Semantic Decision Support Models for Energy Efficiency in Smart-Metered Homes Yan Tang
Ioana G. Ciuciu
VUB STARLab, Department of Computer Science 10G731, Vrije Universiteit Brussel Brussel, Belgium
[email protected]
VUB STARLab, Department of Computer Science 10G731, Vrije Universiteit Brussel Brussel, Belgium
[email protected]
Abstract— A promising approach to Smart Energy Grids is to empower communities of consumers with a novel role in the management of their electricity by sharing excess electricity and therefore becoming energy producers (prosumers). We achieve it using a framework to connect dynamic, contextaware, heterogeneous virtual and real entities on the Internet of Smart Meters (IoSM). We transform the smart electricity meters into fully-fledged intelligent computers on the IoSM and enable them to securely collect data from heterogeneous meters and sensors, detect smart meters with similar goals, exchange and aggregate data from multiple autonomous physical or virtual meters, and manage the actual energy demand and ensure the achievement of demand response for the community involved. Domain ontologies are used to store the semantics of the collected data from IoSM. A problem we have encountered is how our framework can use ontologies for community-grounded decision rules, which need to be consistent based on aggregation of collected data and contexts. We use Semantic Decision Table (SDT), which is a decision table enhanced with ontology technologies, to tackle the problem. This paper records our efforts in 1) the general design of the framework; 2) decision support models, namely SDTs, for fulfilling this requirement; 3) an algorithm of creating invocational SDTs; 4) the implementation of visualizing invocational SDTs. Keywords-domain ontologies, smart meters, decision table, group decision making
I.
INTRODUCTION
Smart meters are advanced meters, which are able to collect data on consumers’ electricity usage, communicate this data to other power system participants on local utility grid via wireless, and get pricing information from Distribution Operator Systems (DSOs) to stop/start household appliances. When we combine smart meters with home management systems, such as Google Power Meter1, we are able to solve the problem of steadily increasing electricity usage. For example, studies in [2] have shown that access to feedback from home management systems could enable consumers to save between 5 and 15% on their monthly bill. By having such home management systems is not enough. The community of users may also potentially offer energy to the grid at an optimized price and conditions. An initial approach is thus to empower communities of 1
http://www.google.com/powermeter/about/
consumers with a novel role in the management of their electricity through home management systems by sharing excess electricity and therefore becoming energy producers (prosumers). For example, a user would like to sell the energy generated by his solar panel when leaving on vacation. This task can be automated by the user, and expressed in terms of rules, initially in natural language and then transformed in machine understandable format using our framework. The user’s energy offer will be available on a community-based web portal, through which other members can buy/bid on it. This paper will show the system design behind such social web portal. A key role is represented by the semantic technologies (more precisely, ontologies) and decision support models built above these technologies. We use semantic decision tables, which are decision tables properly annotated with domain ontologies and are able to organize themselves in an elegant way, to represent those decision support models. The paper is organized as follows. Section II is the related work. The general design of the framework is illustrated in Section III. We show our semantic decision models in Section IV. Section V shows the implementation of SDTs. In Section VI, we discuss and conclude. II.
