View-based and Model-driven Outage Management for the Smart Grid Erik Burger, Victoria Mittelbach, Anne Koziolek Institute for Program Structures and Data Organization, Chair for Software Design and Quality Karlsruhe Institute of Technology, Karlsruhe, Germany E-Mail: burger@kit.edu, koziolek@kit.edu Abstract—The integration of renewable energy resources is Thus, the contribution of this paper is twofold: We have created challenging the traditional electricity network. To manage this, a unified model of the smart grid that makes cross-domain the smart grid has been defined as a cyber-physical system analyses possible, and in addition, we have created view types consisting of a physical component, which is the electricity grid, and a computational component consisting of a communication that implement these analyses. They offer insights into the network, metering network, and software components. Therefore, smart grid that can only be gained by combining information the smart grid can not just be seen as an electrical grid, but also from heterogeneous systems of the grid. as a system of software systems. Currently, control centers of the For the evaluation of the prototype, we have evaluated the smart grid use an outage management software system to react system using historical data of the the German electricity grid. to reported outages. In this paper, we present an extended outage management We have used the SAIDI index to demonstrate how the outage system that solves one main problem of the smart grid: software management systems shortens the outage time per person per systems of the different domains are using different standards. year. The results show that the average annual outage duration Consequently, cross-domain data exchange and analysis are diffi- can be reduced by at least 2 min 28 s if our approach is applied. cult. Therefore, we use the model-driven view-based V ITRUVIUS approach to build a unified model of the smart grid. We then II. F OUNDATIONS combine it with system stability analysis methods presented as views on the model. The result is a model-driven run-time A. Smart Grid monitoring, analysis and control framework to increase the reliability of power supply. The evaluation with statistical data of The smart grid [3] integrates modern IT infrastructure into the German power grid shows that outage time can be reduced the traditional physical infrastructure of the electricity network. with our approach. It is thus a cyber-physical system. The motivation behind its introduction are the integration of renewable energy resources I. I NTRODUCTION into the network, improvement of efficiency, and the reduction The smart grid is a cyber-physical system that spans the of transmission losses. Furthermore, new load profiles, such physical structures of the electricity network and the system of as electrical vehicles and smart homes, shall participate in software systems that monitor, control, and repair the system balancing the grid and its self-healing abilities [4–6]. in case of outages. Currently, many heterogeneous systems and Control centers manage the grid via the supervisory control standards have to interoperate to achieve the desired reliability, and data acquisition system (SCADA), which monitors and stability, and efficiency of the electricity network. Many of controls the technical processes of the electricity system. these standards are based on UML and other metamodeling Smart meters are electronic devices that are installed at the standards, or can at least be expressed as such. Thus, model consumers’ places. They measure electricity production and management systems that were originally designed mainly consumption, and are connected in the advanced metering for the development of software can be used to integrate infrastructure (AMI), which can also receive commands to information from these heterogeneous systems. shape electricity consumption. The wide area monitoring In this paper, we present a concept for an extended outage system (WAMS) connects metering data in real-time. The management system (OMS) for the smart electricity grid. The WAMS consists of Phasor management units (PMU), which extended OMS is used during run-time to monitor and control are high-precision sensors that measure electrical waves with the smart grid, as well as for simulation. The goal of the a frequency of 10 to 30 times per second. A power outage is introduction of the OMS is to increase the reliability of the grid a longer interruption in the electricity supply that can either by detecting outages and by preventing outages by detecting of be planned or unplanned. The causes for unplanned outages instabilities and imbalances. The extended OMS is realized with can be categorized into the following types [7]: the view-based and model-driven V ITRUVIUS approach [1, 2] • Atmospheric Interferences are caused by events of nature, and uses its capabilities for runtime analysis and control of the such as thunderstorms, sub-zero temperatures, avalances, electricity grid. The system is prototypically implemented in the etc. Eclipse Modeling Framework using the description languages • Internal Failures are caused by maloperation of systems, for model correspondences and view generation of V ITRUVIUS. malfunction of network internal systems