Towards Knowledge Infrastructure for Highly Variant Voltage Transmission Systems Mathias Uta and Alexander Felfernig Abstract. The high voltage transmission business uses very breakers separate the faulty part of the grid to prevent blackouts. mature technical solutions to transport electrical energy over large Additionally, current transformers and voltage transformers give distances. New developments in the information technology sector transmission utilities the possibility to measure the transported are now promising opportunities to revolutionize the traditional energy. The collection of all main parts concentrated in one processes within the business. In this paper the opportunities to physical spot is called electric power transmission substation. implement a sophisticated knowledge infrastructure to improve the efficiency and quality of high voltage product manufacturers will be outlined. Therefore, possible solutions have been assessed to establish a non-redundant data structure and create an advanced database system architecture with respect to business specific requirements, considering in particular product configurators. Based on the proposed master data system, the possibility to create as well as integrate a knowledge system has been evaluated. Accordingly, the introduction of a global knowledge manager is proposed to organize inquiries of product configurators to expert systems and introduce a company-wide framework for rules and constraints. To assure communication between all parts of the software architecture, the implementation of a universally understandable format is discussed. Finally, possibilities to integrate recommendation system mechanisms into the suggested system architecture are highlighted. 1 INTRODUCTION The market of high voltage transmission business is highly competitive. The differentiation through up to date processes, high quality and fast response times is highly desirable to manage the Figure 1. Electrical components in the energy grid [3] increasingly complex and fast changing customer requirements. This effect is intensified by general technical developments in Since each of the main modules has to correspond with many measurements initiated by the energy transition from fossil to diversified customer requirements, the big suppliers in the business renewable energy sources. such as Siemens, ABB or Alstom have separated their energy Due to an increasing share of renewable energy into the grid and transmission business in many smaller segments [4] [5]. Each of their decentralized physical arrangement compared to huge power these segments is concentrated on one of the main components in plants, an extension of the electrical grid has to be provided. the grid. Customers expect suppliers to offer solutions for Additionally, the weather dependency of the new energy sources specifications including a complete electric power transmission intensifies this effect. As a result, the utilities are confronted with a substation. As a result, very complex projects with many internal fast changing energy market and need to react to the new energy stakeholders have to be managed. This includes the decentralized mixture [1]. Not only a transformation has to be managed on the data storage of in many cases identical information. In each of the supplier side but also consumers are starting to adopt new involved business segments a unique tool infrastructure is installed technologies with huge electrical energy demand like electrical and requires varying processes. In particular it is possible that, each cars. Therefore, the only solution to assure a secure energy supply business segment uses one or more individual configuration in the future seems to be an extension of the transmission grid [2]. solutions to determine the bill of material derived from individual The structure of the electrical grid and the main components are configuration rules and constraints. These configuration rules are shown in Figure 1. The renewable and non-renewable based power cross segmental on a higher level and do not differ until a finely plants convert several energy sources to electrical energy and are granulated level of configuration is reached. Furthermore, connected via transformers to the grid. The overhead lines and the individual pricing and quotation tools are used which are cables transport the energy over large distances to consumers. The sometimes integrated with the configuration tool, resulting in so switchgears integrate many of these lines on one conductor called configuration prizing and quotation solutions (CPQ). As a (busbar) and build nodes in the grid. In case of failures circuit consequence, very complex processes and redundant actions are decreasing competitiveness in a cost driven market environment. 2 DATABASE EVALUATION Additionally, data and knowledge maintenance cause disproportional effort leading to human and hardware resource Data modelling plays an important role. To create a ubiquitously binding. applicable database, one has to consider a data structure which To tackle the above described challenges, the introduction of a allows multiple views and possibilities to manipulate the stored well-designed data and software infrastructure is necessary. A data via interfaces from manifold applications. Graeme C. et al. completely integrated data and software landscape which provides characterize this situation as follows ‘The data model is a relatively necessary information via interfaces to all quotation related tools small part of the total systems specification but has a high impact should provide synchronization of all involved stakeholders and on quality and useful life of the system’ [6]. Data modelling is improve the maintainability of data and knowledge. concentrated on achieving completeness, non-redundancy, and In an environment which utilizes several ERP (enterprise