Self-configuration of building control system using knowledgebase.
Control sequences are also called sequence of operation (Montgomery and McDowall 2008). These narratives play an essential role in the HVAC system controller, conveying the intent and understanding of the designer and experts. A sequence of operation can be decomposed into a set of rules based on black models (empirical), or white models (physical), or gray models (combined). Appropriate configuration of these sequences is critical for maintaining thermal comfort as well as achieving system-wide energy efficiency. The conventional process of configuring a control system involves the interpretation of narrative control sequences and creating control logic from the interpretation. However, this process has a couple of problems.
First, the problem is created when mixed narrative sequence formats were used in design. Ideally, there are two types of formats to organize the control sequence, i.e. by components, and by operating modes. This is defined in Section 184.108.40.206, ASHRAE Guideline 13-2007 (ASHRAE 2007). This guideline suggests formatting should be based on the intended use of the sequence. However, in practice, control sequence designers sometimes copy and paste narrative control sequences from different projects. As a result, the narrative control sequences turn into a mixed format (i.e., parts of the project are organized by components, while other parts are organized by operating mode.). This mixed format increases the difficulty for other personnel to understand the sequence and confounds programming the logic. This format inconsistency and associated interpretation ambiguity, directly affect the appropriate configuration of the control system, and further contribute to energy waste and poor control. A review study by Roth et al. (2005) summarized that 0.012 quads of energy was wasted due to software programing errors, while other 0.38 quads can be saved if better control of HVAC system and lightings when the space is unoccupied.
Second, the problem is related to the lack of standard configuration procedure. Currently, companies, professional organizations, and researchers have developed different procedures. For example, Guideline: Specifying Direct Digital Control Systems from ASHRAE (2007), online application platform, Control Systems Design Guideline from The California State University (Agaram et al. 2005), and other procedures provided in various books (Underwood 1999; Montgomery and McDowall 2008). The variety risks generating differences and uncertainty by following different guidelines.
The current differences between configuration procedures might also create barriers in technology transfer. Advanced control strategies have been developed include model predictive control (Ma et al. 2011), adaptive control (Soyguder et al. 2009), and other advanced algorithm. However, these new methods are seldom included in sequence libraries or design guideline, which implies the chance of actual deployment would be even smaller. This might be due to the lack of configuration infrastructure that can support the deployment of new and advanced control methods.
Therefore, existing control system configuration processes relying on manual interpretation of the narrative sequence needs to be modernized, and the format of narrative sequences needs to be standardized to avoid ambiguity. This study proposes a new approach to self-configure the control system using control system knowledgebase. As an exploratory effort, the following sections cover the design of knowledgebase, the development of the self-configuration process, and an initial demonstration based on simulation.
2 HVAC CONTROL SEQUENCE KNOWLEDGEBASE
The knowledgebase (or knowledge-base) initially evolved from the expert system (Hayes-Roth et al. 1983). Control system knowledgebase proposed in this study is defined as an integrated database that can store, manage and upgrade generalized control sequences for building operations. Generally, the sequence can be classified into basic sequences, and advanced sequences. Basic sequences are developed based on published control sequences in the Library of system control strategies by Martin and Banyard (1998) and. In the knowledgebase, basic sequences will be enhanced continuously by analyzing and summarizing other published books and guidelines. Advanced sequences/control strategies will be incorporated into the knowledgebase as well. This type will be mostly based on published journal papers and other reliable resources. The following sub sections are to further introduce the knowledgebase structure and a prototype development.
2.1 Structure of the knowledgebase
The object-oriented structure is selected for this knowledgebase development. As shown in Figure 1, it means each output is associated with a series of inputs. Each output belongs to an actuator, or system setpoint.
The external characteristic of each module is represented by names, associated operating mode, etc. The internal characteristic is represented by the control sequences and the associated inputs and outputs. Narrative sequences are provided in two formats. The sequence codes are provided for generating the system-wide control codes.
2.2 Primary knowledgebase
Based on the knowledgebase structure described above, a prototype of this knowledgebase was created in spreadsheet form. This prototype includes most of the components that can be found in common types of HVAC systems. Each component has its own sequences, associated operating mode, required inputs and outputs. Each sequence is represented in narrative description (by component and by operating mode), and sequence codes.