and all other VT2 Control Center VT1 Legend: Operator MM1 VT View Type MM3 uses VT3 MM Metamodel MM2 view view view . . . MIR «instance-of» VT4 ModelJoin Operator Viewtypes Extended OMS operate on show information from Figure 1. The Modular SUM Metamodel concept of V ITRUVIUS Algorithms Smart Grid SUM Metamodel show information from causes that are directly related to the operations of the System Viewtypes network. • Outside Influences are not caused by network operators, «instance-of» but, e.g., by construction work, traffic accidents, etc. view view view . . . • Supply Failures/Cascading Outages are caused by a uses network other than the one controlled by a specific energy Smart Grid Software systems supplier, but have an effect on the supplier’s network. In case that despite the monitoring and control systems, a Figure 2. Concept for the extended OMS disturbance causes a power outage, an outage management system (OMS) becomes active to restore the power supply. From an architectural viewpoint, the smart grid can be represents all the information that is available about a system. seen as a system of software systems that is connected by The metamodel for this model is specific to the domain in a communication network. A large number of standards exists, which the V ITRUVIUS approach is used. It combines several which cover the various aspects of the smart grid systems. The metamodels to form a modular SUM metamodel (see Figure 1). key standards will be presented in the following. The metamodels are included non-intrusively and do not have IEC 61970/61968: The IEC 61970 standard defines the to be adapted. To express the semantic relations between the Common Information Model (CIM), which is used to describe elements of the metamodels, V ITRUVIUS defines the consist- the physical components, measurement data, control and ency description language MIR (mapping/invariant/response). protection elements, and the SCADA system. It is defined Since V ITRUVIUS is a view-based approach, all information in in UML notation. The IEC 61968 standard is an extension the SUM can only be retrieved or manipulated via specialized of the CIM for the distribution network [8]. It is also called views. A view is a special kind of model and conforms to distributed CIM (DCIM) a view type, i.e., its metamodel. For the definition of view IEC 61850: This is a series of standards for substations types and views, V ITRUVIUS uses the ModelJoin language with the purpose of supporting interoperability of intelligent [12]. V ITRUVIUS has been implemented as a prototype1 electronic devices (IED) in substation automation systems. It in the Eclipse Modeling Framework and can thus be used defines the Abstract Communication Service Interface with a with any Ecore-conforming metamodel. So far, it has been mapping to concrete communication protocols, the XML-based applied to software architecture models [13] and model-based Substation Configuration Description Language (SCL), and representations of programming languages [14]. the Logical Node (LN) model, which describes power system III. A N EXTENDED O UTAGE M ANAGEMENT S YSTEM functions [9]. In this section, we present the elements of the extended out- IEC 62056: COSEM (Companion Specification for Energy age management system that are developed with V ITRUVIUS. Metering) is the international standard for data exchange for meter reading, tariff and load control in the domain of electricity A. Concept of the extended OMS metering. It works together with the Device Language Message The Extended Outage Management System is a model-driven Specification (DLMS). Together, they provide a communication analysis and control framework, which is used during run-time profile to transport data from metering equipment to the of the electrical system to monitor and control the smart grid, metering system and to define a data model and communication as well as for simulations. The framework enables network protocols for data exchange [10]. stability analysis, grid balance analysis, failure detection and B. Vitruvius location analysis, and direct controlling interactions with the electricity system to correct faulty sections of the network. V ITRUVIUS [1, 2] is a view-based, model-driven framework The structure of the system, displayed in Figure 2, can for the mangement of heterogeneous models, i.e., models be seen from two perspectives: From an inside perspective, that are instances of different metamodels. It is based on the concept of a single underlying model (SUM) [11], which 1 https://sdqweb.ipd.kit.edu/wiki/Vitruvius, retrieved 2016-09-13 Network for the electricity system provide possibilities for the grid to Grid Control Topology Viewtype Viewtype provide new measurement data, and make system changes to Generator VT VT and keep the smart grid model of the extended OMS system always Generator Consumer up to date. Consequently, system and operator viewtypes are Monitoring VT IEC 61850 Control Viewtype defined based on the smart grid SUM metamodel, and, in case Viewtype MIR VT of the operator viewtypes, on the analysis algorithms. Instances System CIM/DCIM Control of the operator viewtypes are the result of the evaluation of Balance the viewtypes on the instance of the smart grid model. These VT (IEC 61970/61968) VT Area Analysis Viewtype