resource reusability in databases. Additionally, stability, flexibility, and planning), PDM (product data management) and CRM (customer performance have to be taken into account. The implemented relationship management) tools, in future one company-wide database should be capable of modifications and extensions to integrated database should be implemented to deliver one data integrate requirements arising in the future. Nonetheless, the data source for all configuration, pricing and quotation tools. The goal modeler has to start with the given information to create a so called is to enter the information only once into the system so that the ‘conceptual schema’ which can be accomplished by different characteristic is passed on to all connected business segments and course of action like the process-driven, data-driven, or object- their CPQ tools. In consequence, all related objects rely on the oriented approach to name only the most important ones. Based on same data input. This assures cross segmental consistency of data the ‘conceptual schema’ a ‘logical schema’ can be deducted by in all offer documents, technical descriptions and technical using entity-relationship modelling (ERM) or unified-modelling calculations. language (UML) approaches. Finally, a physical design of the Moreover, the integrated representation of data delivers the database can be created. opportunity to implement consistent connections between objects Typically used database types are relational databases (RDBs), in form of rules and constraints. This leads to the extension of data object-oriented databases (OODBs) and a hybrid form which is to knowledge and should support users by preventing false called object-relational database (ORDB). Consistency in RDBs is configurations and incorrect data input, i.e., to contract relevant achieved by using normalization methods for tables as proposed in documents. These relations between the objects need to be non- [7]. The objective of OODBs is to create an abstract view on the redundant as per the data itself. Supplemental to the rules and reality which is easily adaptable to object oriented programming constraints describing simple relationships, interfaces to expert languages. Dependencies between objects are provided by pointers systems capable of very specific and complex calculations need to which allow m:n relationship representations. Therefore, OODBs be established. Mechanical calculations due to earthquake are often adapted for computer aided design (CAD) or technical requirements or ferroresonance calculations of the grid are two calculation programs (expert systems) since these programs deal examples of these expert systems. usually with high complexities and many variants [8] [9]. Finally, based on the achieved integrated data and knowledge, an Unfortunately, OODBs lack query performance compared to analysis of the utilized materials, parameter characteristics and all RDBs, which is why they are still not able to replace traditional other objects during the configuration, prizing, and quotation relational databases for query focused applications [10]. ORDBs process can act as a basis for a self-learning system to improve the try to combine both approaches to achieve an improved CPQ-tools. This system should be able to recommend technical performance and are often adapted by CPQ-tools. Michael solutions and parameter input based on previously chosen solutions Stonebreaker evaluated all three options in [11] and classified them in distinctive situations as customer specific requirements or as depicted in Table 1. The table points out ORDBs combine the special environmental conditions. The user in this scenario still has advantages of RDBs and OODBs in comparison of the most the opportunity to neglect the recommended solution which is important properties for databases, ‘fast queries’ and ‘complex data consequently used to improve the recommendation logic. management’. But ORDBs implicate some issues as well which is Overall, a complex system composition consisting of domain most importantly a low performance in web applications [10]. specific and integrated databases with direct interfaces to Table 1. A classification of database management systems (DBMS) [11] configuration, pricing, and quotation tools has to be established. Further, the user of the quotation tools is supported by knowledge bases inheriting logical connections between relevant parts and an adaptive artificial intelligence which analyses user decisions. The remainder of this paper is organized as follows. In section 2 we discuss the possibility of a holistic database and point out that different use cases lead to contradicting database types. Conducted from this perception we propose system architecture assuring data consistency for the high voltage transmission business. Based on this data consistency section 3 emphasizes the advantages of knowledge integrity. This is followed by a general survey on an approach to implement a global knowledge manager on basis of the proposed system architecture in section 4. Section 5 suggests a A very detailed analysis of the given requirements is precondition concept of a recommendation system in which the knowledge base for implementation of the most appropriate database and data is constantly extended using the results of configuration structure. Since full integration of the whole value chain, including applications used by experts. PLM (product lifecycle management), ERP (enterprise resource 3 SYSTEM ARCHITECTURE planning), CRM (customer relationship management), configuration, pricing, quotation and technical calculations with The data used along the value chain can be described most likely as expert systems, has to be realized, a solution with one centralized ‘master data’ of a company. Master data is defined as ‘data held by database seems with respect to deviating requirements