eChallenges e-2014 Conference Proceedings Paul Cunningham and Miriam Cunningham (Eds) IIMC International Information Management Corporation, 2014 ISBN: 978-1-905824-45-8
Standards and Semantics to Support Interoperable Software Solutions in the Water Distribution Chain Andreas ABECKER1 , TorstenBRAUER1, Babis MAGOUTAS2, Gregoris MENTZAS2, Nikos PAPAGEORGIOU2, Michael QUENZER1 1 Disy Informationssysteme GmbH, Ludwig-Erhard-Allee 6, 76131 Karlsruhe, Germany Tel: +49 721 16006-000, Fax: + 49 721 16006-05, Email:
[email protected] 2 Institute for Computer and Communication Systems (ICCS) at the National Technical University of Athens, Iroon Polytechniou 9, Zografou, 15780 Athens, Greece Tel: +30 210 772 3895, Fax: +30 210 772 4042, Email: (elbabmag|gmentzas|npapag)@mail.ntua.gr Abstract: The goal of the EU-FP7 project WatERP is more interoperability of software systems along the water-supply chain. Having a better data exchange between different stakeholders along the supply chain shall enable more intelligent decision-support for saving water and energy. In order to achieve that higher degree of interoperability, the WatERP project develops a flexible and extensible communication architecture for the different kinds of tools in use, based on a Service-Oriented Architecture (SOA) which is complemented by a Multi-Agent System (MAS) for service identification and orchestration. One central element of the software architecture is the WatERP Water Data Ware-house which acts as a central data hub between different software systems. Both the SOA-MAS architecture and the WDW are based to the most reasonable extent on OGC’s open standards for geospatial data processing and are carefully extended by semantic technologies. In this paper, we sketch the SOA-MAS architecture and explain some details of the WDW.
1. Motivation and Objectives A holistic view on the way of a drop of water, starting with a precipitation in a river catchment area or with a deep-well to an aquifer, and ending with a private household, an industrial or an agricultural consumer, reveals a whole bunch of administrative, commercial and societal actors and stakeholders that altogether create the drinking-water supply chain or value-network, or have some interest in it. This value-network is composed of a number of different actors (if we also consider private consumer households, even a huge number) with significantly differing ways of working, depending on their place and their role in the water-supply chain. Such actors can, for instance, be bulk-water suppliers, dam administrations, local or regional utility companies, large industrial consumers, private households, environmental agencies, NGOs, etc. These actors also comprise different kinds of organizations, such as public authorities, administration unions, special-purpose associations, private companies, public-owned enterprises, public-law institutions, private persons, etc. Their tasks and objectives are not restricted to supplying water to consumers, but also affect constraints and secondary goals, like energy production, flood protection, environment protection, quality control / quality assessment of the drinking water etc. It is a nearby hypothesis that some of the activities of such a multitude of actors, actions and objectives could gain efficiency or effectiveness if there was a comprehensive and prompt exchange of data between different stages in the supply chain. For instance, when Copyright © 2014 The Authors
www.eChallenges.org
Page 1 of 10
optimizing the water distribution in temporarily droughty areas, an early estimation of the upcoming water yield of a bulk-water supplier at a given interconnection point together with a good short-term/mid-term estimation of the different competing consumers’ demand could facilitate the reduction of causes of conflict between different consumers. Further, economic incentives (like flexible water tariffs), organizational measures (timed coordination between different industrial consumers), motivational activities (giving information or appeals to the citizen) or appropriate regulations (setting of quotas) – as methods of Demand Management – could be better coordinated. Or, in the case of a foreseeable peak-demand, the water production could be controlled more efficiently (operation management of deep-wells, water works, reservoirs, etc.). But also in water-rich areas, optimization potentials can be expected, in particular with respect to the energy consumption and/or energy costs (in the case of flexible energy tariffs) needed for operating the water-distribution network (i.e. how to run the pumps for filling elevated tanks and for keeping the water pressure in the pipe network), if one took into account more exact demand forecasts and/or more actual consumption data, or if one coordinated the consumption-plans of large industrial customers. It should also be noted that optimization potentials do not necessarily need to span different organisations; often enough, there are already interoperability deficiencies between the software systems of only one actor. However, up to now, a