views are presented to the operator and the smart grid system Viewtype MIR for them to interact with the smart grid model. COSEM Consumer VT (IEC 62056) VT Monitoring This leads directly to the structure as seen from the outside by Consumer Reachability Viewtype the operators and the electrical system. They see the complete Viewtype VT analysis framework as a ‘black box’, and never interact directly Three Phase Measure- with the models and algorithms inside, but only with the views ment Matrix Viewtype presented to them. The operators can request certain views that Figure 3. The smart grid SUM metamodel and view types for the extended they need for monitoring, analysis, and controlling purposes, outage management system which are delivered to them by the extended OMS system. The extended OMS system presents them the data and results from the smart grid system in a compact and consistently modeled developers can see the inner structure of the system, whereas view, which contains only the information that they need at that operators and existing software systems only have and outside special moment, to reduce the complexity for the operators. perspective on the system. The inner structure of the system consists of three elements: B. The Smart Grid SUM Metamodel The first one is the Smart Grid SUM Metamodel, which is The modular SUM metamodel of the extended OMS consists displayed in detail in Figure 3. Following the V ITRUVIUS of the IEC smart grid standards to model the four required approach, it is a modular metamodel that combines four elements. The necessary elements are the advanced smart meter separate elements of the electricity system: the physical network infrastructure modeled by the IEC 62056 COSEM standard, the topology, the SCADA control system, the WAMS for grid wide area monitoring system for phasor measurements from monitoring, and the AMI for customer monitoring. In the the grid modeled by a part of the IEC 61850 standard, the metamodel, these elements are modeled using the IEC smart substation control functions modeled by another part of the grid standards introduced in section II: The network topology IEC 61850 standard, and the network topology and overview, model is based on the CIM and DCIM standards; the SCADA modeled by the two standards IEC 61970 and IEC 61968, also control system uses the logical node model of the IEC 61850 known as the CIM and DCIM standards. standard; the data collected from the WAMS are modeled COSEM for the Smart Meter System: The smart meter applying the logical nodes for system measurements from the metamodel is not freely available as a digital UML model, so IEC61850 standard; finally, the AMI smart metering data are we built it manually as an Ecore model in Eclipse, based on modeled using the COSEM standard. Since these metamodels the IEC 62056 standard. An excerpt of the smart meter Ecore overlap at certain points, they are complemented by MIR model can be seen in Figure 4: The smart meter is represented correspondence rules to specify the relationships between them. by the PhysicalDevice class, which is identified by its ID. It is Together, the four metamodels of the elements of the electricity associated with the ManagementLogicalDevice class and the system and the correspondence rules form the modular SUM LogicalDevice class. The logical device class has a relation to metamodel. The framework uses an instance of this metamodel each electricity-related COSEM object class. Each COSEM as a basis for all further analysis and control actions. object class implements its interface class. For example, the The second important element are the different outage and class ElectricityValues contains the attributes for current and instability detection and location algorithms, which are used voltage measurement data of the power import and export. For for the different types of failure and stability analysis. later modeling and containment purposes, the physical device The third important element are the viewtypes, which define also references each COSEM object class. the different possible perspectives on the smart grid model. CIM/DCIM (IEC 61970/61968): To use the CIM and DCIM The viewtypes need to be differentiated between viewtypes standard as a part of the SUM metamodel, they both need to for the system operator and such for the electricity system. be in the format of an Ecore model. The CIM User Group2 The viewtypes for the operator implement the algorithms for provides the two metamodels as an integrated Sparx Enterprise balancing input and output of the network to stabilize the Architecture Model ready to download as an xmi file. These voltage, and for failure detection analysis. They define views metamodels are based on the most current release of the two on the network topology and control regions. Furthermore, standards in 2014. The CIM User Group offers the metamodel they define views for the interaction with and control of the physical equipment of the electrical system. The viewtypes 2 http://cimug.ucaiug.org, retrieved 2016-09-13 ... only been defined for the combinations IEC 61850/CIM and COSEM interface classes CIM/COSEM (see Figure 3). An example for such a rule can be seen in Listing 1, where the correspondence is defined for metamodel elements that describe circuit breakers in both the COSEM object classes ... CIM and the IEC 61850 metamodel, which has the package name “substation”. For a complete listing of all rules, we refer the reader to [15]. LogicalDevice PhysicalDevice D. Viewtypes ID: EString ID: EString name: SAPAssignmentCurrent 0..