Rule-based approaches for energy savings in an ambient intelligence environment
Introduction
Contemporary information systems are rapidly undergoing various transformations, influenced by a multitude of emerging trends. One of these trends is Green Computing and the rise of the Smart Grid. Global warming, CO2 emissions and energy waste have directed computing towards greener solutions on both hardware and software levels. At the same time, motivated by renewable energy sources and consumer-produced energy schemes, the idea of the Smart Grid refers to an infrastructure encouraging policies towards marketing and distribution of energy between producers, consumers, “prosumers” (i.e. producers and consumers) and brokers [1]. Conclusively, the concepts of sustainability and energy saving have penetrated all aspects of everyday life, including computing.
Meanwhile, progress in microcontrollers, board computers and–particularly–smartphones brings forward a novel computing paradigm, called ubiquitous computing (ubiComp), where users are being surrounded by powerful, portable computing devices [2]. To take things even further, in pervasive computing (perComp), multiple non-intrusive computing devices are scattered across the environment, attracting little or none of the user’s attention. The vision of Ambient Intelligence (AmI) enriches perComp with Artificial Intelligence (AI) methodologies, aiming at increasing comfort and reducing intrusiveness for the user [2]. Towards this affair, intelligent behaviors, deriving from various areas of AI, such as Planning, Semantics and Reasoning, are typically employed for providing smart automations and intuitive human–computer interactions in such environments.
Unsurprisingly, AmI is heavily based on the widespread usage, performance and affordability of microcontrollers and wireless sensor networks, providing the essential environmental input data required by AI algorithms. Likewise, actions to be performed after the decision making process have to go through actuators in the environment in the form of motors, switches or other endpoints. Since both Green Computing and AmI are facilitated by the increased popularity and low cost of smart meters, environmental sensors, actuators and other devices, these features constitute the common grounds among the two disciplines. Also, both fields share similar goals, i.e. the vision of a comfortable but energy-efficient smart home.
Although AmI domains of application may vary widely, targeting energy savings has yet been only marginally explored. Existing service-oriented approaches in conjunction with smart home equipment have not practically considered intelligence or savings, but rather the issue of communication protocol unification [3]. Multi-agent systems have been employed to provide enhanced functionality or service composition, but certain issues on truly interoperable semantic descriptions still persist [4]. A categorization of AmI application domains in Smart Homes, Health, Transportation, Emergency, Education and Workplace, can be found in [5]. Apparently, most approaches have ignored the aspect of energy saving and are instead focused on multimedia technologies, providing context awareness in smart homes or offices based on UPnP/DLNA multimedia streaming [6].
This work presents a holistic approach to energy savings in a smart building belonging to a Greek Public University complex. The proposed approach lies in the intersection of many disciplines (AmI, Automated Reasoning, Semantic Web Services and Green Computing), facing challenges in each of the corresponding fields. More specifically, a set of state-of-the-art wireless sensors, actuators and smart meters has been deployed at the University premises. Heterogeneous sensor data and actuator functionality are both unified under a universal middleware based on the Service Oriented Architecture (SOA) [7] paradigm and the Web Service Description Language1 (WSDL) for syntactic interoperability. The middleware also complies with some of the latest Semantic Web [8] technologies, like Semantic Annotations for WSDL–SAWSDL [9], to provide semantic interoperability on the application layer.
Towards achieving energy savings, two complementary and mutually exclusive rule-based approaches are proposed. Firstly, using production rules, domain experts are able to commit energy-saving policies maintained and applied by an autonomous reactive software agent. The second approach is, instead, based on a deliberative agent that incorporates a defeasible logic reasoner, greatly enhancing the expressiveness of the committed rule set. The application of defeasible reasoning [10] that offers a sophisticated conflict resolution mechanism is ideal in such environments with incomplete and contradictory information; in our approach three clusters of rules are defined: for preferences, maintenance and emergencies. Using the underlying infrastructure, the University’s yearly consumption has been observed to identify energy behavioral patterns. Consequently, energy-saving and comfort-providing policies have been authored and applied accordingly, showing promising results. Experimental evaluation showed that, although daily savings of more than 4% may seem trivial in such a large infrastructure, the policies do ensure that no energy is wasted in any undesired manner.