systematic, comprehensive and real-time data and software interoperability throughout the whole water-supply chain is not given in practice. Typically, if there is a systematic, electronic communication or coordination between two actors, then (i) it is between two immediately subsequent stages in the value-chain, (ii) it is done only for the purpose of this obvious local collaboration need for the operational process at the intersection of these two organisations, (iii) it is typically realized with simple, tailor-made, technical means, and (iv) it is not always based on most actual data. In order to improve this situation, the EU-FP7 project WatERP (“Water-Enhanced Resource Management – Where Water Supply Meets Demand“) [1, 2] was started with the following goals: 1. to realize a set of decision-support tools (tools for demand forecasting, for energyoptimized network operations, for demand management, etc.) for reducing water consumption or reducing energy consumption through using a better information base from the whole water-supply chain; 2. to implement a data-exchange platform through which the different actors in the watersupply chain can share their observation data, forecasted values, and management decisions; and 3. to set-up an overall communication architecture, based upon widespread standards and semantic technologies which allows to make optimum use of the latter two developments, in realistic domain- and organisation-spanning software environments. While largely ignoring goal (1), this paper will sketch the approach to achieve goal (3) in Section 3 and focus on the approach to achieve goal (2) in Section 4 – which is also the technical focus of the paper. In Section 5, we briefly describe the piloting test beds and expected business benefits before we conclude with Section 6.
2. State-of-the-Art and Project Methodology To improve decision making in water management, several authors have proposed SOAbased interoperable software where Web services based on hydrological procedures perform the connection between different involved systems. For instance, [4] presents an XML / SOAP Web service architecture with a specific XML language for data exchange between services and user interfaces. But most of the solutions in the literature are not standardised with respect to the database structure such that the understanding of hydrological information depends on the concepts implemented in non-standardised XML Copyright © 2014 The Authors
www.eChallenges.org
Page 2 of 10
or in databases. This lack is overcome in approaches that use the OGC® stack complemented by an OpenMI, ODM data model and WaterML2 [1]. But these approaches cannot include new systems without manually re-modelling the architecture. Further, they are only focused on the management of hydrological sensor data without taking into account the interconnection of different management tools available in the water supply and distribution chain. Consequently, the methodological approach of WatERP is (i) to define a serviceoriented data-exchange architecture with communication protocols and data schemas, (ii) to realize a water-data warehouse as the central data-sharing platform, and (iii) populate the architecture not only with real-world example data, but also with exemplary decisionsupport tools aiming at improved water-management decisions based on more comprehensive data exchange. All that based to the most reasonable extent on open standards and grounded in semantic technologies as a means to express shared understanding. Lastly, not only building a practice-oriented solution, but also extensively test in two pilot test beds.
3. Technology Description: WatERP Overall Architecture The WatERP architecture combines a Service Oriented Architecture (SOA) with a Multi Agent System (SOA-MAS) and complements it by the WatERP Water Management Ontology (WMO) to support interoperability and intelligent orchestration of system functionalities. The SOA supports interoperability between systems through OGC® standards (WPS1 for process integration, SOS2 as a data-exchange protocol, WaterML2.05 as a data-exchange format). The architecture permits data exchange in an interoperable way by the use of an ontology which models the hydrological water domain (information and management) to enhance water-data understanding and machine readability. The WMO integrates the conceptual models of WaterML2.0 and other hydrological data and domain models and represents them in OWL. This is a technical means to allow for semantic reasoning about data and it paves the way to easier Linked Data publishing. The combination of SOA-MAS and a semantic knowledge base provides the necessary interoperability through a common communication language which is currently not supported by WaterML2.0. The SOA-MAS architecture [1] shall accomplish the following objectives: (i) link each decisional or informational system to help the integration in a collaborative framework, (ii) provide near real–time information flow, (iii) distribute intelligence to generate actions and alerts related to the water-management processes, and (iv) perform orchestration of existing and new management tools. The proposed architecture can be extended by more systems in a dynamic and automatic way. It permits the water managers to enhance their daily decision-making by combining the SOA-MAS with an ontological data understanding towards orchestration of decisional systems. A central element of the WatERP data-exchange platform is the so-called WatERP Water Data Warehouse (WDW). In the WDW, different kinds of data, information and knowledge created in the water-supply chain are centrally collected, stored, and provided for different tools and purposes in different views, formats and aggregations. Figure 1 gives an overview of the overall WatERP software architecture: The WDW stands at the centre of the system. It is fed by data-producing systems linked to the WDW through OGC® SOS interfaces exchanging WaterML2.0 data. The data-consuming decision-support tools realize a Service-oriented Architecture (SOA), enabled by the OGC WPS (Web Processing Service) protocol for communication, and facilitated by a Multi1 2
http://www.opengeospatial.org/standards/wps http://www.opengeospatial.org/standards/sos
Copyright © 2014 The Authors
www.eChallenges.org
Page 3 of 10
Agent System that helps finding appropriate services and orchestrating complex processing workflows. The conceptual backbone of the system is the WatERP Water Management Ontology (WMO) which represents all decision-relevant aspects of the water-supply chain as an OWL knowledge base (see also [1, 2]). All sensor and measurement transmission activities follow the OGC® SOS protocol with WaterML2.0 as data schema. The internal organization of the mass-data storage also implements a simplified and tailor-made data schema based on the OGC® WaterML2.0 data specification. For implementation, we employed the 52°North SOS Server.3 All data sources are linked to the WDW through SOS – except for the meteorological forecasts where we considered WFS/WFS-T4 more appropriate. Of course, time-series data are not only imported, but also exported through an SOS/WaterML2.0 interface.
Figure 1: Overall WatERP Software Architecture
4. Developments: Semantics-Enabled Water Data Warehouse The Water Data Warehouse copies data from operational systems and offers them in optimized views, aggregations and formats for analysis purposes. Figure 2 sketches the WDW architecture within the overall project context. One major design decision was to have two separate storage areas: Mass data (time-series data describing sensor measurements about hydrology, network operation, water consumption, etc.) is highly repetitive, simply structured, but coming in large volumes. Large amounts of this kind of data can easily and efficiently be managed with “conventional” database technologies, in our case, an object-relational storage area (PostGIS database5). These data populate a data schema which is based on the structure of the OGC / ISO conceptual model „Observations and Measurements (O&M)“6 and on OGC® WaterML2.07 as its instantiation in the water domain. Of 3
http://52north.org/communities/sensorweb/sos/ http://www.opengeospatial.org/standards/wfs 5 http://postgis.net/ 6 http://www.opengeospatial.org/standards/om 7 http://www.opengeospatial.org/standards/waterml 4
Copyright © 2014 The Authors
www.eChallenges.org
Page 4 of 10
course, the database schema had to be extended because WaterML2.0 is mainly focussed on sensors for hydrological observations whereas we also monitor the whole technological-economic distribution and decision system. The necessary extensions have been identified and specified in the Water Management Ontology, based on the domain analyses of our two pilot users in the project. Such extensions are easily possible because O&M is a very generic standard. On the other hand, when one describes all decision-relevant aspects of a water supply chain comprehensively, there will arise the need to model also more irregularly structured knowledge, more complex relationships between objects, definitions of technical terms or interrelationships of them, or generic relationships which abstract away from specific facts (like rule-based knowledge or arithmetic relationships). For expressing such more complex issues, computer science has developed knowledgerepresentation languages for expert systems which have been consolidated and standardized through “Semantic Web” technologies. In the Semantic Web area, socalled triple stores have been introduced for storing and processing complex knowledge. More recently, triple stores have been extended towards geospatial reasoning which allows drawing logical deductions that also include also simple notions of spatial relationships and spatial deductions. For instance, sensors could be geo-referenced and the system could, if appropriately modelled, find all sensors in a given administrative district or in the state which comprises this district; or all sensors situated on the rivers that a given public authority is responsible for. So, the second main storage area of the WatERP WDW is a triple store enabled to do geospatial reasoning.