* ... The second part of the extended OMS are the viewtypes ... designed for the interaction with operators and the smart grid software systems (cf. Figure 2). Each of the three metamodels in ManagementLogicalDevice the modular SUM metamodel is exposed as a legacy viewtype 0..1 to support existing software and visualization methods, and for data exchange with the systems of the smart grid. In addition, Figure 4. The Smart Meter Ecore Model (Excerpt) nine specific view types have been defined using the declarative ModelJoin language. Each viewtype combines information of together with the open source Eclipse plugin “CIMtool”3 . With either a single or multiple sub-metamodels of the modular this tool, it is possible to browse through the models and to SUM metamodel, as displayed in Figure 3. export them as an Ecore model, which can then be directly For network monitoring, the Network Topology Viewtype imported into the V ITRUVIUS environment. defines a view on the electricity system elements and their IEC 61850 for Substation Control and PMU: Analogous to topology. The Control Area Viewtype focuses on the segment- the CIM/DCIM metamodel, the IEC 61850 logical node model ation of the electricity system into control areas, where each needs to be in the format of Ecore in order to be integrated into area is regulated by one control center. the SUM. Our Ecore-based metamodel of the standard realizes Four viewtypes have been defined to support balancing of parts 5 and 7 of the IEC 61850 standard. The metamodel the electricity grid: The System Balance Analysis Viewtype has been created by adapting an Enterprise Architect UML compares production and consumption with predicted values to model by ABB. ABB has donated this metamodel to the IEC detect fluctuations. The Generator Monitoring and Consumer technical committee (TC) 57. It is accessible as a web-based Monitoring view types observe the current situation of generat- UML model on their website.4 The web-based model can ors and consumers. They serve as a basis of decision-making be browsed, but unfortunately, it cannot be downloaded. To on how to react to imbalances in the system. After this decision, integrate the metamodel into the SUM Metamodel as an Ecore operators use the Generator and Consumer Control Viewtype model, we have rebuilt it manually in Eclipse. Since the ABB to control the production and demand of electricity. model is not completely compatible with Ecore, a few changes The Consumer Reachability Viewtype is used to detect had to be made, which are described in detail in [15]. outages. The detection algorithm exploits the fact that when a smart meter is cut off from power supply, it cannot send C. Correspondence Rules any data. Together with the network topology viewtype, it is map CIM.CIM.IEC61970.Wires.Breaker as Breaker possible to see the system in a tree structure, and to mark the and substation.substationStandard.LNNodes.LNGroupX. nodes that failed by searching for the failed consumers in the XCBR as XCBR { topology view. Thus, it is possible to detect the outage, and to when-where { Breaker.mRID == XCBR.NamePlt.IdNs locate the area of its origin. If a failure is detected, the Grid Breaker.mRID = XCBR.NamePlt.IdNs Control Viewtype can be used to stabilize the situation again. } Finally, the Three Phase Measurement Matrix Viewtype } combines measurement data from phasor measurement units Listing 1. Example Mapping Rule for Breaker and smart meters. Thus, instabilities and disturbances in the transmission and distribution network can be detected, so that The correspondence rules for the extended Outage Management outages can be prevented before they occur. For a complete System (OMS) have been defined in the MIR language of definition of all view types, we again refer the reader to [15]. V ITRUVIUS. They describe the semantic overlaps between the metamodels of the modular SUM metamodel. Since it is not IV. E VALUATION mandatory in the V ITRUVIUS approach to define rules for all binary combinations, and since there were no semantic A. The SAIDI Index as Benchmark correspondences between IEC 61850 and COSEM, rules have The overall goal of the extended Outage Management System 3 http://wiki.cimtool.org/HowToValidateCPSM.html, retrieved 2016-09-13 (OMS) is to reduce the outage duration and to lower the amount 4 http://www.nettedautomation.com/download/std/61850/uml/, retrieved 2016- of outages. To evaluate how well the system is performing in 09-13 doing this, the SAIDI index is used. Outages Outages Consumers 49 600 000 Outage Type Low Med. Outages Perc. Outages Low-Voltage 147 800 Voltage Voltage Total Outages Medium-Voltage 26 000 Atmospheric 45.1 % 66 658 11 726 78 384 SAIDI Low-Voltage 2.19 Interferences SAIDI Medium-Voltage 10.09 Internal Failures 29.51 % 43 616 7673 51 289 SAIDI Total 12.28 Outside Influences 22.59 % 33 388 5873 39 261 Table II Supply Failures, 2.8 % 4138 728 4866 A NNUAL O UTAGE AND SAIDI VALUES FOR G ERMANY IN 2014 Cascading Outages Table I O UTAGE T YPES OF AUSTRIA USED FOR G ERMANY 