The next section performs a thorough survey of state-of-the-art in the fields of context-aware applications, service-oriented middleware and the limited application of defeasible reasoning in AmI environments. Section 3 presents the basic principles of defeasible logics, its main characteristics, its syntax and operational semantics as well as the motivation for applying defeasible reasoning in AmI and Smart Building management. The next section presents the architecture of the proposed system in three parts: hardware, middleware and application layer. The two proposed software agents that reside on the application layer are presented subsequently, followed by a real-world application use case scenario, a proposed policy and its experimental results in energy savings. Critical discussion, limitations, future work and conclusions are listed in the final sections.
Section snippets
Related work
The proposed system is built on a synergy of fields such as Ambient Intelligence, Smart Homes, Context-aware Systems, Green Computing, Semantics and Logic. Thus, this section attempts to present the most relevant of existing state-of-the-art, originating from any of the above viewpoints. It also tries to appoint the novelty of the proposed system to each respective work coming from one or more fields.
Defeasible logics
This section describes the basic characteristics of defeasible logic, its syntax and operational semantics as well as the underlying motivation for applying defeasible reasoning in the context of Ambient Intelligence and Smart Building management.
Implementation
The framework proposed in this work is applied at a Greek State University, namely Building A of the International Hellenic University (IHU). The rule-based system is a component of a wider AmI system, aimed to provide automations, comfort and savings, named the Smart IHU project.3 This section presents the overall architecture of the proposed system within Smart IHU and describes each of its components in detail.
Use case: International Hellenic University Building A
As already stated, the proposed framework is applied at the International Hellenic University (IHU) Building A; this section demonstrates the deployment of the policies and their impact on the overall energy consumption. The specific building is a large, appliance-rich building that bears a restrictive and wasteful infrastructure. Thus, it can serve as a stimulating test-bed of our methodology, which can then be applied to other buildings as well.
The percentage of energy savings accomplished by
Critical discussion and future work
This subsection discusses the most significant limitations of the proposed framework and suggests possible ways to tackle each one of them, giving at the same time appealing insights into directions for future work.
The previous sections demonstrated the features of the two distinct agent types (reactive and deliberative), presented their respective pros and cons and discussed their role inside the Smart IHU AmI architecture. To recap, the reactive agent is based on production rules and
Conclusions
This work presents a building-wide real-world rule-based Ambient Intelligence application that aims for user comfort and energy savings. Multiple wireless sensor and actuator networks that comply with different product families and communication protocols have been deployed to collect environmental and energy data across a State University building. The infrastructure integrates a middleware providing service-orientation by unifying all underlying distributed and heterogeneous wireless sensor
Acknowledgments
The Smart IHU project is funded by Operational Program Education and Lifelong Learning, OPS200056 (International Hellenic University, Thessaloniki, Greece). The authors would like to thank the anonymous reviewers for their valuable comments and suggestions to improve the quality of this paper.
References (46)
- et al.
Ambient intelligence: technologies, applications, and opportunities
Pervasive Mobile Comput.
(2009) Developing an energy-saving and case-based reasoning information agent with web service and ontology techniques
Expert Syst. Appl.
(2013)- et al.
Sensor9k: a testbed for designing and experimenting with WSN-based ambient intelligence applications
Pervasive Mobile Comput.
(2012) - et al.
aWESoME: A web service middleware for ambient intelligence
Expert Syst. Appl.
(2013) The path of the smart grid
Power Energy Mag.
(2010)Some computer science issues in ubiquitous computing
Commun. ACM
(1993)- et al.
Home SOA—facing protocol heterogeneity in pervasive applications
- M. Vallée, F. Ramparany, L. Vercouter, A multi-agent system for dynamic service composition in ambient intelligence...
- et al.
MEDUSA: middleware for end-user composition
- M.P. Papazoglou, Service-oriented computing: concepts, characteristics and directions, in: Proc. Fourth Int. Conf. on...
The semantic web
SAWSDL: semantic annotations for WSDL and XML schema
Internet Comput.