Figure 2: Simplified Architecture of WatERP Water Data Warehouse
The incorporation of a semantic knowledge base (i) follows the trend to empower modern software solutions by knowledge-based components, (ii) increases interoperability through ontologies and (iii) makes it easier to provide data according to the Linked Open Copyright © 2014 The Authors
www.eChallenges.org
Page 5 of 10
Data paradigm. However, a complete “semantification” (representation, storage and processing of all aspects with Semantic Web methods and tools) of all data in WatERP seemed not feasible and promising to us, especially regarding the measurement (sensor) data. Instead, the time-series data and the semantic knowledge complement each other and could together be exploited for powerful analyses and queries about the considered water supply system. This was the reason to make the distinction between two storage areas introduced above. For instance, the structured relationships in the triple store could logically describe a water-distribution network and its geospatial aspects, as well as some background knowledge, e.g., about measurement methods for assessing water quality. Then, specific water-quality measurements could continuously be fed as time-series data into the PostGIS database. PostGIS would only contain the raw measurement values whereas the measurement metadata (where is the sensor, what is monitored, which sensor technology is used, etc) would establish the links into the semantic knowledge base. This allows making complex queries which combine time-series sensor data and semantic background knowledge. For instance, one could ask for all measurements made in a certain geographic area and using an analytical sensor technology with certain characteristics; or, about all sensor data from a certain point on in the water supply chain which make statements about a certain group of related chemical or biological pollutants. Analysing the answers could help to find upstream dischargers responsible for a contamination and take appropriate counter-measures, or it could help to identify endangered spots further downstream and take suitable protection or purification measures.
Figure 3: Semantic Reasoning Architecture within the WDW
It should be noted that, as a side-effect of the two-part WDW architecture, the simultaneous evaluation of (i) quantitative measurement-data based and (ii) qualitative, knowledge-based conditions in one query cannot be done within the WDW itself. So, if one would like to find, for example, all sensor observations coming from a certain Copyright © 2014 The Authors
www.eChallenges.org
Page 6 of 10
administrative area (spatial reasoning) and made with a certain type of sensor technology (structural reasoning) which are above a certain threshold (sensor value), this would have to be implemented as a specialized query agent outside the WDW that merges the query results of the SOS sensor-data query and the logic-based query. Figure 3 illustrates the implementation of the WDW geospatial reasoning which is based on the uSeekM und Sesame Java libraries in combination with a PostGIS geodatabase and the Sesame-based RDF triple store OWLIM. Sesame8 is a Java framework for the implementation of RDF stores that offers an extensible API on top of which other stores can be built. OpenSahara uSeekM9 is a Java library which realizes spatial indices and spatial queries as an extension of Sesame Java based triple stores. uSeekM creates a separate R-Tree index for spatial data. It catches GeoSPARQL10 queries and rewrites them as a combination of queries against the spatial index and the RDF triple store. Different triple stores can be used, as far as they support Sesame Sail. The Sesame Sail API allows for functional extensions of RDF stores as a low-level system API for RDF stores and inference engines which abstracts from implementation details and so allows using different stores and inference engines. In WatERP, we employ OWLIM11 as an RDF store that allows reasoning over an OWL ontology, such that, altogether, the WDW Triple Store allows combined reasoning over ontological and spatial knowledge.
5. Use Cases and Business Benefits At the time of preparing this paper, the first fully working prototype of the WDW is up and running and is actually being filled with real-world test data of the two WatERP pilot users, the Catalan Water Agency (Agencia Catalana de l'Aigua / ACA, Barcelona, Spain) and the water utility of the city of Karlsruhe (Stadtwerke Karlsruhe GmbH, SWKA Germany). The whole WatERP solution shall be installed and tested during 2014/15 in these two pilot environments:
Figure 4: Schematic Overview of Ter-Llobregat Pilot (ACA Use Case)
8
http://www.openrdf.org/ http://www.w3.org/2001/sw/wiki/USeekM 10 http://www.opengeospatial.org/standards/geosparql 11 http://www.ontotext.com/owlim 9
Copyright © 2014 The Authors
www.eChallenges.org
Page 7 of 10
