Building Automation and Control Systems: A Review

the main reason for which was detection of several PDES shortcomings, including static configuration between events, the execution of events sequentially, and difficulty in understanding the system.

3.5.2. Parallel Virtual Machine

Parallel virtual machine (PVM) is a software package that allows a heterogeneous set of computers linked together by a network to be used as a single large processor. This very portable software can be classified as a protocol for efficient parallel computing rather than as a simulation-specific protocol.

PVM was developed in 1989 as part of a collaborative research project among the University of Tennessee, Oak Ridge National Laboratory, and Emory University (see Geist et al., 1994). The PVM concept is based on a virtual machine in which a number of computers running on a heterogeneous network act as one large parallel computer. To achieve performance as a virtual machine, applications can be networked together via the APIs defined by PVM libraries.

PVM is still undergoing improvements, and several of its versions are sometimes unstable on certain operating systems. Using PVM to enable interoperability among heterogeneous platforms results in lower performance in comparison to using message passing interface (MPI) because PVM requires that every message be tested to determine the architecture of its sender. PVM attempts to maintain interoperability among programming languages in distributed computing environments (e.g., the ability to communicate between programs written in C/C++ and others written in FORTRAN).

3.5.3. Message Passing Interface

Message passing interface (MPI) is a specification for enabling communication by message passing among many computers. In contrast to PVM, MPI was specified by the 1993-1994 MPI Forum with the intention of unifying the most vendor-specific APIs that exist for different parallel computing systems. When an application requires using routines from several different parallel libraries, most MPI implementations use platform-specific optimisations for different assumptions. This has the negative effect of ensuring that no MPI implementation for a single computer platform can communicate with a different platform. Even though MPI libraries are very portable (i.e., they can be installed on any platform), they lack the ability to achieve interoperability among different platforms (see Geist et al., 1996). Although MPI does not provide for interoperability between languages, it maintains language independent specification (LIS) for function calls and language bindings.

Although MPI is a high-performance message-passing interface that has application to certain areas, it does not provide the interoperability and fault tolerance required for a conceptually interoperable model. In response to this limitation, a project termed PVMPI (parallel virtual machine passing interface) has been initiated to create a programming environment that can access both MPI and PVM libraries with a common software package.

3.6. Building Automation and Control Systems: A Review

Recent developments in BACS architecture indicate that data integration and interchange plays an essential role in ensuring interoperability among equipment and components developed with different network technologies. BACS architecture is an explicitly integrated technology that provides automatic control of different systems used in large functional buildings, such as HVAC&R equipment, lighting components, electrical services, and other facilities. BACS architecture enables automation of the operation of indoor building processes (or variables) in order to provide a comfortable environment for occupants and realise significant savings in energy use. In consequence, modern BACS architecture must imply the integration of advanced control systems to facilitate the development of self-adapting processes for the indoor environment, self-learning functions from the occupants, and self-organising equipment and components along with their emergent actions by using physical abstractions.

For most recent applications, BACS architecture takes the form of a DCS using one or more open network protocols, as shown in Figure 2.6. A typical DCS consists of various device components (sensors and actuators), control systems, and a network interconnecting them. As BACS architecture has become technologically advanced (i.e., well established and more sophisticated by offering a vast diversity of control functions for all integrated equipment and components that operate within the building environment), several network protocols (e.g., BACnet, LonWorks, and Modbus) have been

developed using different technologies for better exploitation of the architecture. These networking protocols generally exploit different formats of object-oriented data structure that BACS architecture must integrate to communicate between all building facilities (i.e., HVAC&R equipment, lighting components, computers, and so on), as shown in Figure 3.3. Indeed, the idea of shared technologies is one solution that may enable communication among the components of different technologies that operate within a building (e.g., BACS architecture can use both BACnet and LonWorks protocols in the same installation). To exploit BACnet and LonWorks components in the same BACS architecture, it is necessary to have a gateway or router in order to translate between BACnet object properties into LonWorks network variables. Yahiaoui et al. (2007) described an approach using services or web-based XML and SOAP applications in order to provide great flexibility in communication among components and devices and through various network protocols designed with different technologies.

