Elsevier

Energy Conversion and Management

Volume 39, Issue 14, 15 September 1998, Pages 1471-1482
Energy Conversion and Management

PI-control of dual duct systems: manual tuning and control loop interaction

https://doi.org/10.1016/S0196-8904(98)00020-XGet rights and content

Abstract

Good HVAC control schemes in buildings help reduce energy use and maintain occupant comfort. To this end, PID (proportional integral derivative) controllers are widely used in commercial buildings to keep variables, such as temperature and pressure, at predefined setpoints. These controllers, when they do not include an autotune feature, must be carefully tuned. Also, interactions occurring among the control loops must be considered during the first fine-tuning phase, so one loop response is not improved to the detriment of another.

This study concentrates on PI (proportional integral) control of dual duct systems. Three tuning methods were selected and tested on a small scale dual duct system with four decentralized control loops. Each controller was tuned according to the best performing set of tuning parameters, and the system was then operated under design conditions to observe the interactions among the control loops. The effects of interactions were examined through a poorly tuned controller in each loop, one controller at a time.

Introduction

The goal of HVAC design in buildings is to provide comfort to the occupants. Because heating and cooling loads vary with the time of the day and of the year, an HVAC system must be complemented with a good control scheme to maintain comfort under any load conditions. Good control will also reduce energy use by keeping the process variables (temperature, pressure etc.) to their setpoint efficiently. The efficiency of a control scheme to maintain comfort largely depends on proper controller tuning. Tuning can be performed manually, automatically with autotuners which have a special feature to be turned on and off by an operator, or automatically with adaptive controllers which are self-tuning. Currently, most commercial and institutional buildings are equipped with PID (proportional integral derivative) controllers which are manually tuned or with autotuners. In these cases, interactions between the different control loops of the system can occur. This results in one loop being tuned to the detriment of another.

Until now, tuning rules were developed by testing them through several simulations or on one specific control loop. Some researchers compared their new set of rules with older ones to show their improved responses and robustness. However, their comparisons were based on one control loop only. Also, the interaction occurring between the numerous control loops of HVAC systems is rarely discussed. One article by Krakow et al. (1995)[7] was found to assess directly a problem of interaction between compressor and evaporator fan speed loops during temperature and relative humidity control. More recently, research has been mainly directed towards the development of adaptive controllers including fuzzy logic controllers and neural-network-based controllers.

A typical local-loop control configuration is shown in Fig. 1. For such control loops, tuning based on experimental tests was first reported by Ziegler and Nichols[12] in 1942. This method is still used in the industry and is said to provide a good basis to obtain the optimum settings. Cohen and Coon (1953)[3] and Pessen (1994)[9] developed tuning rules for cases where Ziegler-Nichols settings are not applicable. Pessen (1994) also proposed new values for the proportional gain, reset rate and derivative time to improve the PID controller-tuning parameters found by Ziegler and Nichols’ method. The sensitivity method, as described by Ziegler and Nichols, requires several closed-loop tests before the critical gains can be defined but Yuwana and Seborg (1982)[11] developed a new on-line method to determine the process parameters needed for tuning. The method requires only one experimental closed-loop test, and once the process parameters are established, they can be input in any tuning rule to determine the controller settings. The simplified IMC-PID tuning rules developed by Fruehaut et al. (1994)[4] are used in the chemicla industry. The rules were established for general, flow and level loops. The researchers showed that the simplified IMC-PID rules are more robust than Ziegler and Nichols’ rules or Cohen and Coon’s rules. Finally, some tuning rules for PI (proportional integral) controllers used with first-order HVAC processes, such as the rules developed by Bekker et al. (1991)[1], rely on a more theoretical background to determine the tuning parameters using techniques such as root locus and pole zero cancellation.