2014 the faulty section. In this case, the outage is repaired after 2 minutes. If this is not the case, the detection and restoration of the failure takes up to one hour. Since these two cases Definition 1. The SAIDI (System Average Interruption Dur- cannot be differentiated here, an average outage restoration ation Index) index is the average outage duration for each duration is calculated based on them. Since this evaluation is served customer. It is an indicator for the reliability of power conducted in a modern smart grid, it is assumed that the first supply of an electricity system. It is calculated as: case happens slightly more often than the second one (60:40). C Kj Consequently, the average outage restoration duration is about N j,k 30 minutes (using 7.5 minutes for the outage reporting time): SAIDI = ∑ ∑ ϕ j,k j=1 k=1 N 0.6 ∗ 2 min + 0.4 ∗ 60 min + 7.5 min = 31.5 min. Besides these, some more general assumptions are made for where C is the amount of areas the grid is divided into, K j all of the following evaluations. is the number of annual outages in area j, and ϕ j,k is the • The percentages of the types of outages from the Austrian duration of the k-th outage in area j, and N j,k is the number report will be used for the German system, since they of consumers in area j affected by outage k, and N is the total are neighbors with similar infrastructure and weather (see number of consumers in the system [16]. Table I). B. Datasets used for the Evaluation • The outage restoration duration is 30 minutes. • Based on this duration and the SAIDI values in Table II, To evaluate the extended outage management system in the the average number of people affected by an outage in the German smart grid, four main datasets are used: low and medium voltage network is 24 in the low-voltage 1) Germany: Each year, the federal electricity agency of network and 640 in the medium-voltage network. Germany releases a monitoring report about the electricity system. Besides data about production, renewable production C. Evaluation Views and consumption, it includes numbers about the reliability of power supply and the amount of outages per year. The report For the evaluation of our approach, we have analysed several is freely available and will be used as a basis to calculate the properties of a smart grid using model and the views as SAIDI index and for comparison, since it includes the SAIDI described in subsection III-D. index for the current German system. For this evaluation, the 1) Grid Balancing: The main purpose of the viewtypes report from 2014 will be used [17]. for grid balancing is to prevent outages by keeping electrical 2) Austria: Similar to the German report, E-Control pub- inflows and outflows at balance. We have recalculated the lishes the outages and disturbances statistic for Austria each SAIDI index to measure the improvement gained from using the year. The report includes data about the number and type of viewtypes. The calculation is based on the numbers from 2014 outages in Austria in 2014, together with the SAIDI index. in Germany and Austria and makes some specific assumptions: Since the German and Austrian systems are neighbours, we • The fourth outage type Supply failures and cascading will assume for the evaluation that the types of outages (as outages combines two kinds of outages, of which only introduced in section II) are the same [7]. the supply failures are of interest. For this evaluation, it 3) entsoe: To analyze balancing the grid, statistical data is assumed that they have each a share of 50 % of the about the actual and forecasted production and consumption for total 2.8 % of their appearance. This is 2069 outages in 2015 provided by entsoe is used. The dataset includes quarter the low-voltage network and 364 in the medium-voltage. hourly data for every day and can be used to detect differences • Since the views currently only focus on production and between forecasted and actual production and consumption. consumption and no models for the electricity market The duration of the system restoration after an outage is an and special forecasts are included it is assumed that the important aspect. According to literature [4, 16], two cases need views will not perform very well in preventing imbalance to be differentiated: In both cases, the detection of an outage outages that currently occur. Therefore it is assumed, that relies on customer feedback. On average, an outage is reported only 30 % of the outages can be prevented. This makes after 5 to 10 minutes. In the first case, the system has an 629 prevented outages in the low-voltage network and implemented infrastructure to automatically detect and restore 109 in the medium voltage. The total amount of outages would then be 147 151 in the low-voltage network and Outages Outages Outage Type Low- Medium- Outages 25 891 in the medium-voltage. Voltage Voltage Total The results for the low- and medium-voltage system are: Atmospheric 66 658 11 726 78 384 147151 4 4 Interferences 24 SAIDI low = ∑ ∑ 30 min ∗ 49 600 000 = 2.14 min, Internal Failures 21 808 3837 25 645 j=1 k=1 Outside Influences 33 388 5873 39 261 25891 Supply Failures, 2061 364 2425 4 4 640 Cascading Outages SAIDI medium = ∑ ∑ 30 min ∗ 49 600 000 = 10.02 min j=1 k=1 Table III Thus, the total is 12.16 min