Figure 3.3. A modern BACS architectural solution

As described by van Paassen (1986) and Yahiaoui et al. (2005b), a computer-based control system is much more flexible than are traditional control systems because the software installed on the computer allows adaptation of the control strategy to changing conditions. Specifically, the central computer defines the control loops suitable for activation and the appropriate control setpoints for the current situation. This computer is connected to substations that constitute the automation level that collects messages from sensors and regulates actuators via a network. To enable communication, a protocol is used to exchange data between the computer and building equipment and components. A message typically includes the address of the component and the series that refers to the real physical quantity (e.g., the desired position of a valve). It is important to consider the use of advanced control systems within the BACS architecture during the early design phase. In doing so, distributed simulations are then required to study the impact of advanced control systems on building performance applications.

3.6.1. Advanced control systems

The design of advanced control systems in ABs requires the performance of trade-off studies to analyse various configuration options and different strategies possible. The importance of performing M&S studies early in and throughout the design and development phase has been demonstrated many times. The major benefit of integrating M&S into the building development process is a significant reduction in total cost of ownership. As most current BPS software tools do not have a flexible way of dealing with advanced control strategies, there is a need to improve BPS in order to be able to study the impact of advanced control systems and strategies on building performance applications.

The integration of advanced control systems in building environments is required in order to constantly maintain indoor building processes at or between the desired preferences (or references) of occupants

while minimizing the energy consumption in buildings. In general, the choice of control systems often depends on the task under consideration. Although it is relatively easy to attune and maintain simple controls, managing several indoor building processes altogether requires control by more sophisticated algorithms. Currently, the higher-level tasks of optimisation and supervision are typically carried out by human operators. With the advent of modern technology and advances in the field of intelligent control systems, these higher-level tasks can become automated (see e.g. Meystel et al., 2002). In particular, the installation, operation, and integrity of modern control systems can be supervised by higher-level automation. As the main challenge in BACS architecture concerns with automation and operational integrity of various building equipment and components in an efficient and manageable fashion, a computer-based control system is used and desirable because it is much more flexible and allows the use of more complex control algorithms. Yahiaoui et al. (2003) provided a comprehensive view of the hierarchical layers (multi level multi-objective) in integrated modern control analyses. A more complete understanding of these hierarchical layers in control systems is provided in Figure 3.4.

Figure 3.4. The practical implementation of hierarchical layers in control systems These hierarchical layers are classified as follows:

 Regulation (or direct control) occurs when a control forces a building process and/or plant to behave in a desired way in order to achieve certain control setpoints.

 Optimisation (plus monitoring and supervision) occurs when a control needs to determine the optimal setpoints of the regulators; optimise (minimise or maximise) defined performance criteria in order to attain certain objectives regarding state operation, energy efficiency, and time response; and monitor and supervise control operations at multiple remote locations from a central computer.

 Adaptation occurs when a control directly or indirectly adapts the variations in coefficients describing the model of a system and the variables in equations describing the control laws so that the control responses follow the references more closely, even in the presence of errors.

 Self-learning and organising occurs when a control is based on the choice of structure of the model and the control as a function of the changing environment. A typical example of such a control system is a multi-control architecture that can be used to switch among alternative control designs to adapt the control operations to the present situation.

The relevant literature on control, monitoring, supervision, and optimisation techniques is extensive, with many papers exhorting a certain solution to a particular problem. However, it is generally acknowledged that there is currently no one technique that will resolve all the control problems that can manifest in modern practice. Therefore, a flexible approach that will allow easy incorporation of alternative strategies is needed. Yahiaoui et al. (2006g) described one such flexible approach to BACS architecture using hybrid systems to model large-scale systems typically arising in MASs.

3.6.2. Distributed Control Systems

DCS architecture is composed of several interconnected devices that integrate a central computer, communication networks, control programs, sensors and actuators into different levels of abstraction, as shown in Figure 2.6. DCS architecture is widely used in a number of industries, such as building and home automation, manufacturing systems; intelligent vehicle systems; and advanced aircraft and spacecraft systems. In the literature, two important terms often used to refer to DCS architecture are integrated communication and control systems (ICCSs) (Ray, 1989; Wittenmark et al., 1995) and networked control systems (NCSs) (Nilsson, 1998; Walsh et al., 1999; Branicky et al., 2000; Lian et al., 2001), although the latter term is more commonly used. NCS architecture consists of control systems within which computer-based controls, sensors, actuators, and other system components exchange data through communication networks. Usually, computer-based control systems are carried out in two modes, supervisory control mode (SCM) or direct digital control mode (DDCM or DCC), as shown in Figure 3.5. In SCM, the computer sends signals to establish setpoints and remotely activate or deactivate local controls. In contrast, in DDC the computer can define suitable control loops for activation and appropriate setpoints for the current situation. Both modes are widely used in industrial applications, especially in third-generation ABs, because they allow and facilitate integration of control technologies. DCC is currently much more widely used because it uses common software – for control systems design – installed on a computer instead of hardware embedded on local controllers. This computer can be as simple as an inexpensive PC or as sophisticated as a large scale computer.