RELATED WORK
A series of initiatives exist at European and worldwide level concerning innovative infrastructures for energy efficiency (e.g. SmartHouse/SmartGrids 2 , BeyWatch 3 , BeAware 4 ), the Internet of Energy (e.g. Smart Watts 5 ), semantic technologies for energy efficiency [16] (e.g. SESAME6 and SESAME-S7), demand response (e.g. DR 38,
2
http://www.smarthouse-smartgrid.eu/ 3 http://www.beywatch.eu/ 4 http://www.energyawareness.eu/beaware/ 5 http://www.smartwatts.de 6 http://sesame.ftw.at/ 7 http://sesame-s.ftw.at 8 http://www.ieso.ca/imoweb/consult/demandresponse.asp
SmartCamp9), or smart solutions for home automation (e.g. EcoGrid10, DiYSE11). The approach in this paper extends the methods and techniques developed by these projects in the following ways: (i) it develops an architecture that will assist the individuals to form communities and influence the way they purchase and sell energy in the Smart Grids platform; and (ii) it extends the interoperability on the Internet of Smart Meters via the use of ontologies, which map user incentives expressed in natural language and modeled with semantic decision tables [13] [14] and ontologies to technical concepts processable by the smart devices. The framework proposed in this study pursues promotion of energy awareness via its community services on the existing social network sites, following up already established practical experiments in this area [8]. In order to produce energy smartly and to consume energy efficiently, a crucial move is to integrate the cyber world and the physical world. This leads to the recent development of Cyber-Physical Systems (CPS) that is simultaneously computational and physical. One of the biggest challenges of CPS research is the real-time, secure, and safe group communication methods, protocols, algorithms in a dynamically changing environment amongst various computing devices and equipment. To undertake this challenge, in this study we have used the Internet-ofThings (IoT) computing paradigm [4]. To fulfill the architecture of Internet of Things, we adopted the ResourceOriented Architecture (ROA) style which was first proposed in the work of Richardson et al. [10], in which four key concepts were defined for a ROA: Resources, URL, Representation, and Links. Jagatheesan and Helal [6] proposed a framework fostering the service discovery amongst hierarchical communities and a set of related protocols built upon the private UDDI nodes – Syndication Matchmaker. The notion of service here can be interpreted as utility in the context of our study. Verma et al. [17] provide a scalable Web services discovery architecture – WSDI – composed of multiple registries, namely community gateways. More importantly, the project MSDI employed and customized the JXTA framework [3] as the peer-to-peer protocol to facilitate the federated cooperative service (i.e. utility information) discovery amongst these distributed ontology-augmented registries (community gateways). The most widely used methods for community based context aware demand forecasting are based on adaptive neural networks and have been proposed by Sestito et al. and by Szkuta et al. in [11][12]. No work has been done for 9
http://arc.gov.au/pdf/LP10_R2/Curtin_University_of_ Technology.pdf 10
http://energinet.dk/EN/FORSKNING/Energinetdkresearch-and-development/EcoGrid/Sider/EU-EcoGridnet.aspx 11
http://dyse.org:8080/pages/viewrecentblogposts.action?key= hometest
demand forecasting at a community level as envisaged in this study. It represents a crucial element for the real-time decision support. In this paper, we will use ontologies and decision tables for the real-time decision support. Decision tables [1] are a means to capture process semantics in a decision support system. It is a table containing an exhaustive set of mutually exclusive decision rules [17]. Compared to other decision tools, such as if-then-else statements, flowcharts, decision trees and Bayesian Networks, decision tables have the following advantages: 1) easy to use and observe; 2) compact presentation; 3) easy to produce and learn etc. [9][14]. A semantic decision table (SDT) is a (set of) decision table(s) properly annotated with a domain ontology(-ies) in order to ensure the completeness and correctness of the decision table(s) [14]. It bridges appliance control data expressed in natural language by the users, on the one hand, to applications around the framework (knowledge representation, which is use case specific), on the other hand, at the executive layer (the real decision making processes) through the SDT reasoning engine [13]. An SDT is, first of all, a decision table. Hence it contains all the known features of decision tables, such as completeness and syntactical consistence. In addition, it provides a means to capture and examine decision makers’ concepts, as well as a tool for refining their decision knowledge and facilitating knowledge sharing in a scalable manner. With SDT, hidden semantics from decision tables are specified through ontological specification and annotation. SDT also makes interoperability possible among the agents. Including how to fully test the functionalities in SDT and its engine, one of the challenges is how to evaluate the usability of SDT and decision tables. In this paper, we will illustrate our innovative approach to modeling users’ preferences using SDTs. To make it clearer, we will first illustrate the framework and use cases in the next section. III.