As shown in Table 1, an example of mixing box (dampers) illustrates the typical categories in a component module setup. This example indicates that there are four control statuses in mixing box. Each status is associated with a number of inputs and rules (listed in the same column). Control algorithms are represented in narrative description and sequence codes. This example provides an explicit structure and format to describe the control sequence. Meanwhile, the knowledgebase needs to be supported by strong querying capability. Narrative sequences can be arranged and sorted by components, or by operating modes.
3 SELF-CONFIGURATION OF A CONTROL SYSTEM
With the knowledgebase served as a platform, the self-configuration of control system becomes possible. The following sections explain the definition and conceptual process of self-configuration.
3.1 Definition and Concept
The concept of self-configuration was primarily introduced to autonomic computing by Kephart and Chess (2003). It was used to describe when a new component is introduced into an autonomic system, it will automatically learn about the system, and configure itself and modify its behavior accordingly. Akinci et al. (2011) adopted this concept, and identified 10 functional requirements for self-configuring HVAC system. The major requirements include, communication with neighboring systems, generating and evaluation of alternative configurations, using less resources, being able to learn and unlearn, and being modular.
In this study, we focus on control system, and define self-configuration of control system as being able to generate sequence codes based on identifying the HVAC system type and components. Then, it configures the sequence codes into the controllers. That is to say, to realize self-configuration of the control system, it first requires achieving self- generation the appropriate control sequence. However, there is not any self-configuration procedure available in HVAC control system, or even the fundamental framework. Therefore, we framed an overview of the process, shown in Figure 2. The highlighted step illustrates the difference between conventional approach and the self-configuration. The new approach proposed here uses knowledgebase to automatically generate sequence codes and narrative sequences based on acquired information. Conventional approach requires engineers to design the narrative control sequences using expertise and/or a sequence library and then manually interoperate and program the narrative sequence into sequence codes. Thus, the new approach can promote automation in HVAC control design and avoid individual interpretation of the sequence narrative.
3.2 A prototype work flow
To realize the self-configuration, a five-step process is established. Details of each step are illustrated in Figure 3
The first step is to acquire information of the building and system, and from the requirements of client and other parties. Information here can be classified into two types, i.e., mandatory and optional. Mandatory information includes building location, building type, building size, HVAC system type, and component list. Optional information includes (but not limited to) occupant type, schedule, temperature set point preference, building energy models. These two types of information will be identified and prepared as the inputs for the self-configuration process.
The second step is to process the information with the knowledgebase. This includes five sub-steps. The first substep is to identify the major components (sensors, actuators, and equipment) from the knowledgebase. Based on the identified list, if any mandatory component(s) is missing, configuration process will halt, and report a message as "lack of mandatory components". Otherwise, it goes to the second sub-step to identify the list of operating mode. The third sub-step is to identify control strategies. The fourth sub-step is to identify the control sequences. The last sub-step is to identify any optional control strategies.
The third step is to generate sequence codes for controllers at local, coordinating, and building level. The first sub-step is to select an appropriate organization structure of controllers. The second sub-step is to generate sequence codes for local, coordinating and building level controllers accordingly. Those sequences are component-based. The third sub-step is to explore if any optional control strategy is available for this case. The last sub-step is to generate narrative control sequences in two formats (by component and by operating mode).
The fourth step is to package, deliver and implement the configured control sequences. The first sub-step is to assign a case serial number and keep a copy in the knowledgebase under the category of "Cases". The second sub-step is to implement the sequences into local, coordinating, and building level controllers. Functional test will be conducted as the third sub-step to make sure the sequences work as designed.
The fifth step is to upgrade the control during the operation and maintenance, and to activate the adaptation features if embedded in the control sequences. This step is critical for any retrofit or regular system repair and maintenance. The implemented case will interact with the knowledgebase to get recommendations to upgrade control strategies and equipment. Identified applicable strategies will be self-configured and applied in this existing control system with the permission.
3.3 Automation of self-configuration
Automating the self-configuration process through a computer platform is under development. This part is necessary to evaluate the applicability of the proposed knowledgebase structure and self-configuration procedure.
4 INITIAL EFFORTS VIA CASE STUDY
The case study uses a single zone system. Major components include mixing box, heating coil, cooling coil, supply fan, and occupied zone. From the control perspective, the heating coil and cooling coil provide conditioned supply air at designed temperature; the supply fan together with a damper maintaining a certain supply air pressure; the mixing box manages the percentage of recirculation air and fresh air. Thus each component operates on its duty and together achieves the building control. So we followed the self-configuration procedure in Section 3. First, the list of the components is identified. Second, we use this list to sort out the knowledgebase and generate the sequence code for each component. Then, we copy and paste those sequence codes in the virtual controllers in the simulation platform and test the functionality of control sequences.
