚动优化控制策略, 构建了一个非线性多步预测控制器. 仿真结果表明, 文中提出的多步预测控制器具有较好的控制效果.     

A Nonsmooth Nonlinear Programming Based Predictive Control for Mechanical Servo Systems with Backlash‐like Hysteresis

In this paper, a multi‐step‐ahead predictive control approach for dynamic systems with preceded backlash‐like hysteresis based on nonsmooth nonlinear programming is proposed. In this approach, a nonsmooth multi‐step‐ahead predictive model is developed for long‐range prediction of the controlled dynamic systems with preceded backlash‐like hysteresis. Then, the predictive control strategy is treated as a problem of on‐line nonsmooth nonlinear programming. Subsequently, the stability of the nonsmooth predictive control system is analyzed and the corresponding stability condition is derived. Afterward, a numerical example and a simulation based on a mechanical servo system are presented, respectively.     

Bridging the gap between the linear and nonlinear predictive control: Adaptations for efficient building climate control

The linear model predictive control which is frequently used for building climate control benefits from the fact that the resulting optimization task is convex (thus easily and quickly solvable). On the other hand, the nonlinear model predictive control enables the use of a more detailed nonlinear model and it takes advantage of the fact that it addresses the optimization task more directly, however, it requires a more computationally complex algorithm for solving the non-convex optimization problem. In this paper, the gap between the linear and the nonlinear one is bridged by introducing a predictive controller with linear time-dependent model. Making use of linear time-dependent model of the building, the newly proposed controller obtains predictions which are closer to reality than those of linear time invariant model, however, the computational complexity is still kept low since the optimization task remains convex. The concept of linear time-dependent predictive controller is verified on a set of numerical experiments performed using a high fidelity model created in a building simulation environment and compared to the previously mentioned alternatives. Furthermore, the model for the nonlinear variant is identified using an adaptation of the existing model predictive control relevant identification method and the optimization algorithm for the nonlinear predictive controller is adapted such that it can handle also restrictions on discrete-valued nature of the manipulated variables. The presented comparisons show that the current adaptations lead to more efficient building climate control.     

APPLICATION OF A MODEL‐BASED ITERATIVE LEARNING TECHNIQUE TO TRACKING CONTROL OF A PIEZOELECTRIC SYSTEM

This paper presents the application of a model‐based iterative learning control technique to position tracking of a piezoelectric system. Identification of the closed‐loop piezoelectric system was undertaken first, and then an iterative learning control methodology based on the identified model was implemented for dynamic tracking control of the actuator. The methodology differs from the conventional iterative learning scheme in that it takes into account the difference of the one‐step‐ahead predictive input between two successive iterations. The methodology compensates for the predictive input difference as well as the causal error in the previous iteration. The results of the experiments prove the excellence of this technique for precision tracking control of the piezoelectric actuator.     

Neural network modelling and control for underwater vehicles

Neural networks are currently finding practical applications ranging from ‘soft’ regulatory control in consumer products to accurate control of non-linear plant in the process industries. This paper describes the application of neural networks to modelling and control of a prototype underwater vehicle, as an example of a system containing severe non-linearities. The most common implementation strategy for neural control is model predictive control, where a model of the process is developed first and is used off-line to design an appropriate compensator. The accuracy and robustness of this control strategy relies on the quality of the non-linear process model, in particular its ability to predict the plant response accurately multiple-steps ahead. In this paper, several neural network architectures are used to evaluate a long-range model predictive control structure, both in simulation and for on-line control of vehicle depth, achieving accurate control with a smooth actuator signal.     

基于神经网络模型的直接优化预测控制

针对具有时延的非线性系统提出了一种基于神经网络模型直接优于的预测控制。     

基于Hammerstei模型的非线性预测控制

本文提出了一种针对 Hammerstein模型的预测控制策略.该策略将Hammerstein模型中的无记忆非线性静态增益环节,改进成易于由中间变量求取控制量的环节,避免了求解高阶方程根的困难,又对线性环节采用线性系统的广义预测控制.由于引入了广义预测控制中多步预测的思想,抗噪声的能力显著提高.仿真结果验证了该策略的有效性.     