difficult to an organization that describes the entities that are both independent implement. Table 2 shows adapted from Alejandro Vaisman’s et al. and fundamental for that organization, and that it needs in a holistic comparison between these tools and their resulting reference in order to perform its transactions’ [13]. By other utilized database types [12]. means, master data is all the data stored about customers, employees, products, materials, suppliers. The management of this Table 2. Requirement comparison between ERP, CRM, PDM, CPQ and data is called master data management (MDM) and can be expert systems provided by mainly four different architectures as shown in Figure Aspect ERP / CRM / CPQ Expert system 2. PDM User type Operators, Customer, sales Engineers office employees employees Content Current, Current, Current, detailed data detailed detailed knowledge knowledge Data According to According to According to organization operational operational and operational needs analysis needs needs Data structures Optimized for Optimized for Optimized for small complex queries complex queries transactions Access High High Medium frequency Access type Read, insert, Read, insert, Read, insert, update, delete update, delete update, delete Number of Few Many Many records per access Response time Short Short Can be long Concurrency High High From medium level to low Update High From medium From medium Figure 2. Classification of architecture approaches [14] frequency to low to low Data Low Medium Can be high (no redundancy (normalized (normalized normalization The architecture in this approach is classified by two major tables) tables and methods and categories: objects) inheritance) ∂ Are global master data attributes defined or has each Resulting Relational Object Object involved application its own master data definitions? database type database relational oriented ∂ Where is data created and maintained – in a (RDB) database database centralized system or decentralized in each connected (ORDB) (OODB) application? In the centralized master data system (1) all global master data Whereas ERP, CRM, and PDM tools are designed for fast access to attributes are created and maintained in one database. Associated distinctive data to satisfy operational queries, CPQ and expert applications receive these attributes directly from this centralized systems are focused on very complex queries that require system via identical primary keys whereas additional specific combinations between many data sources. Additionally, CPQ-tools attributes can be defined individually in the single applications. are facing a high access frequency with expected short response The leading system methodology (2) instead has no global master times which leads to an even higher level of complexity in the data attributes but a system which is responsible for data creation and structure. maintenance. These attributes are transmitted to connected In conclusion, based on these manifold requirements it seems more applications where the primary key of the attribute is mapped to the likely to implement a sophisticated system architecture including corresponding attribute in the application. several customized databases for each application instead of Decentralized data creation and maintenance approaches on the establishing one ubiquitous database. The system architecture of other hand rely on company-wide standard definitions (3) or an such a conglomeration of databases is closer examined in the next implementation of a repository (4). Company-wide standards do chapter. not assure consistency in each dataset of the applications. Instead they focus on the ability to create the same understanding of each attribute which leads to easier manageable comparisons between those datasets. A physical connection between applications is in this approach not mandatory. The installation of a repository In another scenario a company-wide leading system located in at a follows the idea of metadata storage to connect applications with distinct place is established to create and maintain all master data. each other whenever an interaction is necessary. Therefore, This leading system is additionally used as basis for connected primary keys of all involved applications for a certain attribute are central master data systems installed in distributed factories of the stored in the repository and matched for cross-program query company. These centralized master data systems receive the events with each other. company-wide master data from the leading system and can be In case of the high voltage transmission business, an extended with manufacturing side specific master data as basis for implementation of one integrated MDM solution seems, equivalent connected applications. to the result of the second chapter for databases, not feasible. Two reasons are mainly responsible for implementation of a more complex structure. First of all the structure of the business includes, as already pointed out, several different independent departments responsible for distinct parts of the substation. These departments are not necessarily located at the same location but are distributed over the world to assure an increased satisfaction of locational requirements. Consequentially, one centralized system located, for instance, in Europe will lead to performance issues when it comes to queries from an application located in Asia. Live connections over long distances and many servers are due to the congestion control of the transmission control protocol not advisable [15]. A system which synchronizes independent and asynchronous to operational tasks is therefore preferable. Secondly, each department has sometimes not only one factory but several production sides each with their own portfolio. The main structure of the portfolio is same as per the department. The differences can Figure 4. Two