Figure 3.5. Different modes of a computer based control systems

DCS architecture is a very complex research field because it involves at least two areas of focus concerned with communication protocols and control system design. For example, Lian et al. (2002) explained that proper message transmission is necessary to guarantee network quality of service (QoS), whereas advanced control design is desirable for guaranteeing quality of performance (QoP). DCS architecture offers several advantages for networked control applications, including improved system modularity, efficiency, dependability, and flexibility (e.g., distributed processing and interoperability);

features that ease installation and reconfiguration (e.g., only a small number of wired connections is required); and powerful system diagnosis and maintenance utilities (e.g., automated alarm handling).

However, the use of network in DCS architecture makes the analysis and design of control systems complex because the network-induced time delay makes the traditional study of time delay systems different. Regarding this fact, Lelevé et al. (2000) pointed out that because the communication delay in DCS architecture is considered either constant or varying in time interval, conventional control theories with many ideal assumptions, such as synchronised control and non-delayed sensing and actuation, must be re-evaluated before they can be applied to DCS architecture. The further network challenges of packet losses and frame scheduling should be considered during the early design phase of control systems. Halevi and Ray (1990), Zhang et al. (2001) and Juanole and Mouney (2005) attributed certain types of failures that can degrade control system performance or even destabilise the system to the poor design of DCS architecture. Nevertheless, DCS architecture uses a network bus that has several advantages over CCS architecture and point-to-point control systems, such as decreased wiring and maintenance costs, processors that offer faster and better performance than can mainframes, expanded flexibility of control architecture and lack of diagnosis capabilities, enhanced reliability by preserving a part of the system, improved reconfigurability, and increased interchangeability.

With regard to the timing issues of communication and computation time delays, processor idling, and transient errors in DCS architecture, Wittenmark et al. (1995) and Sanfridson (2000) concluded from different case studies that the total maximum network (or communication) delay is less than one sampling period. Bauer et al. (2001) described an approach in which the network-induced time delay is compensated by using a control-based Smith predictor. In this approach, it is assumed that communication between senders and receivers is instantaneous when communication is possible and that the transmission and reception times are known. However, these assumptions are not always met because the application only applies to the delay over the switched Ethernet³⁴. Moreover, the control-based Smith predictor model lacks robustness. In response, Georges et al. (2006) proposed an approach to compensating for the effects of the delay by using a robust control and estimating the network-induced time delay by using network calculus theory. Juanole and Mouney (2005) and Juanole et al.

(2005) established a relationship between the distributed simulation system QoS and the closed loop network based control system QoP. In their studies, the design of distributed system mechanisms and the control systems are performed in a separate manner and communicate through Fieldbuses³⁵. Arzen et al. (1999), Branicky et al. (2002), Sename et al. (2003), Branicky et al. (2003) and Cervin (2003) presented a new approach to NCS co-design by studying the design of control systems and schedules in an integrated manner. In their studies, Fieldbuses were also used for connecting sensors and actuators to I/O control systems. However, the use of Fieldbuses is sometimes problematic because they are very costly and difficult to interface with products using different technologies. To overcome these problems, a computer network technology, especially Ethernet protocol, is used. Lian et al.

(2001) characterised the network delays for the different industrial networks of ControlNet, DeviceNet, and Ethernet and studied the inherent tradeoffs between network bandwidth and control sampling rates.

Based on this, it was concluded that Ethernet can be used for periodic/non-time-critical and large data size communication, such as communication among machines. For control systems with short and/or prioritised messages, DeviceNet provides better performance. When Lee and Lee (2002) evaluated both switched Ethernet and control system performances; he observed that the control performances on the network-based control system using the switched Ethernet were affected very little by the network delay. Consequently, DCS architecture using a switched Ethernet network can still satisfy the desired design specifications. Recently, the development of switched Ethernet has demonstrated that it has very promising prospects for industrial applications due to the elimination of uncertainties (collisions) in the network operation, thus improving the performance and effectiveness of DCS architecture.