4.1 Sequences codes generating
The primary result of control sequences are generated for major components, as shown in Figure 4. Control sequences (code) in the knowledgebase are developed in a commercially available software program.
4.2 Simulation and results
As shown in Figure 5, a single zone dynamic simulation was constructed using a commercial software program. It was developed based on the HVAC component library developed by Chen and Treado (2014). This simulation case serves as a test bed to test the functionality of the sequence code generated from the knowledgebase. Due to the limited number of sequence codes available in the knowledgebase, PI controller modules were used for temporally representing unavailable PI control codes.
These control codes were tested in summer (cooling season) and winter (heating season). The simulation results were shown in Figure 6. In summer condition, as shown in Figure 6 (a), the control sequences modify the zone temperature setpoint based on schedule, and modulate the flow rate of cooling coil water to maintain the setpoint. In winter condition, as shown in Figure 6 (b), the control sequences modify the zone temperature setpoint based on schedule, and modulate the flow rate of heating coil water to maintain zone temperature. This indicates the control sequence code is functioning appropriately.
In summary, this study establishes the control system knowledgebase and develops process of control system self-configuration. Given examples of the knowledgebase and the case study demonstrate the layouts of modular components in the knowledgebase and the framework of self-configuration process. Primary results indicate that this simple and straightforward process would potentially facilitate the control system configuration and standardize the knowledge of operation towards smarter and more efficient system operation.
However, several issues were also identified. First, in the narrative sequence library, the PI controller parameters are not provided. This usually requires field tuning of the parameters. To realize self-configuration, auto-tuning PI controller must be programmed in the sequence codes in the knowledgebase. Second, optional advanced control strategies need to be included in the knowledgebase accordingly.
This study is currently in progress from following perspectives,
* Transfer the knowledgebase from spreadsheet to knowledgebase software.
* Add logic flow chart for the sequences.
* Develop an executive software program to automate the self-configuration process.
* Develop graphic interface display.
The authors would like to thank the financial support from the Department of Energy (DOE) through the Energy Efficient Building (EEB) Hub graduate student assistantship.
Agaram, J., D. Effenberger, et al. (2005). Control Systems Design Guideline. Long Beach, California, The California State University.
Akinci, B., J. H. Garrett, et al. (2011). Identification of Functional Requirements and Possible Approaches for Self- Configuring Intelligent Building Systems. Gaithersburg, MD 20899-8631, NIST Grant 60NANB8D8140.
ASHRAE (2007). ASHRAE Guideline: Specifying Direct Digital Control Systems, American Society of heating Refrigerating and Air-Conditioning Engineers.
Chen, Y. and S. Treado (2014). "Development of a simulation platform based on dynamic models for HVAC control analysis." Energy and Buildings 68, Part A(0): 376-386.
Hayes-Roth, F., D. A. Waterman, et al. (1983). Building expert systems, Addison-Wesley.
Kephart, J. O. and D. M. Chess (2003). "The vision of autonomic computing." Computer 36(1): 41-50.
Ma, Y., G. Anderson, et al. (2011). A distributed predictive control approach to building temperature regulation. American Control Conference (ACC), 2011.
Martin, A. J. and C. P. Banyard (1998). Library of system control strategies. Bracknell, UK, Building Services Research and Information Association.
Montgomery, R. and R. McDowall (2008). Chapter 9--Control Diagrams and Sequences. Fundamentals of HVAC Control Systems. Oxford, Elsevier: 216-249.
Roth, K. W., D. Westphalen, et al. (2005). Energy Impact of Commercial Building Controls and Performance Diagnostics: Market Characterization, Energy Impact of Building Faults and Energy Savings Potential. J. R. (DOE). Cambridge, MA, TIAX LLC.
Soyguder, S., M. Karakose, et al. (2009). "Design and simulation of self-tuning PID-type fuzzy adaptive control for an expert HVAC system." Expert Systems with Applications 36(3, Part 1): 4566-4573.
Underwood, C. P. (1999). HVAC Control System: Modelling, Analysis and Design. New York, E & FN Spon.
Student Member ASHRAE
Stephen Treado, PhD, PE
Mr. Yan Chen is a PhD candidate in the Department of Architectural Engineering at The Pennsylvania State University, University Park.
