Salem A.K. Al-Assadi

Design of decentralized robust controllers for multizone space heating systems

Temperature control of multizone environmental spaces is studied. A bilinear model of a multizone space heating (MZSH) system is developed. The MZSH system is acted upon by constant unknown disturbances and step changes in the desired set-points. Robust decentralized controllers are designed with linear and nonlinear models, using two different methods based on optimization of quadratic performance criteria. In each case, the output responses of the resulting closed-loop bilinear system are examined. Simulation results showing the regulation properties of the decentralized and centralized controllers are presented. The results show that the system performance with decentralized control is very close to the performance of the system with centralized control. A decentralized controller is also designed for the bilinear model based on minimizing not only the output error and energy input, but also a measure of sensitivity to parameter variations. Simulation results for the corresponding bilinear closed-loop system are presented. >

Decentralized preview control for multiple disturbance rejection in HVAC systems

Abstract A decentralized preview controller is designed for temperature control of multizone indoor environmental spaces. A two-zone space heating system is considered. The physical system consists of a boiler, heat pumps, distribution network and two environmental zones. By assuming that the outdoor temperature variations are “previewable”, a decentralized preview controller is designed by using a parameter optimization method. The output responses of the resulting decentralized closed-loop bilinear system acted upon by single and multiple disturbances with and without preview action are compared. Also, results showing the robustness property of the controller, and the 24-hour building operation with unoccupied and occupied setpoint tracking using preview control are given.

Computer-aided optimisation of simplified discrete models for control systems

A computer-aided method for the computation of optimal discrete models with reduced orders for control systems is presented in this paper. The method used for simplification utilises a time-domain-optimisation technique to search for optimum values for the parameters of prespecified reduced-order discrete-time models for control systems. The measure of success of simplification is taken to be the minimum integral squared error between the step responses of the simplified discrete-time models and that of the actual system at sampling instants. Two examples are considered to illustrate the usefulness of the method in simplifying control systems given in various forms. One of the examples considered was to obtain reduced-order discrete-time models for a fourth-order Butterworth filter. The method proved to be powerful in finding the optimal discrete models with reduced orders for both short and long sampling periods.

Computation algorithms for disturbance rejection in multivariable systems

In this paper, one way of solving an important problem in linear time-invariant multivariable systems control for synthesizing feedback controllers to make the outputs of a physical system respond in a desirable manner to reference inputs and external disturbances is presented. The proposed algorithms can be regarded as a logical extension of the pole assignment problem in that, measurable or unmeasurable multiple disturbance acting on multivariable systems described by the 4-tuples (A,B,C,E) or the 6-tuples (A,B,C,D,E,F) can be rejected at the outputs in steady state. This is done by assignment of the `disturbance blocking zeros at specified locations using the concept of pole assignment for computing the state feedback controllers. The performance of the algorithms is illustrated by a numerical example.

Temperature Control of Multi-Zone Space Heating Systems Using Decentralized Robust Controllers

Abstract In this paper, temperature control of multi-zone environmental spaces is studied. Decentralized and centralized controllers are designed based on solving constrained servomechanism problems for a bilinear system. Simulation results are presented which show the regulation properties of the decentralized and centralized controller designed for the bilinear model. A decentralized controller is also designed for the bilinear model based on minimizing not only the output error and input energy but also a measure of sensitivity to parameter variations. Simulation results for the corresponding bilinear closed-loop system are presented.

An approach for pole assignment by dynamic compensator with prespecified poles

In this paper, a new approach for the design of physically realizable dynamic output feedback with prespecified poles to achieve pole assignment in multivariable systems is presented. The proposed approach computes a non unity-rank compensator with lower order in two steps. In the first step, we assign a number of poles by means of constant output feedback using a method based on the implicitly shifted QR algorithm for solving the algebraic eigenvalue problem. In the second step, the assigned poles are preserved and a number of additional poles are placed using a unity-rank dynamic compensator with prespecified poles computed entirely in the frequency domain. Numerical example is provided to illustrate the performance of the proposed approach.

Frequency-weighted least-squared error for optimal discretization of continuous-data control systems

Abstract This paper is concerned with developing a computer-aided method for optimum discretization of continuous-data control systems. In this method, we can assume different orders for the digital controllers, and a search for optimum values for their parameters is carried out by minimizing the sum of the magnitude-squared error function formulated in the frequency domain. Results of all examples considered show that the method is efficient and has advantages when compared to other existing methods in minimizing the magnitude-squared error in both the frequency and time domains for different sampling rates.

Disturbance rejection in multivariable systems by state feedback controller

Abstract In this paper, computational algorithms for designing a state feedback controller to reject the disturbances acting on linear time-invariant multivariable systems is presented. The theoretical basis for these algorithms is a factorization procedure for the transfer function matrix between the outputs and the disturbance. This enables us to use the concept of a minimal order inverse to determine the position of the ‘disturbance blocking zeros’ which can be assigned by state feedback to desired locations, such that all measurable or unmeasurable exponential disturbances are rejected in the steady state. The performance of the algorithms are illustrated by two numerical examples.