THE FRAMEWORK
Based on the related work [6], [10], [17], we create the framework as illustrated in Fig.1. The key architectural components are the demand response small and medium enterprises (DR SME), DR Community, and DR Fabric. A Demand Response SME is an IoT framework-embedded computing unit that has a physical (wire or wireless) connection with a smart meter using the IoT Device, and sets out its functionalities to the DR Community and DR Fabric through IoT Overlay realized in the DR Fabric. Each smart meter collects consumption information and sends it back to the DR SME node. The DR SME may transform this information into a DR event processed by the IoT Kernel, which may then liaise with other DR SME through the DR Community. DR applications (desktops, Web, mobile) use IoT API to interact with IoT Context to retrieve interesting application scenarios and to further control the DR system. DR Fabric consists of innetwork system functions (routing, admission control) that
are provided either by applications themselves (e.g. extending DR Community) or by a customizable DR Fabric reference implementation. Utility Company
DR Community IoT API IoT Context IoT Overlay
DR SME
DR SME IoT Overlay
DR SME
IoT Kernel
IoT Overlay
IoT Device
IoT Kernel
IoT Overlay IoT Kernel IoT Device
IoT Device
DR Fabric IoT API IoT Context
DR Mobile Apps
IoT Overlay
DR Desktop Applications
DR Community
DR Web Apps
IoT API IoT Context IoT Overlay
DR Developers
DR Prosumers
DR SME
DR SME IoT Overlay
DR SME
IoT Kernel
IoT Overlay
IoT Device
IoT Kernel
IoT Overlay IoT Kernel IoT Device
IoT Device
Figure 1. The Community-based, Semantic Decision Making Framework for Energy Efficiency in Smart Metered Homes.
Semantic decision tables are the decision models that can be illustrated by a DR application. Examples of modeling the user preferences in a smart home and milk farm, which will be used to illustrate SDT models, will be shown in the next subsections. A. Use Case 1: Model User Preferences in Connected Smart Homes Suppose a user in a smart home wants to control the lights, ventilation and air conditioners etc. with the help of the framework illustrated in Fig. 1. He may create a set of decision rules in SDT, which stores his user specific rules. A rule set can be, for instance, the decision on whether or not turn on the air conditioner, ventilation and lights based on the temperature in the house, the air quality inside/outside the house, and the power needed to switch the devices (and probably also the current power load in the house). The user sends a query like “turn on the air conditioner” to a sort of Universal Control Box (UCB) enabled with SDT, which stores this rule set. The SDT engine will provide a decision and provide the feedback to the UCB. When he is away for a vacation, he wants to have his ventilation working from time to time, but not other devices. Then he may update the preferences, and consequently the SDT models. Another interesting insight is to share his energy with his neighbors if he has an installed solar panel which provides
excess energy at certain points in time. This is where the community aspect of our framework comes. He puts the available dates and prices on a community portal, through which people from the same neighborhood can purchase. Smart meters are used in the places where the user wants to get as the inputs to the SDT models and, also as an appliance the user wants to monitor and control. For example, it is connected to the air conditioner, ventilation, light and the solar panel. The user can remotely create rules (SDT models), which are automatically executed by the whole system. The scenario concerns solar panels but can as well be applied to renewable energy in general as shown in the following subsection. B. Use Case 2: Manage Energy Resource of a Farm Community This use case comes from a milk farm. When a farm is producing and storing (freezing) milk, the demand of energy is high. Therefore, it might need to purchase energy from the other farms around. This kind of energy can be solar energy, wind energy, gas, or any other form of renewable energy. We compare the energy demand, the energy available and the offers from the other farms in the system. SDTs are used to model the preferences of the demand and responses. All the farmers from this farm community put together their energy excess and sell it to the grid (DSO – Distribution System Operator). This is happening after the SDT engine decides (based on rules and conditions and context) that it is indeed about energy excess, i.e. they do not need it at that time. As indicated by the title of this paper, we will focus on the previous use case. We illustrate the use case of a farm community for showing how diverse the applications from our framework can be. In the next section, we will show how SDTs are modeled for the use case of smart-metered homes. IV.