Defeasible logic
A survey of service composition in ambient intelligence environments
Towards a rule-based approach for context-aware applications
A rule-based approach towards context-aware user notification services
SESAME-S: semantic smart home system for energy efficiency
Inf. Spektrum
OWLIM—a pragmatic semantic repository for OWL
Cloud-enabled adaptive activity-aware energy-saving system in a dynamic environment
Discovery of Semantic Web Services Compositions based on SAWSDL Annotations
Middleware architecture for ambient intelligence in the networked home
Cited by (34)
Modelling building HVAC control strategies using a deep reinforcement learning approach
2024, Energy and BuildingsUniversal workflow of artificial intelligence for energy saving
2022, Energy ReportsSmart-valve-assisted model-free predictive control system for chiller plants
2021, Energy and BuildingsCitation Excerpt :The third type of AI algorithm is the rule-based system (RBS). In 2015, Stavropoulus et al. reported a RBS for energy saving in an ambient intelligence environment [8]. The RBS comprised a hardware layer, middle (service) layer, and application layer.
Computational sustainability and the PHESS platform: Using affective computing as social indicators
2019, Future Generation Computer SystemsCitation Excerpt :In order to address this problem, automatic systems are being proposed in the literature to control environment attributes and, as a consequence, modify user behavior to act according to sustainable policies. For instance, in [17], a rule based system is proposed to this end, where rules control an environment attributes such as temperature through middleware which results in a stated 4% daily savings. This example is not a unique case, in fact, in [18] a survey of rule based system was conducted supporting the strategy to control user routines through intelligent systems that reduce the occurrence of suboptimal behaviors.
Control problems in building energy retrofit and maintenance planning
2017, Annual Reviews in ControlCitation Excerpt :The power electronics layer involves energy optimization that focus on the maintaining and improving the power quality (Abo-Al-Ez, Elaiw, & Xia, 2014; Esmaeli, 2016; Liu, Zhang, Wang, & Wang, 2014; Mokgonyana, Zhang, Zhang, & Xia, 2016; Nikkhajoei & Lasseter, 2009; Sao & Lehn, 2005; Wilson, Robinett, Weaver, Byrne, & Young, 2016; Yu, Khambadkone, Wang, & Terence, 2010), which is essential to guarantee the performances of all electricity consuming components in the building. The smart appliance layer improves the building energy efficiency by bringing in energy efficiency intelligence to the appliances in addition to the built-in control logic (Arens, Federspiel, Wang, & Huizenga, 2005; Bijker, Xia, & Zhang, 2009; Catherine, Wheeler, Wilkinson, & de Jager, 2012; Mei, Zhu, & Xia, 2015a; 2015b; Portmess & Tower, 2015; Setlhaolo & Xia, 2015; 2016; Setlhaolo, Xia, & Zhang, 2014; Stavropoulos et al., 2015; Stavropoulos, Koutitas, Vrakas, Kontopoulos, & Vlahavas, 2016; Wang, Zhang, & Xia, 2013). The energy flow layer focuses on the energy efficiency opportunities from balancing different energy sources (Elaiw, Xia, & Shehata, 2012; 2013; Ntsaluba, Zhu, & Xia, 2016; Nwulu & Xia, 2015a; 2015b; 2015c; 2017; Sichilalu, Tazvinga, & Xia, 2016; Sichilalu & Xia, 2015a; 2015b; Tazvinga, Xia, & Zhang, 2013; Tazvinga, Zhu, & Xia, 2014; 2015; Wu, Tazvinga, & Xia, 2015a; Xia & Elaiw, 2010; Xia, Zhang, & Elaiw, 2009; 2011; Zhang & Xia, 2011; Zhu, Tazvinga, & Xia, 2015).
DemaWare2: Integrating sensors, multimedia and semantic analysis for the ambient care of dementia
2017, Pervasive and Mobile ComputingCitation Excerpt :The aWESoME-S middleware [9] integrates sensors and actuators together with knowledge representation based on semantic service descriptions. Rule-based agents and defeasible logics automatically manage a smart building environment, with an aim to reduce energy consumption while ensuring comfort [10]. Other middleware, such as AIM [11] and Hydra [4], similarly integrate sensors and actuators using various formalities for semantic service descriptions.