with an improvement of 0.16 min. O UTAGES OF THE TOTAL EXTENDED OMS 2) Outage Detection: The main objective of these views is to detect outages automatically and faster, since currently, this relies on customer feedback, which can take up to 10 is 2069 outages in the low-voltage network and 364 in minutes. Since in the modern smart grid with automatic remote the medium-voltage. control, it is possible to restore the power supply after an • It is assumed that the external system to compare the outage already after two minutes, it is important to reduce the values from the views with historic data has been built. time for detection in order to lower the total outage restoration • Internal failures, especially if a device is broken, cannot duration. always be prevented even if the disturbance is detected. The evaluation is based on the following assumptions: However, due to the N-1 criterion it should be guaranteed • The low-voltage networks attached to a medium-voltage in most of the cases that the outage can be prevented. But network do all have the same amount of customers. to estimate it carefully it is assumed, that only 50 % of • Each smart meter sends data once every 30 minutes. the internal outages can be prevented. • The smart meters in a medium-voltage network segment • Concerning cascading outages, since they can be prevented do not send all at the same time, but are distributed equally by separating the faulty section from the rest of the grid, over the 30 minutes. They are rotating through the low- it is assumed that 70 % of them can be prevented. voltage networks attached to the medium-voltage. The results for the low- and medium-voltage system are: • To differentiate between short temporal outages (< 1 min) 124 539 4 4 and long permanent outages, the detection time needs to SAIDI low = ∑ ∑ 30 min ∗ 49 600 24 000 = 1.81 min, be longer than 1 minute. j=1 k=1 21 909 We use the formula of [15, section 5.3.4] to calculate the 4 4 640 SAIDI medium = ∑ ∑ 30 min ∗ 49 600 000 = 8.48 min. outage detection time based on the number of customers in the j=1 k=1 640 30 min Thus, the total is 10.29 min with an improvement of 1.99 min. system: 3 ∗ 25 ∗ 640 = 3.6 min. As explained in the general assumptions, there are now two cases of outage restoration. D. The SAIDI Index of the overall extended OMS The same restoration times will now be combined with the new detection time. With 60 %, the restoration time is 2 minutes, The different views evaluated above are all used together and with 40 %, it is 60 minutes, which gives an average outage in the extended outage management system. Therefore their detection and restoration time of 0.6 ∗ 2 min + 0.4 ∗ 60 min + functionalities can be combined to combine outage detection 3.6 min = 28.8 min. This duration will become lower the higher with prevention. This will further improve the SAIDI index of the share of remote controls in the grid gets. the extended OMS, since the single improvements are added The results for the low- and medium-voltage system are: up. In order to do that, this section summarizes the single 4 147800 improvements from the previous sections and calculates a 4 24 combined SAIDI index. The main achievements are: SAIDI low = ∑ ∑ 28.8 min ∗ 49 600 000 = 2.06 min, j=1 k=1 26000 • Reduction of supply failure outages by 30 %, which are 4 4 640 SAIDI medium = ∑ ∑ 28.8 min ∗ 49 600 000 = 9.66 min. 629 prevented outages in the low-voltage network and j=1 k=1 109 in the medium voltage. Thus, the total is 11.72 min with an improvement of 0.56 min. • Detection of outages in 3.6 minutes with a total reduction 3) Instability and Cascading Blackouts detection: The main of average outage detection and restoration time to 28.1. objective of these views is to prevent outages due to instabilities • Reduction of internal failure outages by 50 %, which in the system and to detain cascading outages. It is the goal to means 21 808 prevented outages in the low-voltage net- prevent outages of two types: internal outages and cascading work and 3837 in the medium voltage. outages. To evaluate the functionality of these views, some • Reduction of cascading outages by 70 %, which are 1448 further assumptions have to be made: prevented outages in the low-voltage network and 255 in • The fourth outage type Supply failures and cascading the medium voltage. outages combines two different kinds of outages of which This results are displayed in Table III. The results for the only the cascading outages are of interest for these views. extended OMS are: 123 915 For this evaluation it is assumed that they each have a 4 4 24 share of 50 % of the total 2.8 % of their appearance. This SAIDI low = ∑ ∑ 28.8 min ∗ 49 600 000 = 1.72 min j=1 k=1 21 800 4 4 methods. This can be found several times in literature [30–33]. 