SEMANTIC DECISION MODELS
An SDT provides a means to capture and examine decision makers’ concepts, as well as a tool for refining their decision knowledge and facilitating knowledge sharing in a scalable manner. With SDT, hidden semantics from decision tables are specified through ontological specification and annotation. SDT also makes interoperability possible among the agents. Each SDT records the result of a group effort. Let us first show the structure/model of an SDT. A. Structure of Semantic Decision Tables A semantic decision table (SDT) contains three parts (see TABLE I). The first part is a decision table. According to the standard [1], a decision table DT is modeled as a triple , , , where C is a set of conditions, A is a set of actions, and F is a set of decision rules. Each condition c ( ) is defined as , where is a condition stub (label), and is a condition entry (value or value range); where M is a set of action stubs (labels) and U is a set of labels for Boolean values (for example, T/F, 0/1, Yes/No) that are used as action entries. Each decision
rule is defined as a function : where as usual denotes the set of all assignments of N to V. The second part is called SDT lexon base – a set of binary fact types. A lexon l is a binary fact type and defined , where as a quintuple , , , , , T is a finite set of linguistic terms; and represent two concepts in a natural language. R is a finite set of roles; and ( corresponds to “role” and to “co-role”) refer to the relationships that the concepts share with respect to one another; is a context identifier that refers to a context, which serves to disambiguate , into the intended concepts, and in which they become meaningful. For example, the lexon L1 in TABLE I explains a fact that “PV is Solar Panel”, and “Solar Panel is PV” in the context identified by ; is, i.e., a URI that points to this fact. The third part is called SDT commitment layer – a set of axioms. A commitment is a formal agreement on how an ontology-based application will possibly make use of lexons in a consistent way. We use Decisional Commitment Language (DECOL) to write the axioms. For instance, we have a DECOL statement as below. L1 = [Solar Panel, is, is, PV]: EQUAL (L1 (Solar Panel), L1 (PV)). It means that the sets ‘Solar Panel’ and ‘PV’ are equivalent. A DECOL statement can be mapped into OWL, if needed. In [13], how to create the map between DECOL and OWL is discussed. TABLE I. SDT1-DECISION ON WHETHER OR NOT TO SWITCH ON A DEVICE BASED ON POWER GENERATED BY SOLAR PANELS (WATTS/HOUR), DEVICE CONSUMPTION (WATTS/HOUR) AND INDOOR TEMPERATURE (°C) Condition Power generated by PV Consumption of X Indoor temperature Action Switch on X ID L1 L2 L3 L4 L5 L6 L7 L8 L9 L10 ID K1
1
2 3 4 >=CL+[X] 100 =25 =25
5
6 7 8 =0.
TABLE I shows an SDT example, with which we decide whether or not to switch on a cooling device based on the total power generated by the solar panels (PV), consumption of this device and the current indoor temperature. This device can be, for instance, a ventilator or an air conditioner. Column 2 in TABLE I contains a decision rule as – switch on device X if the following conditions are satisfied. • The power generated by PV is more than the total of current load and the consumption of device X, and • The consumption of device X is less than or equal to 100 watts/hour, and
•
Current indoor temperature is more than or equal to 25 °C. As pointed out earlier, an SDT is a result of a group effort. How to guide a group of decision makers to create SDTs can be found in [14]. In this paper, we would rather focus on the practical aspects of the relations amongst a set of SDTs. A fundamental problem of a decision table is that we should keep a table small enough to have a clear view. Otherwise, it will lose its visibility [7]. And if we look at the definition of decision table, then we will see that the main factor that affects the size of a decision table is the number of conditions. Hence, a decision table should not contain too many conditions. One way to solve this problem is to carefully separate unrelated conditions from a table, make a new decision table by compiling those conditions, and link the current table with the new table through an invocation, which, according to [5], can happen at the levels of conditions or actions, or both. In the next subsection, we will illustrate an algorithm