layer leading master data system approach be found on a more detailed layer of the products as, for instance, in the differentiation of high-end products and cost-optimized Further combinations are imaginable as well and have to be products which serve generally the same technical requirements evaluated based on several reference values in further publications. but are differentiated in more sophisticated configuration options Based on these results and taking all requirements for the high as better operation monitoring or increased maintainability of the voltage transmission business into account finally a decision on the electrical component. From these considerations follows, a solution system architecture can be reached. Another important decision is incorporating combinations of the previously presented options the selection of the tool used as master data management system. seem more feasible to encounter the high complexity in the high The use of the ERP, CRM or PDM tool is imaginable since the voltage transmission business and causes the introduction of more most data is created in these databases. Additionally, the adaption than one MDM layer. of the CPQ-tool as main input tool for data might also be a Figure 3 shows an option where a centralized master data system is reasonable choice. implemented storing master data of one manufacturing location Up to that point we focused on data structure and system whereas the centralized database is connected to other locations via architecture neglecting the value which can be created out of a repository to assure a synchronization of all business relevant integrated, non-redundant data. The following chapters will discuss data independent from daily operational tasks. Furthermore, the possibilities given by such a dataset. standards across the whole company are defined to realize a homogenous creation of new data since each master data system is allowed to create and maintain its data by itself. 4 KNOWLEDGE INTEGRITY A well-known use case for master data management systems is the implementation of data warehouses. Operational data is put through an ‘extraction, transformation, integration, and cleansing process (to) store the data in a common repository called data warehouse’ [13]. Data in the data warehouse is in reference to W.H. Inmon saved ‘subject-oriented, integrated, time-variant, nonvolatile (to) support management’s decision making process’ [16]. This means an OODB is implemented that saves data over a long time to enable analysis of developments in the observed business. Besides the data itself metadata – data about the stored data – is stored in the data warehouse. In big organizations, data is not directly used by queries to the data warehouse. Instead, data marts are introduced as an additional layer between the user and the data warehouse. Similar to the already analyzed central master data systems data is transmitted from the data warehouse (central master data system) to the data mart (application database) which fits to the requirements of a specific department. The user of the Figure 3. Three layer repository - master data system – standards approach data mart is finally able to create reports, perform data analysis and technical realization of this requirement can be established by to mine data. introducing interfaces between the product configurators and the The restriction of this concept is already given in the definition of PDM system. However, with this step an integrated data the data warehouse. The data warehouse should ‘support administration can be introduced but the knowledge administration management’s decision making process’ [16]. The area which is still redundantly managed in each configuration tool. besides the management decisions is widely neglected in To achieve knowledge integrity - a major target formulated in the centralized data and knowledge storage is the configuration and visions at the beginning of this paper - we propose steps connatural engineering sector. While huge effort is put into preparation of to the actions necessary to reach data integrity [7] [8] and which management reports and supporting management decisions, are also related to frameworks as formulated in [19]. First a configuration tasks, technical calculations and dimensioning ‘conceptual schema’ of the knowledge base has to be created using decisions are left completely to engineers and expert departments. the object-oriented approach since several connections between the This leads to a complete dependency of the company on their team given objects have to be considered. Data in the PDM system is of experts. The problem worsens if the dependence is only on a most likely stored, as analyzed in chapter 2, in RDBs and has to be single expert. Attempts to consolidate the knowledge of addressed to objects in an OODB or ORDB. These objects have to engineering experts to more than one employee usually collapse be connected via rules and constraints on a level applicable for all because of budget and time restrictions. The experience of the product configurators to implement a ‘logical scheme’. Therefore, expert cannot be transferred to other colleagues within in short normalization methods comparable to data normalization need to period of time [17]. It sometimes requires years to reach the level be formulated. In other words, a framework for product of an expert and could result in loss of knowledge in case the configurators needs to be established ‘making the common parts expert retires. common’ [19] and making general rules and constraints applicable Another issue which is more and more visible to companies arises for all configuration tools to prevent redundant code with the implementation of more than one configuration software. implementation. Due to the generalization of rules and constraints Since the beginning of the new millennium a rising