640 SAIDI medium = ∑ ∑ 28.8 min ∗ 49 600 000 = 8.10 min. j=1 k=1 Since currently the mapping of signals between the CIM In total, this is 9.82 min with a total improvement of 2.46 min, and IEC 61850 standards needs to be performed manually, which is 2 minutes and 28 seconds. these papers suggest a mapping between the SCL and the This means if the extended OMS is used in the German smart CIM configuration file with ontology matching. This approach grid, the average annual outage duration for each consumer follows the same purpose like the model-driven approach could be reduced by at least 2 minutes and 28 seconds. presented in this research, however using a different method. Consequently, the analysis and control framework built can VI. C ONCLUSION indeed help to improve the reliability of power supply. In this paper, we have presented a model-driven and view- V. R ELATED W ORK based framework, called the extended outage management Related work on improving the reliability of the power supply system, for run-time analysis and control of the smart grid. using model-driven methods can be grouped in two categories: The main purpose of the system is to increase the system Approaches that treat the problem of improving the reliability reliability by faster detection of outages and by preventing of power supply, and approaches that examine the problem of the outages through the detection of system instabilities and combining information of the smart grid using model-driven imbalances. The system is built in V ITRUVIUS, a model-driven or related methods. and view based model management framework, which was There is quite an extensive amount of research about the originally created for software development purposes, but can reliability of electrical systems, also in connection with the be used with any kind of metamodel-based data. This paper smart grid. E.R. Brown [18] examines the challenge of a reliable has shown that the mechanisms and languages of V ITRUVIUS power distribution system, and D. Elmakias introduces in his for defining correspondences and viewtypes can successfully book methods to examine and improve the reliability of an be applied to domains that do not purely concern software. electrical system [19]. Chowdhury et al. [20] and Waseem et al. The framework has been evaluated using the SAIDI index. [21] focus especially on the impact of distributed generation on Using statistical data of the German power grid, a possible system reliability, since the integration of distributed generation improvement of 2 minutes and 28 seconds in annual outage is a new challenge for the power system. Related to power time. The system could be further improved by including further distribution, Russell et. al [22] work on improving the reliability data sources, such as real-time data from the energy market. of the distribution system equipment in their research. R EFERENCES Since the unification of different smart grid elements in one model has been identified early in the NIST smart grid [1] E. Burger, ‘Flexible Views for View-Based Model- roadmap [3], different approaches have been developed: Driven Development’, in Proceedings of the 18th inter- The Electric Power Research Institute (EPRI) has harmonized national doctoral symposium on Components and archi- the CIM model with the IEC 61850 substation equipment with tecture, ser. WCOP ’13, Vancouver, British Columbia, a CIM-based unified UML model that includes the physical Canada: ACM, 2013, pp. 25–30. [Online]. Available: elements of both standards [23–27]. In contrast, the approach http://doi.acm.org/10.1145/2465498.2465501. presented in this paper leaves the existing smart grid models [2] M. E. Kramer, E. Burger and M. Langhammer, ‘View- unchanged and builds mappings to combine them. The EPRI centric engineering with synchronized heterogeneous approach includes only the substation equipment from the IEC models’, in Proceedings of the 1st Workshop on View- 61850 SCL model, while our research focuses on the substation Based, Aspect-Oriented and Orthographic Software functions in the LN model. Modelling, ser. VAO ’13, Montpellier, France: ACM, Andrén et. al. use model-driven QVTo transformations to 2013, 5:1–5:6. [Online]. Available: http://doi.acm.org/ transform IEC 61850 LN models into IEC 61499 models, a 10.1145/2489861.2489864. standard for process automation. This is done to automate the [3] D. von Dollen, ‘Report to nist on the smart grid substation processes. The authors use model transformations interoperability standards roadmap’, Aug. 2009. to combine the two standards, like it is done in this research. [4] B. Buchholz and Z. Styczynski, ‘Smart grids – fun- However, they focus on those two standards and do not have damentals and technologies in electricity networks’, the goal to enhance smart grid integration [28]. English, in, Springer Berlin Heidelberg, 2014, pp. 1–17. Byunghun et. al. present in their work mappings between [Online]. Available: http://dx.doi.org/10.1007/978-3- datatypes from the CIM and the IEC 61850 standard. The two 642-45120-1%5C_1. standards can only be unified if their datatypes match. The [5] D. Mah, P. Hills, V. Li and R. Balme, ‘Smart grid IEC 61850 however defines special datatypes for the standard applications and developments’, English, in, ser. Green that need to be mapped onto the datatypes defined in the CIM Energy and Technology, D. Mah, P. Hills, V. O. Li model. For deeper research about the mapping of the two and R. Balme, Eds., Springer London, 2014, pp. 3–20. standards, this paper is of importance [29]. [Online]. Available: http://dx.doi.org/10.1007/978-1- Another different approach to map the two standards CIM 4471-6281-0%5C_1. and IEC 61850 is the use of ontologies and semantic web [6] J. DongLi, M. Xiaoli and S. Xiaohui, ‘Study on [20] A. Chowdhury, S. Agarwal