of creating invocational (semantic) decision tables. B. Creating Invocational (Semantic) Decision Tables The algorithm is illustrated below. DEFINE T: an empty set DEFINE : condition stub set of sdt DEFINE : a constant Integer, e.g. 3 BEGIN PROCEDURE BOOLEAN Check (SDT sdt) IF | | THEN IF ( , ) THEN sdt1 = Make (sdt, C1) sdt2 = Rest (sdt, C1, sdt1) RETURN Check (sdt1) RETURN Check (sdt2) ELSE ADD sdt to T RETURN TRUE END Check BEGIN PROCEDURE SDT Make (SDT sdt1, C C1) SDT sdt2 sdt2.C = C1 sdt2.A = relevant action stubs from sdt1 RETURN sdt2 END Make BEGIN PROCEDURE SDT Rest (SDT sdt0, C C1, SDT sdt1) SDT sdt2 sdt2.C = sdt0.C – C1 sdt2.A = relevant action stubs from sdt0 Add GOTO sdt1 command to sdt2.A RETURN sdt2 END Rest
As illustrated, this algorithm first groups condition stubs and action stubs for different sub-contexts. Then, it recursively creates tables by two, one of which contains a “GOTO” command that invokes the other. When it stops, we will get a set of empty SDTs, which contains condition stubs and action stubs. The follow-up tasks will be executed by the decision modeler, who will enter the decision rules and, if necessary, create a call loop. For example, a user wants to be able to use both devices for cooling and heating. He accepts to turn off other devices if he does not have enough power. He wants to buy energy from his neighbors if he does not have enough energy. And he also wants to sell his energy when he is away. In this example, we have the condition stub set as {Power generated by PV, Indoor temperature, Consumption of X/Y, Lights are on, People in the room, People at home, Ventilator is on, Air quality, More communities available, Offer available, Price} and the condition stub set as {Switch on X, Switch on Y, Turn off lights, Turn off ventilator, Buy, Search another community, Sell}. Those stubs are under the sub-contexts identified as “switch a device on/off”, “turn off lights”, “turn off ventilator”, “buy energy”, “sell energy”. After we run the algorithm and manually set the decision columns, we get a set of SDTs as shown in TABLE II ~VII. TABLE II. SDT2- DECISION ON WHETHER OR NOT TO SWITCH ON A DEVICE (COOLING OR HEATING) BASED ON POWER GENERATED BY SOLAR PANELS (WATTS/HOUR), DEVICE CONSUMPTION (WATTS/HOUR) AND INDOOR TEMPERATURE (°C) (TABULAR VIEW ONLY) Condition Power generated by PV Indoor temperature Consumption of X Action Switch on X Switch on Y GO TO SDT3 GO TO SDT5
1 2 >=CL+[X] =18,=25
100
100
6
7 =CL
TABLE VI. SDT6 – DECISION ON WHETHER OR NOT TO BUY ENERGY BASED ON THE AVAILABILITY OF COMMUNITIES THAT OFFER ENERGY, THE AVAILABILITY OF THE OFFERS AND THE PRICE (€ PER KWH) (TABULAR VIEW ONLY)
TABLE IV. SDT4 – DECISION ON WHETHER OR NOT TO TURN OFF A VENTILATOR BASED ON THE CURRENT STATUS OF THE VENTILATOR AND THE AIR QUALITY IN THAT ROOM (TABULAR VIEW ONLY) Condition Ventilator is on Air quality Action Turn off ventilator GO TO SDT2
1
•
Instantiation of conditions and actions; how we deal with the instantiation is as follows. Firstly, we map a data record from a database to a condition stub or action stub. Then, we interpret a condition or action through a middleware, such as ZigBee, Z-Wave, Bluetooth and 6LoWPAN etc., in order to get the
•
input from a physical device and provide an output to a physical device. For instance, in TABLE IV, the condition stub “ventilator” is mapped to the ventilator in the room, which has the room number 731, from a particular address. And we get the commands for the condition , via a middleware. We use smart meters with regard to the measurement of consumptions of any devices. Interpretation of “GOTO” commands; the file locations of the origin and the target of these commands are indicated. For example in TABLE III, the GOTO command is interpreted using the following DECOL statements. P6 = [GOTO SDT4, is instance of, has instance, INVOKE_COMMAND]. P7 = [GOTO SDT4, has, is of, FILE_LOCATION]: P7 (FILE_LOCATION) = “C:\\x.xml”.
•
Initialization of variables and constants; for example in TABLE II, X and Y are two variables of devices. If we set X as a ceiling fan in a particular room, then [X] – the consumption of X – will be replaced by 50 watts. The DECOL statements of this example is illustrated as follows. P8 = [X, has, is of, Consumption of X]: P8 (X) = ‘ceiling fan 2341’. P9 = [[X], is, is, Value]: MEASUREMENT_UNIT (P9 (Value), watts), P9 (Value) = 50.