In the building domain, very few relevant research studies document in detail the different networking platforms that support data communication for the automation of building HVAC&R equipment, lighting components, and other facilities. Practically no studies have analysed the behaviour of NCS architecture regarding delay and performance issues, even though networking protocols such as BACnet, LonWorks, and Modbus are widely used in the context of ABs. On the other hand, the importance of using building control applications using BACnet or LonWorks as a network protocol has been shown in many studies. For example, ASHRAE (1997) described the evaluation of HVAC&R control algorithms and strategies using computer simulation across a BACnet protocol.

3.6.3. Building Control HVAC&R Equipment and Lighting Components

The automation of building HVAC&R equipment, lighting components, and other facilities requires using advanced control methods (or techniques) to increase building environmental performance and energy efficiency. Moreover, many efforts in the control of building HVAC equipment and lighting components have typically paid on the local level controls (see Goswami, 1986; Rishel, 2003; Moore and Fisher, 2003; Fredrik and Dennis, 2004; Zhang et al., 2005; and Bourgeois et al., 2006). Therefore, there is great interest in the integration of advanced control systems in buildings, especially in BPS for better energy and comfort management. In third-generation ABs, BACSs have replaced so-called hardwired controls with control strategies implemented in software.

34 Switched Ethernet is by definition an Ethernet LAN that uses switches to connect individual hosts or segments (i.e., sections of a network).

35 Fieldbuses are serial communication buses frequently used in industrial applications, such as controller area networks (CANs), Ethernet for Control Automation Technology (EtherCAT), Aeronautical Radio Incorporated (ARINC), FIB, and DeviceNet.

The advent of computer-based control systems has fuelled the investigation of building HVAC&R equipment and lighting components that easily control their functions in an efficient and rational manner while reducing fossil fuel consumption and greenhouse gas emissions (see IEA 2002). Modern control methods offer an efficient means of handling emergency issues in buildings, as a BACS central computer can devise optimal control strategies for specific urgent situations. For example, Lute et al.

(1995) attempted to identify a cost-effective means of supplying heat to a building using a predictor for the indoor temperature while maintaining a comfortable air temperature in the building within a certain range of variation. Such control of temperature in heating or cooling mode is maintained between two predefined limits (or setpoints) instead of maintaining it as long as possible at the particular desired setpoint. In another study, Peng and van Paassen (1998) successfully used a state space model derived from CFD for predicting and regulating control responses to the temperature of indoor air zones.

Regarding control strategies developed for a double skin façade (DSF), Park et al. (2003) proposed the use of optimal control strategies to minimise energy use in buildings with various DSF configurations.

Within their proposal, a number of performance criteria are defined in terms of several parameters, such as predicted mean vote (PMV), transmitted heat, solar radiation and light to inside, cooling and heating load reduction due to the DSF, and louver slat angle (i.e. façade opening for ventilation). The results showed that using DSF facilities may reduce heating demands up to 70% and cooling demands up to 10% when compared to the use of an on/off (or manual) control. Moreover, Stec and van Paasen (2005) described different HVAC&R strategies for single facade and DSF configurations, in which the early design phase of a building includes a number of requirements regarding the indoor parameters related to heating, cooling, airflow, noise reduction, the dimensions of the facade, the HVAC&R optimisation to consider with the facade, and a control system that regulates the entire building

Within their proposal, a number of performance criteria are defined in terms of several parameters, such as predicted mean vote (PMV), transmitted heat, solar radiation and light to inside, cooling and heating load reduction due to the DSF, and louver slat angle (i.e. façade opening for ventilation). The results showed that using DSF facilities may reduce heating demands up to 70% and cooling demands up to 10% when compared to the use of an on/off (or manual) control. Moreover, Stec and van Paasen (2005) described different HVAC&R strategies for single facade and DSF configurations, in which the early design phase of a building includes a number of requirements regarding the indoor parameters related to heating, cooling, airflow, noise reduction, the dimensions of the facade, the HVAC&R optimisation to consider with the facade, and a control system that regulates the entire building