and D. Koval, ‘Reliability technology system of self-healing control in smart modeling of distributed generation in conventional distribution grid’, in Advanced Power System Automation distribution systems planning and analysis’, Industry and Protection (APAP), 2011 International Conference Applications, IEEE Transactions on, vol. 39, no. 5, on, vol. 1, Oct. 2011, pp. 26–30. pp. 1493–1498, Sep. 2003. [7] E-Control, ‘Ausfall und Störungsstatistiken für Öster- [21] I. Waseem, ‘Impacts of distributed generation on the reich 2014’, 2015. residential distributed network operation’, PhD thesis, [8] ‘Iec 61970 energy management system application pro- 2008. gram interface (ems-api) - part 301 common information [22] B. Russell, C. Benner, R. Cheney, C. Wallis, T. Anthony model (cim) base’, Aug. 2011. and W. Muston, ‘Reliability improvement of distribution [9] ‘Iec 61850 communication networks and systems for feeders through real-time, intelligent monitoring’, in power utility automation’, Jul. 2015. Power Energy Society General Meeting, 2009. PES ’09. [10] D. U. Association, ‘Excerpt from companion specific- IEEE, Jul. 2009, pp. 1–8. ation for energy metering cosem interface classes and [23] E. P. R. Institute, ‘Harmonizing the international electro- obis identification system’, Sep. 2014. technical commission common information model (cim) [11] C. Atkinson, D. Stoll and P. Bostan, ‘Orthographic and 61850: Key to achieve smart grid interoperability Software Modeling: A Practical Approach to View-Based objectives’, May 2010. Development’, in Evaluation of Novel Approaches to [24] ——, ‘Harmonization of iec 61970, 61968, and 61850 Software Engineering, ser. Communications in Computer models’, Dec. 2006. and Information Science, L. A. Maciaszek, C. González- [25] ——, ‘Harmonization of cim with iec standards: Draft Pérez and S. Jablonski, Eds., vol. 69, Berlin/Heidelberg: report for cim and other iec working groups’, Mar. 2006. Springer, 2010, pp. 206–219. [26] ——, ‘Harmonization of utility common information [12] E. Burger, J. Henß, M. Küster, S. Kruse and L. Happe, model (cim) with other iec power system management ‘View-Based Model-Driven Software Development with standards’, Mar. 2007. ModelJoin’, Software & Systems Modeling, vol. 15, no. [27] ——, ‘Integration of advanced automation and enterprise 2, R. France and B. Rumpe, Eds., pp. 472–496, 2014. information system infrastructures’, Dec. 2007. [13] M. E. Kramer, M. Langhammer, D. Messinger, S. [28] F. Andren, T. Strasser and W. Kastner, ‘Model-driven Seifermann and E. Burger, ‘Realizing change-driven engineering applied to smart grid automation using iec consistency for component code, architectural models, 61850 and iec 61499’, in Power Systems Computation and contracts in vitruvius’, Karlsruhe Institute of Techno- Conference (PSCC), 2014, Aug. 2014, pp. 1–7. logy, Department of Informatics, Karlsruhe, Tech. Rep., [29] B. Lee, D.-K. Kim, H. Yang, H. Jang, D. Hong and 2015. [Online]. Available: http://nbn-resolving.org/urn: H. Falk, ‘Unifying data types of iec 61850 and cim’, nbn:de:swb:90-456541. Power Systems, IEEE Transactions on, vol. 30, no. 1, [14] M. Langhammer and K. Krogmann, ‘A co-evolution pp. 448–456, Jan. 2015. approach for source code and component-based architec- [30] R. Santodomingo, J. Rodriguez-Mondejar and M. Sanz- ture models’, in 17. Workshop Software-Reengineering Bobi, ‘Using semantic web resources to translate existing und-Evolution, vol. 4, 2015. [Online]. Available: http: files between cim and iec 61850’, Power Systems, IEEE //fg-sre.gi.de/fileadmin/gliederungen/fg-sre/wsre2015/ Transactions on, vol. 27, no. 4, pp. 2047–2054, Nov. WSRE2015-Proceeedings-preliminary.pdf#page=40. 2012. [15] V. Mittelbach, ‘Model-driven Consistency Preservation [31] R. Santodomingo, J. Rodríguez-Mondéjar and M. Sanz- in Cyber-Physical Systems’, Master’s thesis. Bobi, ‘Ontology matching approach to the harmoniza- [16] A. Avritzer, S. Suresh, D. S. Menasche, R. M. M. tion of cim and iec 61850 standards’, in Smart Grid Leao, E. de Souza e Silva, M. C. Diniz, K. Trivedi, Communications (SmartGridComm), 2010 First IEEE L. Happe and A. Koziolek, ‘Survivability models for the International Conference on, Oct. 2010, pp. 55–60. assessment of smart grid distribution automation network [32] A. Mercurio, A. Di Giorgio and P. Cioci, ‘Open-source designs’, in Proceedings of the fourth ACM/SPEC implementation of monitoring and controlling services International Conference on Performance Engineering for ems/scada systems by means of web services iec (ICPE 2013), ser. ICPE ’13, Prague, Czech Republic: 61850 and iec 61970 standards’, Power Delivery, IEEE ACM, New York, NY, USA, 2013, pp. 241–252. [Online]. Transactions on, vol. 24, no. 3, pp. 1148–1153, Jul. Available: http://doi.acm.org/10.1145/2479871.2479905. 2009. [17] Bundesnetzagentur, ‘Monitoringbericht 2014’, Nov. [33] R. Santodomingo, S. Rohjans, M. Uslar, J. Rodriguez- 2014. Mondejar and M. Sanz-Bobi, ‘Facilitating the automatic [18] E. Brown, ‘Electric power distribution reliability, 2nd mapping of iec 61850 signals and cim measurements’, ed.’, in, CRC Press, 2002. Power Systems, IEEE Transactions on, vol. 28, no. 4, [19] D. Elmakias, ‘New computational methods in power pp. 4348–4355, Nov. 2013. system reliability’, in, Springer, 2010.