•
Interpretation of fuzzy values; let us take TABLE IV as the example, the condition entry “Good” for “Air quality” means that the time-weighted average (TWA) limit for carbon monoxide is less than 25 ppm. Otherwise, “Air quality” is considered “Bad”. As this task is non-trivial, we use an annotation in natural language for its explanation. The function is implemented through a middleware (see the above discussion in “Instantiation of conditions and actions”). SDT2
SDT4
GOTO SDT3
SDT3
GOTO SDT5
GOTO SDT4
GOTO SDT2
GOTO SDT7
V.
THE IMPLEMENTATION
Our tool is developed in the Eclipse 12 Plug-in Development Environment (PDE), which reuses the Eclipse APIs for modeling and visualizing dependencies. Fig. 2 shows a screenshot. From left to right and from up to down, the views are listed as follows: • Dependence model list, which contains the following three elements o Domain. Each domain can have several diagrams. o Diagram. Each diagram can have several SDTs, but can belong to only one domain. o SDT. Each SDT can belong to several diagrams, but only to one domain. • SDT dependency visualizer, which currently supports invocations, aggregation and sequence. In this paper, we focus on invocational models. Aggregation is also called sub-table dependence. Generally speaking, a selection of rows and columns in an SDT can be recomposed into another SDT. The former is the parent table of the latter; and, the latter is the sub-table of the former. Sequence is similar to invocation. The only difference is, in a set of sequential SDTs, it is not necessary to use commands like “GOTO”. It can be restricted at a meta-level. • Simple table viewer. When a user clicks on an SDT in the SDT dependency visualizer, it will show a decision table with relevant decision rules. Only the columns with actions are illustrated. The ones without actions are not illustrated. Let us take TABLE II as an example. The columns 3 and 4 will be hidden in this view. It simplifies the view and can assist users with a neat interface. • Properties. It shows the file location of the currently selected SDT. VI.
SDT5 GOTO SDT6
After we get all the SDTs, we should draw a dependency map (as illustrated in Fig. 2) and test them with transactions. Note that dead loops are not allowed. How to detect dead loops is one item of our on-going work. In the next section, we will show the implementation of visualizing invocational SDTs.
SDT6 GOTO SDT6
SDT7 Figure 2. The Dependency Map of SDT 2~7
CONCLUSION
The paper proposes an innovative framework for energy efficiency based on a demand-response system built on top of ontology engineering technologies. We have used the Internet-of-Things (IoT) computing paradigm and adopted the Resource-Oriented Architecture (ROA) style. In the past, no work has been done for demand forecasting at a community level as envisaged in this study. It represents a crucial element for the real-time decision support. The study also demonstrates the application of SDTs for community-grounded decision rules. 12
http://www.eclipse.org
Within the scope of this paper, SDT plays the following roles as: 1) A community-grounded means to create sharable decision making rules. In our case, for example, it is important for an energy buyer to understand the purchasing rules designed by a seller. 2) A bridge between domain ontologies and rule modelers/end users. It is often a challenge to “deploy” the semantics defined in an ontology to the rule-based applications – it is often a non-trivial task. In this paper, we want to use the technologies of SDT to turn this task into non-trivial and make our rule base more scalable. 3) A way to ensure the consistency and completeness of the decision rules, which is necessary for a rule-based system. The above advantages brought forward by SDT makes our work more advanced than the related work. Including how to fully test the functionalities in SDT and its engine, one of the challenges is how to evaluate the usability of SDT and decision tables. Future work consists of the implementation of a visualization client for setting the user preferences (rules) for both PC and mobile devices. When we decompose a large decision table into several dependent small tables, we substitute the static complexity of a larger table with the dynamic complexity of invocation chains at a certain level. How to deal with this dynamic complexity is another interesting research objective in the future.
