Control of a reverse flow reactor for VOC combustion

https://doi.org/10.1016/j.ces.2004.10.020Get rights and content

Abstract

A flow reversal reactor for VOC combustion is controlled by the linear quadratic regulator (LQR), which uses dilution and internal electric heating as controls to confine the hot spot temperature within the two temperature limits, in order to ensure complete conversion of the VOC and to prevent overheating of the catalyst. Three phases of operation, i.e., dilution phase, heating phase and inactive phase, are identified. In dilution and heating phases, the cost functions of the LQR control are defined in quadratic forms. In the inactive phase, the controllers are inactivated. A linear model is derived by linearization of a countercurrent pseudo-homogeneous model at two nominal operating conditions in the dilution phase and the heating phase, respectively. The feed concentration and the temperature profile are estimated on-line by using a high-gain observer with three temperatures measurements and are used in the LQR feedback control. Experiments are carried out on a medium-scale reversed flow reactor to demonstrate the proposed LQR control strategy. Results show that the LQR controller is highly efficient in maintaining normal operation of the reactor.

Introduction

Reverse flow reactor (RFR) is a packed-bed reactor with the flow direction periodically reversed. It was conceived more than 20 years ago and ever since has evoked wide interests among academia and industry. Through periodic flow reversal, heat released by reaction is first trapped in the packing, which is then used to heat up the feed when the flow direction is reversed. In the RFR, heat exchange between the feed and the effluent is highly efficient. As a result, autothermal operation is possible even if the feed has an adiabatic temperature rise as low as 10–15 K. Moreover, owing to the large heat capacity of the packing, the high-temperature plateau established in the packed bed is not sensitive to abrupt changes in the inlet concentration, making the reactor easy to operate and complete conversion of the reactant easily achievable. These features make RFR highly competitive for VOC combustion.

For efficient temperature control, different reactor configurations of RFR have been proposed. For rich feed, one can use cold gas injection, hot gas withdrawing, or heat recovery through internal heat exchangers to suppress temperature run-away; while for lean feed, hot gas supply, or internal heating can be applied to prevent extinction (Cunill et al., 1997).

Despite the large number of publications contributed to maintain safe and smooth operation of the RFRs, only a few papers have been found dealing with control of the RFR. In the pioneering work of Budman et al. (1996), a PID and a feed-forward controller were employed to control a reversed flow reactor for carbon monoxide combustion. Pikes were observed at times of flow reversal when a PID controller was used. The feed-forward controller provided more efficient control of the conversion. However, it required on-line measurement of inlet concentration. The other is the work of Xiao et al. (1996), who controlled an RFR for sulfur dioxide conversion and compared the performance of proportional, feed-forward proportional and bang-bang control. The feed-forward proportional control was found slightly better than the simple proportional control, but it also required on-line concentration measurement. The bang-bang control was easily applicable, however, at the cost of more cooling and resulting in lower sulfur dioxide conversion. Both research groups used a reversal flow reactor with long cycles of 1200 at 2500 s, and their investigation was made by simulation.

Owing to the intrinsic unsteady-state nature of the RFRs, the state variables change periodically even if the reactor has reached a cyclic steady state. Therefore, the controller is needed only to confine the variation of state variable within a certain range. This is relatively easy if good control variables have been chosen. However, again because of the unsteady-state nature, the dynamic response of control, and the control cost in particular, may be disregarded.

In this paper, the LQR control strategy with well-defined control costs and performance is proposed, which is then validated experimentally on an RFR for VOC combustion.

Section snippets

Process description

A medium-scale RFR for VOC combustion, as schematically shown in Fig. 1 (Ramdani et al., 2001), is considered in this work. Liquid xylene is employed as the VOC, which is injected into the air through a capillary tube and is then vaporized before entering the reactor. Cordierite monoliths (Corning) of square cross-sections with channels of 1×1mm2 are packed in the reactor. Monolith in the core region is catalytically active and is inert in both end regions. Nieken et al., 1994a, Nieken et al.,

Controller design

The controller design that we propose in this article is based on the LQR techniques, and feedback control is derived from the minimization of a linear quadratic cost that characterizes the control objective.