基于Matlab和“组态王”的预测控制实验平台的建立

通过DDE（动态数据交换）可以实现组态软件与Matlab之间的通信。本文以AB2000E型过程控制设备为控制对象,以Matlab和”组态王”软件平台为工具,建立了双容水箱液位预测控制的实验平台。由“组态王”软件生成人机界面,Matlab完成预测控制算法的运算,通过实验,证实其具有较好的控制效果。通过DDE通讯,对其他先进控制算法的实测也有很好的实用价值。     

Closed-loop identification of systems with uncertain time delays using ARX-OBF structure

Closed-loop identification of systems with known time delays can be effectively carried out with simple model structures like Autoregressive with Exogenous Input (ARX) and Autoregressive Moving Average with Exogenous Input (ARMAX). However, when the system contains large uncertain time delay, such structures may lead to inaccurate models with significant bias if the time delay estimate used in the identification is less accurate. On the other hand, conventional orthonormal basis filter (OBF) model structures are very effective in capturing the dynamics of systems with uncertain time delays. However, they are not effective for closed-loop identification. In this paper, an ARX-OBF model structure which is obtained by modifying the ARX structure is shown to be effective in handling closed-loop identification of systems with uncertain time delays. In addition, the paper shows that this advantage of ARX-OBF models over simple ARX model is considerable in multi-step ahead predictions.     

基于径向基函数网络的一步超前预测控制研究

提出一种基于径向基函数（RBF）神经网络的一步超前预测控制算法。该方法只用于一个网络，控制量的获取只求几步迭代，算法简单并有较好的实用性。通过对离散非线性系统的仿真证明了算法的有效性。     

基于阻尼最小二乘法的神经网络自校正一步预测控制器

针对非线性控制器设计中遇到的模型结构及模型参数辨识问题,采用多层前馈神经网络去逼近任意的非线性系统,并使用收敛速度快且稳定性好的阻尼最小二乘法在线学习网络的仅植。基于估计的神经网络模型,依据辨识与控制的对偶原则,设计了基于阻尼最小二乘法的一步向前预测控制器。仿真研究表明,这种神经网络自校正控制器不仅具有很好的性能,而且不会产生参数爆发现象。     

Modular modelling of an evaporator for long-range prediction

This paper presents the development of a neural network model of a pilot-scale, three-effect falling-film evaporator for use in a model predictive control system. The data used in its development are from a simulation model of the evaporator. The approach taken in the neural network modelling is to divide the full model into a group of sub-networks, each modelling a specific element of the overall system. The model decomposition is determined through prior knowledge of the system. Localised computation is also used within the sub-networks to simplify the model further. The modular nature of the model gives it the capability of representing theoretical information through its topology as well as empirical information through the weights of the sub-networks. The performance of the modular evaporator neural network model is demonstrated for n-step-ahead prediction by comparing it with the analytical model of the evaporator. The results show that the model can perform satisfactory long-range predictions and hence is well suited for implementation within a model predictive control scheme.     

A model predictive control approach with relevant identification in dynamic PLS framework

In this paper, a generalized predictive control (GPC) scheme under a dynamic partial least squares (PLS) framework is proposed. At the modeling stage, a model predictive control relevant identification (MRI) approach is used to improve the identification of the model. Within PLS framework, the MIMO system can be automatically decomposed into several SISO subsystems in the latent space. For each subsystem, MRI is implemented and GPC is designed independently. With the advantage of MRI and PLS framework, fewer parameters are needed to be estimated in the identification stage, nonsquare and ill-conditioned system can be handled naturally, control parameters tuning is easier and better control performance can be obtained. Furthermore, the computing time of control action which is very crucial for GPC on-line application decreases since each GPC is running in SISO subsystem in parallel. The results of two simulation examples and a laboratory experiment demonstrate the merit of the proposed method.     

