Hongye Su

Robust stability for uncertain discrete singular systems with time-varying delays:

Abstract This paper is concerned with the problem of delay-dependent robust stability for uncertain discrete singular systems with time-varying delays. Without introducing the free-weighting matrices to deal with the cross terms, a new and improved delay-dependent condition is established for the nominal systems to be regular, causal, and stable via a new Lyapunov function employing the integrated lower and upper time-delay bounds. The condition is also extended to the uncertain case. A numerical example is given to illustrate that the proposed methods are effective and lead to less conservatism than the existing ones.

H∞ model reduction for discrete singular Markovian jump systems

Abstract The problem of H∞ model reduction is investigated for discrete singular systems with Markovian jump. A stochastic admissibility condition and bounded real lemma (BRL) are established for the discrete singular Markovian jump system in terms of the linear matrix inequality (LMI) approach. Under partial knowledge of the transition probabilities, the strict LMI-based sufficient condition is obtained for the existence of desired reduced-order model. The numerical example is proposed to show the validity of the developed result.

A pragmatic approach for selecting weight matrix coefficients in model predictive control algorithm and its application

Selecting appropriate coefficients often requires large expensive computation duty, manual designating the coefficients in weight matrix in model predictive control(MPC) algorithm often leads infeasible control results. In this paper, a pragmatic method used to efficiently select weight matrix coefficients in MPC objective function is proposed. The controllability and the observability Gramians of a stable controlled system are needed to calculate, the controllability and observability Gramians are singular value decomposed, and the controllability and observability Gramians singular value matrices are calculated. Then, Gramians singular values of the system are used as a benchmark to select the weight coefficients corresponding to different variable in the objective function of MPC. Meanwhile, a simple theoretical justification of the method based on a energy analysis is provided. The proposed methodology is validated and estimated by two different industrial application examples and the simulation results show a better agreement comparing to manual selection weight values method.

Generalized predictive control for a precise crude oil blending process

The refinery units frequently need advanced process control (APC) strategies and near smooth real-time feed stream composition of crude oil for increasing economic profit in petroleum refining industry. So crude oil blending is a crucial process for managing crude transitions in logistics. In this paper we model the blending process as multi-input multi-output (MIMO) CARIMA model according to the characteristics that there is no coupling among variables, we control the blending process based on generalized predictive control (GPC) strategy. Meanwhile a nuclear magnetic resonance (NMR) spectroscopy analyzer is used to precisely measure the important crude oil composition of different steam of the blender. Then the real time data is used to adapt the parameters of CARIMA model. Finally, the accurateness and suitability of the proposed method is demonstrated with the case study of a crude oil blender.

A method to coordination for decentralized multi- MPC controllers in integrated plant

In this paper, a method to coordinate multi-decentralized model predictive controllers(DMMPC), which is used as a strategy to control the integrated plant through decomposing the total system model into different functional sub-models, is proposed. Usually, an integrated plant is also called a large scale systems. The main aim is to relieve the expensive computation load and for improving the flexibility and reliability for operation. How to control and coordinate the possible dynamic coupling and constraints among subsystems has already become an important problem. In order to accomplish the overall objective, coordinating the relationship between optimizer located in upper layer and multi-decentralized MPC controllers located in down layer is a challenging work. A stochastic probability density functions(PDFs) coordination rule is applied to coordinate optimizer with decentralized MPC controllers in the hierarchical control structure. This approach provides the guarantee that local optimal solutions for dominant controlled variables approximate their global optimal solutions under considering uncertainties and interactions among subunits on closed-loop stability condition. As a study case, an integrated plant which consists of two subunits connected via other process and some intermediate tanks under severe constraints while considering maximization economic performance is used to demonstrate feasibility and efficiency of this framework.

H∞ filter design for singular time-delay systems

Abstract This paper aims to solve the H∞ filtering problem for singular time-delay systems. Two new and improved delay-dependent bounded real lemmas (BRLs), which are equivalent to each other, are proposed. Based on one of them, an H∞ filter is designed via a linear matrix inequality (LMI) approach. Numerical examples are given to illustrate that the newly proposed methods introduce less conservatism than the existing ones.

Parallel adaptive hybrid genetic optimization algorithm and its application

A pragmatic hybrid genetic algorithm named parallel adaptive genetic simulated annealing (PAGSA) is developed. The proposed hybrid approach combines the merits of genetic algorithm (GA) with simulated annealing (SA) to construct a more efficient genetic simulated annealing (GSA) algorithm for global search, while the iterative hill climbing (IHC) method is used as a local search technique to incorporate into GSA loop for speeding up the convergence of the algorithm. In addition, a self-adaptive hybrid mechanism is developed to maintain a tradeoff between the global and local optimizer searching then to efficiently locate quality solution to complicated optimization problem. The computational results and application have illustrated that the global searching ability and the convergence speed of this hybrid algorithm are significantly improved.

Optimal RGC Control with Regional Poles and Control Input Constraints: the Actuator Failure Case

Based on linear matrix inequalities (LMI) approach, the problem of optimal reliable guaranteed cost (RGC) control with control input and regional poles constraints is investigated for a class of uncertain discrete-time systems subject to actuator failures. The state-feedback controller is designed to meet the given upper bounds on the amplitudes of control signals and guarantee, for all admissible time-varying uncertainties and possible actuator failures, the closed-loop system satisfying the pre-specified regional pole index and having the optimal quadratic cost performance. Thus, with the constrained control input, the resulting closed-loop system can provide satisfactory stability, transient property and cost performance despite of possible actuator faults.