1) The Ter-Llobregat water supply distribution system, located in North-Eastern Spain and managed by ACA is the WatERP pilot focussing on the upper part of the water-supply chain. The system regarded consists mainly of the Ter and Llobregat river basins, which have a combined area of 7,912 km2 and an annual average water volume of 1,492 hm3. Due to the Mediterranean climate, the amount of available water can be very variable, from below 500 hm3 to over 4,000 hm3. Water demand in the Ter-Llobregat system (with about 6 million inhabitants and over 140 municipalities) is primarily for agricultural, urban and industrial uses as well as municipal. The basic regulation of surface water is made by 5 reservoirs (total capacity 600 hm3) distributed across the two basins, supplemented by desalinisation plants that can provide 70 hm3 per year and groundwater resources up to 160 hm3 per year. Currently, a number of benefits from data exchange are envisaged: - Better water availability predictions by integration of information from weather radar and numerical weather prediction. - Better water availability predictions by long-term calibration using rain-gauges and by considering ground echoes and orographic screening as well as different available radars. - Better water availability predictions by directly feeding the precipitation forecasts into the hydrological model of the basin. - Better negotiations with (competing) industrial and agricultural water consumers through earlier and more precise availability estimates. - Possibilities for demand management of private households by more transparency on usage data and predicted availability. 2) The city of Karlsruhe distribution network is managed by SWKA as public water utility which provides water to its inhabitants. The City of Karlsruhe in South-Western Germany obtains all of its water from a sole source, the underlying groundwater, which stores about 1.3 Bill m3 of ground water in the aquifer in the region around Karlsruhe. SWKA produces 23 to 24 Million m3 of drinking water per year. The drinking water network has a length of 900 km. This pilot case is focussed on the lower part of the distribution network, the pipe-network in the city and its immediate neighbourhood. In contrast to the water-saving goal that dominates in the Southern-European countries, we here focus on operations management aiming at an energy-saving goal related to the energy-consumption incurred by water utilities for operating their facilities. Here we envisage the following effects enabled by increased data exchange: - Access to historical water-consumption data allows for a case-based approach to devise and improve water-demand predictions for the near-future (considered period: day to week). - Fine-grained (daily water-use profiles on the household-level that can be aggregated over time and over kinds of users or over neighbourhoods/ network zones) monitoring and documentation of water-consumption allows for better understanding of consumers which supports both efficient daily operations and long-term network maintenance. - More precise demand-predictions allow more sophisticated operations management (When to fill how much water into which elevated tanks? Which pressure to maintain in a pipe-network zone when?). Although first estimations in the project suggest that water savings in the dimension of 8% in the ACA pilot and energy savings in the dimension of 5% in the Stadtwerke Karlsruhe pilot may be achievable through more integrated and smarter software, it would be dubious to make such promises at this point in time. First, we deliver and test a technical solution. The concrete impact of applying such solutions highly depends on the exact usage Copyright © 2014 The Authors
www.eChallenges.org
Page 8 of 10
situation and the exact way of usage. But we hope to gain more insights about that in the piloting phase 2014/15. In general, the success of many specific use cases will also highly depend on non-technological issues, like the regulatory framework, the incentive systems for data exchange, etc. It might also be the case that for an operational roll-out of a WatERP-like solution, a very strong privacy and data-protection layer would have to be integrated because stakeholders might not send their data to a central system if it was not ensured that unauthorized access is impossible.
6. Conclusions The main scientific-technical contribution consists of: (1) in the engineering and the proofof-concept implementation of the WatERP overall SOA-MAS architecture which is both flexible and highly interoperable, based on standards and semantic technologies. This includes also the WMO which is the first ontology combining aspects of hydrology, technical water-supply system and management issues of water supply. (2) The WDW architecture with its different kinds of data and knowledge and different kinds of reasoning possible. This concerns, on one hand, the stability and usefulness of the selected and integrated elements (52°North SOS server, SOS protocol, WaterML2.0 data schema, OWLIM, uSeekM, etc) and, on the other hand, the way of integration (separate storage areas for measurement data and semantic knowledge). In general, we believe that such a combination of a relational database with a spatial-reasoning enabled technology is unique. There are identified deficiencies and desiderata. Technically, the prototype is not yet suitable for processing many high-frequency data streams in (near)real-time (for instance, continuous gauge values from thousands of metering points in a flood-endangered river catchment area, or continuous smart-meter data about water consumption of thousands of households). However, both pilot application scenarios are already useful when considering only several hundreds of domain