Supervisory nonlinear MPC for emergency voltage control using pattern search

The design of a higher-layer controller using model predictive control (MPC) is considered. The higher-layer controller uses MPC to determine set-points for controllers in a lower control layer. In this paper the use of an object-oriented model of the system for making predictions is proposed. When employing such an object-oriented prediction model the MPC problem is a nonlinear, non-smooth optimization problem, with an objective function that is expensive to evaluate. Multi-start pattern search is proposed as approach to solving this problem, since it deals effectively with the local minima and the non-smoothness of the problem, and does not require expensive estimation of derivatives. Experiments in an emergency voltage control problem on a 9-bus dynamic power network show the superior performance of the proposed multi-start pattern-search approach when compared to a gradient-based approach.     

Output constrained IMC controllers in control systems of electromechanical actuators

Electromechanical actuators are widely used in many industrial applications. There are usually some constraints existing in a designed system. This paper proposes a simple method to design constrained controllers for electromechanical actuators. The controllers merge the ideas exploited in internal model control and model predictive control. They are designed using the standard control system structure with unity negative feedback. The structure of the controllers is relatively simple as well as the design process. The output constraint handling mechanism is based on prediction of the control plant behavior many time steps ahead. The mechanism increases control performance and safety of the control plant. The benefits offered by the proposed controllers have been demonstrated in real-life experiments carried out in control systems of two electromechanical actuators: a DC motor and an electrohydraulic actuator.     

局部高斯分布拟合的脑MR图像分割及有偏场校正

为实现对灰度不均匀脑核磁共振(MR)图像分割的同时进行有偏场估计并校正,提出一种基于局部高斯分布拟合(LGDF)模型的多相水平集方法.通过分析图像有偏场模型的局部特性,将有偏场乘性因子引入到图像局部灰度均值的表达中,从而使有偏场乘性因子成为新的能量函数的变量.能量函数的迭代最小化既实现了目标组织分割,又有效估计了有偏场.合成图像和仿真脑MR图像实验结果表明,本文方法比现有多种方法分割性能更好,且利用本文方法估计的有偏场校正后的图像有更好的视觉效果.     

Neural network model-based automotive engine air/fuel ratio control and robustness evaluation

Automotive engines are multivariable system with severe non-linear dynamics, and their modelling and control are challenging tasks for control engineers. Current control of engine used look-up table combined with proportional and integral (PI) control and is not robust to system uncertainty and time varying effects. In this paper the model predictive control strategy is applied to engine air/fuel ratio control using neural network model. The neural network model uses information from multivariables and considers engine dynamics to do multi-step ahead prediction. The model is adapted in on-line mode to cope with system uncertainty and time varying effects. Thus, the control performance is more accurate and robust compared with non-adaptive model based methods. To speed up algorithm calculation, different optimisation algorithms are investigated and performance compared. Finally, the developed method is evaluated on a well-known engine benchmark, a simulated mean value engine model (MVEM). The simulation results demonstrate the effectiveness of the developed method.     

Avoiding the Look‐Ahead Pathology of Decision Tree Learning

Most decision‐tree induction algorithms are using a local greedy strategy, where a leaf is always split on the best attribute according to a given attribute‐selection criterion. A more accurate model could possibly be found by looking ahead for alternative subtrees. However, some researchers argue that the look‐ahead should not be used due to a negative effect (called “decision‐tree pathology”) on the decision‐tree accuracy. This paper presents a new look‐ahead heuristics for decision‐tree induction. The proposed method is called look‐ahead J48 ( LA‐J48) as it is based on J48, the Weka implementation of the popular C4.5 algorithm. At each tree node, the LA‐J48 algorithm applies the look‐ahead procedure of bounded depth only to attributes that are not statistically distinguishable from the best attribute chosen by the greedy approach of C4.5. A bootstrap process is used for estimating the standard deviation of splitting criteria with unknown probability distribution. Based on a separate validation set, the attribute producing the most accurate subtree is chosen for the next step of the algorithm. In experiments on 20 benchmark data sets, the proposed look‐ahead method outperforms the greedy J48 algorithm with the gain ratio and the gini index splitting criteria, thus avoiding the look‐ahead pathology of decision‐tree induction.