Some other researchers used a laboratory scale system to run tests on control loops. Nesler and Stoecker (1984)[8] analyzed the effect of proportional and integral gains on a heating coil discharge air-temperature loop. Pinnella (1994)[10] studied the control of supply duct static pressure in an HVAC system. He suggested using an integral only controller and found that it eliminates the offset and is easier to tune than the PI controller. Green (1994)[5] worked on a hydronic system in which the usual control valves were replaced by variable-speed pumps. He used Ziegler and Nichols’ open-loop tuning method and the Chien et al. method (1988)[2] to define two sets of parameters for PI-control and selected the Chien et al. method to tune his controller.

In spite of the multivariable and interconnected nature of the HVAC processes, industry still prefers controller tuning at the local-loop level, that is, one loop at a time. This practice is simple and cost effective and, as such, enjoys wider acceptance among engineers. As a result, several tuning methods, as noted above, have been developed for tuning local controllers. These methods also form the basis on which some autotuners and adaptive controllers are designed.

In the context of this background, the current study focuses attention on two important issues: (1) given the fact that HVAC processes consist of both fast and slow dynamics, is there any advantage of using one tuning method over the other for a specific loop? and (2) in a system with multiple local loops, what is the effect of one poorly tuned controller on the performance of the other controllers?

The first issue is related to the appropriate choice of tuning method for a given local loop, and the second issue is related to the consequence of tuning one controller at a time in a multiloop system in which one of the controllers could have been poorly tuned or subjected to a large disturbance. We have addressed these two issues in this paper by:

  • 1.

    Comparing the results obtained from three tuning rules applied on two control loops of a laboratory scale dual duct system.

  • 2.

    Quantifying the interaction occurring among four control loops through simulation of a poorly tuned controller.

Section snippets

System description

The main section of a multizone air handling unit (AHU) with four decentralized control loops was built in the laboratory. Fig. 2 shows a schematic diagram of the AHU. It consists of the supply fan, the two-ducts—hot and cold—the mixing box and the mixed air duct. The cold duct is equipped with a chilled water cooling and dehumidifying coil and two dampers to regulate the air flow. One damper is located in the duct itself while the other is located in a branch to exhaust air from the system.

Manual tuning of the controllers

The three tuning rules covering PI-control selected for this study are: (1) Ziegler and Nichols’[12] tuning rules (closed-loop method only) (1942), (2) simplified IMC-PID tuning rules (Fruehauf, 1994[4]), (3) Bekker et al. tuning rules (1991)[1]. They were tested on the four controllers of the dual duct system. Here, we present the results obtained for the fan speed and the cooling valve controllers. The selection of the most reliable set of tuning parameters for a controller depended on how

Operating the HVAC system under real conditions

With the four control loops individually tuned, the overall system response was studied to determine if the HVAC system could maintain the four process variables at their setpoints efficiently. The system was first operated under the design conditions, and then, the gains of each controller, with the exception of the heater controller, were changed, one controller at a time, to analyze the interaction effects.

Conclusions

PI-control of a laboratory size dual duct system was analyzed. The air handling unit included four decentralized control loops which allowed comparison between three different tuning methods and evaluation of the interactions among the loops.

Three tuning rules were tested on two control loops of the dual duct system. It seems Ziegler–Nichols’ tuning method would be more appropriate for control loops operating in slow time scales, while the simplified IMC-PID tuning method would be more

Acknowledgements

The authors acknowledge the financial support provided by the National Sciences and Engineering Research Council (NSERC) and the Fonds pour la formation de chercheurs et l’aide ála recherche (FCAR-Centre de recherche).

References (12)

  • P.S. Fruehauf et al.

    ISA Trans.

    (1994)
  • J.E. Bekker et al.

    ASHRAE Trans.

    (1991)
  • Chien, , Hrones, and Reswick, , in Controllability Study of Unit Processes of Air Conditioning System, ed. P....
  • G.H. Cohen et al.

    ASME Trans.

    (1953)
  • R.H. Green

    ASHRAE Trans.

    (1994)
  • Jetté, I., PI-Control in Dual Duct Systems: A Study on Manual Tuning and Control Loop Interaction. Master Thesis at...
There are more references available in the full text version of this article.

Cited by (0)

View full text