Dr. Stephen Treado is an Associate Professor of Architectural Engineering at The Pennsylvania State University, University Park.
Table 1 Prototype module (mixing box) in the knowledgebase Name Mixing Box (EA, OA, RA dampers) Output EA, OA, RA dampers positions Associated optimum low return optimum night operating start water cooling cooling mode(s) heating, low temperature, start zone low outside temperature, air temperature Value Fully Maintain Fully outs de air (EA, recirculation current OA full op en, RA full (OA & EA status close) RA full close, open) Input 1 BMS BMS BMS BMS Value or optimum low return optimum night Rule 1 start water cooling cooling heating, low temperature, start zone low outside temperature, air temperature Input 2 Fan status -- T_amb T_amb Value or failure -- T_amb < T_amb < Rule 2 T_z T_z-2 [degrees]C Logic 1 OR -- AND AND Logic 2 -- -- OR Narrative * Fully recirculation position (OA&EA close, RA sequence full open): when AHU is not operating, or when BMS (component) in (fan overrun, boost, optimum start heating, low zone temperature condition), or fan failure signal is ON. * No Action (maintain current status): when in low return water temperature, or low outside temperature. * Full outside air (OA, EA full open, RA full close): when BMS in optimum cooling start and T_amb<T_zone, or when BMS in night cooling operation and T_amb < T_z-2[degrees]C. * Temperature control (PI control OA, EA, RA, modulate for supply air temperature setpoint, for minimum OA position setpoint (0.6[m.sup.3]/s), for schedule to fan speed control): when BMS in normal operation. * Interlocks: OA EA damper proven signal interlock with fan operation Narrative When BMS in When BMS is When BMS in sequence (fan in low optimum cooling (operating overrun, return water start and mode) boost, temperature, T_amb<T_zone, optimum or low or when BMS in start outside night cooling heating, low temperature, operation and zone should T_amb < T_z- temperature maintain 2[degrees]C., condition), previous mixing box or fan mixing should be at failure dampers full outside signal is positions. air position ON, mixing (OA, EA full box should open, RA full be at fully close) recirculation position (OA&EA close, RA full open) Sequence function [d_EA, d_RA, if SHU==1 || LZA==1 code d_OA]= MixingBox || OSH==1 || F_fail==1 (SHU, LTA, LTW, LZA, d EA=0.01; OSH, OSC, NCO, d_RA= 1; F_fail, T_amb, T_z, d OA=0.01; d_EA_0, d_RA_0, d_OA_0) elseif LTW= = 1 || LTA==1 d_EA=d_EA_0; d_RA=d_RA_0; d_OA=d_OA_0; Name Mixing Box (EA, OA, RA dampers) Output Associated normal operation operating mode(s) Value Modulation Input 1 BMS Value or normal Rule 1 operation Input 2 T_sa Value or -- Rule 2 Logic 1 -- Logic 2 -- Narrative * Fully recirculation position (OA&EA close, RA sequence full open): when AHU is not operating, or when BMS (component) in (fan overrun, boost, optimum start heating, low zone temperature condition), or fan failure signal is ON. * No Action (maintain current status): when in low return water temperature, or low outside temperature. * Full outside air (OA, EA full open, RA full close): when BMS in optimum cooling start and T_amb<T_zone, or when BMS in night cooling operation and T_amb < T_z-2[degrees]C. * Temperature control (PI control OA, EA, RA, modulate for supply air temperature setpoint, for minimum OA position setpoint (0.6[m.sup.3]/s), for schedule to fan speed control): when BMS in normal operation. * Interlocks: OA EA damper proven signal interlock with fan operation Narrative When BMS in sequence normal (operating operation, OA mode) EA damper proven signal (interlock with fan operation), PI control loop should control OA, EA, RA positions to maintain supply air temperature setpoint (for minimum OA position setpoint (0.6[m.sup.3]/ s), for schedule to fan speed control) Sequence elseif (OSC==1 &&T_amb<T_z) || code (NCO==1 &&T_amb<T_z-2) d EA=1; d_RA=0.01; d OA=1; else % PID control code under development d EA=PID; d RA=PID; d OA=PID; end
|Printer friendly Cite/link Email Feedback|
|Author:||Chen, Yan; Treado, Stephen|
|Date:||Jul 1, 2014|
|Previous Article:||Optimal model-based control of chiller tower fan and cooling water pump.|
|Next Article:||Principal component analysis for building combined cooling heating and power application potential.|