objects (water-transmission points between stages in the supply chain, deep-wells, water works, reservoirs or elevated reservoirs, head pipes, pressure zones, …) and a data actualization interval in the dimension between an hour and a couple of days. Scaling-up our solution towards a more real-time and more fine-grained data-collection mode would certainly create new challenges regarding performance and software architecture. Here, recent methods from Big Data processing could be employed. Nevertheless, already the current approach offers manifold potential benefits to our end users. It is not obvious whether and when they would technically and organisationally be able to create and reasonably employ huge, high-frequency data streams. The main emphasis of the last project phase in October 2014 – October 2015 will be the piloting in the two test beds. Technically, this could raise some additional work with respect to software usability, stability and efficiency. Also, the integration of the WDW with the complex WatERP overall infrastructure and its deployment in the real-world scenarios is not completely trivial. The next interesting question will be whether the expected benefits for the pilot users can be realized in practice. Within the area of water supply, there are certainly further additional potential benefits from increased data exchange between different parties that were not yet considered in depth. Regarding the expected benefits, the energy-savings use case in SWKA seems to be relatively unique in the current water-management research. So, it will be thrilling to see whether the envisaged benefits can be realized in practice. In the long-term, there is certainly even much more potential in exploiting this “energy-water nexus” because the water-supply system can be used as energy storage for the unpredictable energy production by renewable sources such as wind and sun. This is even more interesting when seen together with the very volatile prices on the energy markets. Regarding water-savings, there are many projects dealing with leak detection, and several ones dealing with demand management; but WatERP seems to be unique in the sense that it tries to cover the whole Copyright © 2014 The Authors
www.eChallenges.org
Page 9 of 10
supply (and associated tool) chain from the predicted rainfall down to the real water usage. In the future, there will be more near real-time consumption data available through smart meters and more near real-time data on distribution-network status and operations management through telemetric sensors along the whole system. Hence, the work presented here on standards- and semantics-based data exchange and tool interoperability will probably become even more important. At the moment, there are almost no generally accepted, widely used IT standards in place for this whole area, except for (i) the OGC standards (and their predecessors) regarding geodata exchange and hydrological information that we settle upon and which must be extended thematically, and (ii) few ECadministration standards about geodata and environmental data reporting (in the INSPIRE context). Within utility companies, we have mostly commercial (more or less) de-facto standards set by big system providers. A holistic and synergetic view has probably not yet aimed at – which is understandable because of the multi-disciplinarity of such an endeavour. However, an even broader view, also encompassing the other water-related aspects and organizations – such as flood prevention, climate, biodiversity and ecologic aspects, health aspects, land-management aspects, energy aspects, etc – seems to be promising. Lastly, we expect that the technical solution of WatERP for multi-stakeholder data exchange can be applied far beyond the scope of water resource management, e.g. in the general broader scope of Smart Cities.
Acknowledgment The work presented here is partially funded by the European Commission under grant EUFP7-ICT-318603 (WatERP - http://www.waterp-fp7.eu/ ). We gratefully acknowledge the contributions of Barcelona Digital and Staffordshire University (group of Prof. Wenyan Wu) who also participate in the WDW workpackage of the WatERP project, as well as INCLAM S.A. which greatly supports all software-architectural discussions in the project.
References [1] Anzaldi, G., Chomat, C., Rubión, E., Corchero, A., Moya, A., Moya, C., Ciancio, J. and Helmbrecht, J., A Holistic ICT Solution To Improve Matching Between Supply And Demand Over The Water Supply Distribution Chain, In: SDEWES2013, 8th Conference on Sustainable Development of Energy, Water and Environment Systems, Dubrovnik, Croatia, 2013. [2] Anzaldi, G., Wu, W., Abecker, A., Rubión, E., Corchero, A., Hussain, A. and Quenzer, M., Integration of Water Supply Distribution Systems By Using Interoperable Standards To Make Effective Decisions, In: HIC-2014, 11th International Conference on Hydroinformatics. New York, USA, 2014. [3] Goodall J., Robinson B., and Castronova A., Modelling Water Resource Systems Using Service-Oriented Computing Paradigm, Environmental Modelling & Software, 2010, Vol. 26, pp. 573-582. [4] Kanwar R., Narayan U., and Lakshmi V., (2010), Web Service based Hydrologic Data Distribution System, Computers & Geosciences, Vol. 36, 2010, pp. 819-826.
Copyright © 2014 The Authors
www.eChallenges.org
Page 10 of 10
