On-line parameter and state estimation of an air handling unit model: experimental results using the modulating function method

Abstract

This paper considers the on-line implementation of the modulating function method, for parameter and state estimation, for the model of an air-handling unit, central element of HVAC systems. After recalling the few elements of the method, more attention is paid on issues related to its on-line implementation, issues for which we use two different techniques. Experimental results are obtained after implementation of the algorithms on a heat flow experiment, and they are compared with conventional techniques (conventional tools from Matlab for parameter estimation, and a simple Luenberger observer for state estimation) for their validation. * ai@mci.sdu.dk † ramezani@mci.sdu.dk ‡ jerome@mci.sdu.dk arXiv:1810.08929v2 [cs.SY] 15 Nov 2018 state and parameter estimation on a continuous-time ROM based on a simple thermal-electrical analogy [15]. In comparison with more conventional approaches using discrete-time such as Kalman filtering, performing both state and parameter estimation in continuous-time relates better to the formalism in which the system is originally modeled and ensures the convergence of the estimates to the real values when sampling time approaches zero [39]. Compared to their discrete-time counterparts estimating either parameters or state components is the necessity of considering time derivatives. The modulating function technique, originally introduced by Shinbrot [38], allows to circumvent this issue elegantly, with the use of fixed-length time integrals of the measured signals, akin continuous-time FIR filtering. This offers the additional advantages of giving estimates after a fixed and predetermined amount of time, contrary to usual Kalman-based filtering methods. Originally proposed towards parameter estimation, the modulating function method was more recently extended to include state estimation [23] [19].Many studies considering the modulating function approach showed satisfactory performance, but mostly in simulations using artificial data [19] or sources generated from real profiles [5]. In contrast to that and to the best of our knowledge, not many applications can be found in the literature dealing with experiments done on actual measurements, to the notable exception of [11], where Hartley modulating functions are used to estimate the parameters of a thyristor driven DC-motor.

Identification and control of a heat flow system based on the Takagi-Sugeno fuzzy model using the grey wolf optimization algorithm

“…The temperature control has an extensive range of applications in petrochemical, biochemical and pharmaceutical industries. Modeling and control of the heat flow experiment, which is the plant in this research as well, was done in papers [1,2,3]. The proposed method in [1] is based on a set of frequency-domain data to design fixed-order controllers capable of providing satisfactory performance profiles and constrained control inputs.…”

“…The design problem is formulated as a constrained optimization problem, which has been solved using the genetic algorithm (GA) to find suitable parameter values of a selected controller to achieve satisfactory change in temperature. Paper [2] examines the on-line implementation of the modulating function method, for parameter and state estimation, for the model of an air-handling unit, central element of HVAC systems. In [3] authors have conveyed and stressed the issue of performance analysis of fractional-order controller designs for integer first-order plus time delay systems.…”