[5]
[6]
[7] [8]
[9]
[10]
[11]
[12]
[13]
[14]
ACKNOWLEDGMENT The work has been partly supported by the Open Semantic Cloud for Brussels project (OSCB) founded by the Brussels Capital Region. We are very pleased to thank Prof. Tharam Dillon for his input concerning the demand response architecture in this paper. REFERENCES [1] [2]
[3] [4]
CSA, (1970): Z243.1-1970 for Decision Tables, Canadian Standards Association, 1970. Darby, S. (2006): The effectiveness of feedback on energy consumption: A review for DEFRA of the literature on metering, billing and direct displays. Tech. report, Environmental Change Institute, University of Oxford, 2006. Gong, L. (2001) JXTA: A network programming environment. IEEE Internet Computing, 5, 88 – 95. Guinard, D., Trifa, V., and Wilde, E. (2010): A resource oriented architecture for the Web of Things, in Proc. of Internet of Things (IoT), pp. 1–8.
[15]
[16]
[17]
[18]
Henry Beitz, E., Buck, N.H., Jorgensen, P. C., Larson, L., Maes, R., Marselos, N. L., Muntz C., Rabin, J., Reinwald, L. T., and Verhelst, M. (1982): A modern appraisal of decision tables, a Codasyl report, ACM, New York. Jagatheesan, A., Helal, S. (2003) Sangam: Universal Interop Protocols for E-Service Brokering Communities using Private UDDI Nodes. IEEE Symposium on Computers and Communications. ISCC’2003. Kohavi, R. (1995): The Power of Decision Tables, Machine Learning: ECML-95, LNCS 912/1995, 174-189. Papazoglou, M. P., Kramer, B. J., Yang, J. (2003) Leveraging WebServices and Peer-to-Peer Networks. Springer-Verlag Berlin Heidelberg 2003. Pooch, U.W. (1974), Translation of Decision Tables, ACM Computing Surveys (CSUR), Vol. 6, Issue 2, pp.125 -151, ISSN: 0360-0300, ACM, New York, NY, USA. Richardson T, Wellings A, Dianes J, Díaz M. (2009) Providing temporal isolation in the OSGi framework. Proceedings of 7th International Workshop on Java Technologies for Real-time and Embedded Systems, Madrid, Spain. Sestito, S., Leung, S., Dillon, T.S. (1991) Short term load forecasting using an adaptive neural network, Journal of Electrical Power and Energy Systems, vol. 13(4), pp. 186-192. Szkuta, B.R., Sanabria, L.A., Dillon, T.S. (1999) Electricity price short-term forecasting using Artificial Neural Networks. In IEEE transactions on power systems, 14(3):851-857. Tang, Y. and Meersman, R. (2011): Towards Directly Applied Ontological Constrains in a Semantic Decision Table, in proc. of 5th International Symposium (RuleML 2011) Springer LNCS 7018, pp. 193-207, F. Olken, M. Palmirani, D. Sottara (eds.), Nov. 3-5, 2011, Ft. Lauderdale, FL, USA Tang, Y. (2010): Semantic Decision Tables - A New, Promising and Practical Way of Organizing Your Business Semantics with Existing Decision Making Tools , ISBN 978-3-8383-3791-3, LAP LAMBERT Academic Publishing AG & Co. KG, Saarbrücken, Germany. Tang, Y., and Meersman, R. (2009): SDRule Markup Language: Towards Modelling and Interchanging Ontological Commitments for Semantic Decision Making, Handbook of Re-search on Emerging Rule-Based Languages and Technologies: Open Solutions and Approaches, IGI Publishing, ISBN: 1-60566-402-2, USA. Tomic, S., Fensel, A., and Schwanzer, M., Veljovic, M.K. and Stefanovic, M. (2012): “Semantics for energy efficiency in smart home environments”, In Applied Semantic Web Technologies, chapter 16, pp. 429–454. Vanthienen, J. (1994): A more general Comparison of the Decision Table and Tree, Communications of the ACM, Vol. 37, No. 2, pp.109-113 Verma, K., Sivashanmugam, K., Sheth, A., Patil, A., Oundhakar, S., Miller, J. (2005) METEROR-S WSDI: A Scalable P2P Infrastructure of Registries for Semantic Publication and Discovery of Web Services. Information Technology and Management, 6
Figure 3. The Tool for visualizing invocational SDT