The implementation of this controller requires a discrete space model together with an observer, which generates online estimation of the states used for feedback control.

Simulation

Before the controller is implemented to the real reactor, its reliability needs to be validated by simulation. The disturbance in feed concentration used for validation is randomly chosen and is shown in Fig. 5 in terms of adiabatic temperature rise. From Fig. 6 one can see that if the reactor is not controlled, the hot-spot temperature will either be higher than the upper limit or lower than the lower limit.

With the use of the high-gain observer, the full states of the system can be estimated

Conclusion and discussion

In this paper, on the basis of a high-gain observer, an LQR control strategy is proposed and investigated to control a reverse flow reactor for VOC combustion. The new key issue of the control is that the temperature is to be confined in an envelop of two limit temperatures, instead of maintaining the temperature at a set point. To achieve this, three phases of operation are identified and two LQR controllers are incorporated in a switching control structure. With the LQR strategy, the

Notation

acspecific solid–fluid surface area, m-1
cpmgfluid heat capacity, Jkg-1K-1
cpssolid heat capacity, Jkg-1K-1
hsolid–fluid heat transfer coefficient, Wm-2K-1
Htotal length of monolith, m
kDsolid–fluid mass transfer coefficient, ms-1
MVOC molecular weight, kgmol-1
Nnumber of transfer units for heat loss, dimensionless
PPeclet number for solid–fluid heat transfer, dimensionless
Paxaxial Peclet number for heat conduction, dimensionless
PθP corrected for the finite frequency, dimensionless
Qjexternal power

References (14)

There are more references available in the full text version of this article.

Cited by (29)

  • Error-feedback temperature regulation for a reverse flow reactor driven by a distributed parameter exosystem

    2022, Journal of Process Control
    Citation Excerpt :

    Moreover, feedback control of infinite-dimensional systems has attracted lots of attention and efforts (see e.g [12,22,11]) and variety of techniques have been adopted for many types of PDEs systems, including hyperbolic and parabolic PDEs (see e.g. [23,14]). In [24], state-feedback linear–quadratic regulator has been designed for VOC combustion in a CFRR by manipulating internal electric heating and dilution as inputs to control the hot spot temperature. The problem has been investigated by implementing finite-difference discretization method of the linearized plant to regulate the temperature at a single position.

  • Control of regenerative catalytic oxidizers used in coal mine ventilation air methane exploitation

    2020, Process Safety and Environmental Protection
    Citation Excerpt :

    Energy can be added to the reactor with an internal or external heater, using electricity or hot fluids, or by adding a fuel to the reactor feed (Fissore and Barresi, 2008; Hua et al., 1998). Reactor temperature can be decreased by extracting energy from the reactor with an internal or external refrigeration system, by diluting the feed with air (at the reactor inlet or in the middle point), or by by-passing part of the cold feed to the reactor centre (thus, this fraction of the feed is not pre-heated in the hot first part of the bed) (Balaji et al., 2007; Edouard et al., 2005b). Other possibility consists of withdrawing part of the hot gas, generally in the central part of the reactor (Fuxman et al., 2007; Li et al., 2013a, b; Li et al., 2017).

  • Demonstration of mitigation and utilization of ventilation air methane in a pilot scale catalytic reverse flow reactor

    2017, Fuel Processing Technology
    Citation Excerpt :

    Even though RFR has advantage of auto-thermal operation, it is hard to operate due to its hybrid and forced unsteady-state nature as well as the extensive variation in the methane concentration and flow rate [17,19,20]. A great deal of effort has been made on the control of the RFR recently, including feed forward control [21], model predictive control (MPC) [19,20,22–25], linear quadratic regulator (LQR) control [26,27], robust control [28], logic control [5,29–34] and fuzzy logic control [35]. Besides operating the RFR auto-thermally, recovering part of heat is also very tempting due to the huge emission of VAM.

View all citing articles on Scopus
View full text