number of on a higher level, a non-redundancy of the knowledge in the companies implement product configurators to treat a phenomenon separated specialized configuration solutions is achieved and called ‘mass customization’ [18]. ‘Mass customization’ is an therefore, a centralized maintainability of the knowledge. oxymoron referring on the one hand to the growing production Redundancy detection algorithms as proposed in [20] can charges of companies to serve the market and on the other hand the consequentially be applied under these conditions to assure increasing amount of individualization requirements by customers. continuous knowledge integrity. Furthermore, the knowledge of The configuration technology relies on a subject-oriented, experts can be transferred in incremental steps into the knowledge integrated, time-variant, nonvolatile dataset similar to the data base avoiding redundancy in several tools and decreasing warehouse. Combinations of entities and their attributes are dependency on single persons in the company. Finally, a ‘physical configured according to combination and configuration rules to scheme’ of the system architecture has to be created and is create a producible bill of material. investigated in more detail in the next chapter. This technology was immediately adopted by the high voltage transmission business since the modular design of its products is predestinated to realize product configurators. Unfortunately, the 5 A GLOBAL KNOWLEDGE MANAGER degree of standardization in the business does not allow a full An initial approach to set up a system architecture analog to the switch from CAD (computer aided design) applications to product requirement formulated above was made in the paper ‘On configuration solutions. Special-purpose solutions define the high Knowledge-Base System Architectures’ and is illustrated in Figure voltage transmission market up to fifty percent. The result is an 5 [21]. The paper introduced a unit called global knowledge increasing redundancy in the configuration data and knowledge. manager (GKM) to centrally handle inquiries by other knowledge- While the CAD system inherits engineering data accomplished by based systems and data type processors (e.g. traditional databases). rules and constraints for the product with a high degree of freedom To assure common semantics a translator or interface was to support the engineering expert, the product configuration proposed. A request incoming to the GKM is first scheduled and application is designed for sales people or even for the customer optimized by using the GKM’s knowledge base and information and includes a separated engineering dataset complemented by delivered by the source of the request. The result is an access plan even more sophisticated rules, constraints and methods to prevent a stored in the GKM’s internal database. The monitor/interpreter wrong configuration. Nevertheless, a redundancy of engineering uses this plan to process the request by applying the GKM’s data and knowledge is introduced; leading to a high maintenance knowledge base rules, constraints and methods. The outcome is effort and complicated processes whenever the research and returned to the original source using common semantics. development department (R&D) releases technical innovations or This approach presumes a fully centralized inference of all the product lifecycle management department (PLM) disables the inquiries. Any information in the system is gathered in the GKM, use of certain parts of the product. scheduled and interpreted. In consequence, a centralized expert Based on these evaluations, a centralized engineering database has system is introduced responsible to create and maintain all business to be implemented as foundation for all product configurators. relevant rules, constraints and methods irrespective of the PDM systems fortunately inherit all relevant product information complexity and domain of the request. A configuration task to and can, as the major tool for PLM and R&D departments, be calculate the mechanical forces on a circuit breaker in case of an utilized as a centralized engineering database. The result is a earthquake would as well be processed as a simple request for database valid for all configuration applications which are no earnings before interest and taxes (EBIT) calculation. The longer maintained decentral by each product configurator but are consequence would be a very complex set of knowledge centrally maintained by the product responsible departments. The representations neglecting all domain specific requirements. Therefore, a master data and master knowledge approach assuring consistency between all connected tools might be a more feasible solution. Figure 6. Three layer repository – global knowledge management – standards approach In Location 1, which is in shown in Figure 7, a system architecture including two CPQ-tools and the usual databases is established. CPQ-tool 1 is responsible for very standardized configuration tasks while CPQ-tool 2 is an expert tool to configure specialized customer requirements. Simple if-then relationships which generally describe a part of the substation are maintained in the GKM. For instance: ‘If product A is chosen then the cross section of the rectangular conductor cannot exceed 2400 mm².’ The information about the product is received via an interface to the PDM database and is maintained in a fast accessible RDB Figure 5. Architecture with global knowledge manager [21] database. In the GKM two objects (product A and rectangular conductor) are described by parameters (cross section, ambient Figure 6 shows, based on the system architecture presented in temperature, body temperature, rated current) matched to the Figure 3, an overview of the system architecture we propose to information delivered by the PDM database and are connected via provide the requirements of an easy maintainable, integrated the above mentioned relationship. This general rule is system for the high voltage transmission business. A company- accomplished by a more specific rule in CPQ-tool 1 by the wide standard or framework for creation of data, rules and following relationship: ‘If the rated current exceeds 2500 A at a constraints is defined to assure semantic consistency between all ambient temperature of 35 °C then a rectangular aluminum locations. The master data system is accomplished by the global conductor needs at least a cross section of 2400 mm² to not exceed knowledge manager to facilitate data and knowledge integrity in 65°C body temperature.’ [22]. one location. While the master data management part is responsible A user of CPQ-tool 1 is in consequence not allowed to choose for the data integrity of all databases, the global knowledge product A if the rated current exceeds 2500 A and an ambient manager is comparably responsible for knowledge integrity in all temperature of 35°C is given. Contradicting to this rule, an expert connected configuration applications and expert systems. in CPQ-tool 2 is not restricted by this constraint. An expert could Additionally the global knowledge manager organizes special decide to use product A if the current exceeds 2500 A at a ambient calculation inquiries by CPQ-tools to expert tools. Therefore, a temperature of 35 °C knowing that the conductor is allowed to universally understandable format (e.g. extensible markup exceed 65 °C if the conductor is not touchable by humans. This language – XML) is introduced to enable communications between expert knowledge includes besides product knowledge as well the GKM and all connected tools. Knowledge integrity during knowledge about spatial constraints and human interactions with operation is assured by redundancy detection algorithms the product and is with its complexity not easily describable in a considering all knowledge bases of the software infrastructure. non-expert configuration tool. Nevertheless both configuration Additionally, a repository is tracking comparable data, rules and systems will lead to a correct solution. CPQ-tool 1 will lead to the constraints of the separately operating locations by defining costlier product B but is operable by the customer herself while company-wide primary keys. Changes in the master data and CPQ-tool 2 with a higher degree of freedom will deliver a cheaper global knowledge management system are synchronized solution of the configuration task but needs expert knowledge. asynchronous to operational inquiries to maintain consistency Therefore, two options for the customer are established, a slightly between all local data and knowledge sets by utilizing the costlier but very fast configuration or a very accurate solution with repository primary keys. The behavior of the system is illustrated a more time consuming configuration. in more details by the two following examples. beginning of this paper also includes recommendation mechanisms and possibilities to implement self-learning algorithms the proposed system architecture needs to be accomplished by continuative considerations. 6 CONSTRAINT-BASED RECOMMENDER SYSTEMS Recommender systems are since the 1990s an increasingly used service in mainly e-commerce applications to recommend simple products to users [23]. The user gets recommended products based on previously taken buy decisions (individual information), social background information and a knowledge base that proceeds given user input in form of determined attribute and using concerned domain and contextual knowledge. But according to Felfernigs et Figure 7. Multiple CPQ-tool system architecture al. definitions recommender systems could be used in a much more general way: ‘Any system that guides a user in a personalized way Location 2 is in detailed visible in Figure 8 and includes the to interesting or useful objects in a large space of possible options standard databases, CPQ-tool 1 and an expert system. The or that produces such objects as output.’ [24]. previously mentioned configuration task to calculate the Besides the possibilities given for e-commerce sellers, a second mechanical forces on a circuit breaker in case of an earthquake is very large field of application opens with the usage in expert given. This calculation task requires information which is not systems. Mainly product configuration applications are adopting stored in the PDM database or any other standard available master recommendation techniques which can be classified in: data. Additionally, very specific calculation methods have to be applied not handled in the CPQ-tool. An expert system is necessary 1. collaborative recommendation – relying on the choices to solve the calculation task. The calculation inquiry is sent from made previously by other users with the same social and the CPQ-tool via the universally understandable format to the demographical background GKM that interprets schedules and finally processes the request to 2. content-based recommendation – relying on choices the correct expert tool. In consequence, the GKM inherits besides made previously by the user herself generalized rules and constraints also information about the expert 3. knowledge based recommendation – relying on user tools in the system and administrates access to them. The inquiry is requirements and domain knowledge handled by the expert system using the information received via 4. hybrid forms – try to combine the other types to avoid the interface and by utilizing additional information like specific disadvantages of each stand-alone solution knowledge on the occurring earthquake forces in the concerned region to calculate the forces on the circuit breaker. The result is With respect to the high voltage transmission business the only processed back to the GKM and further to the CPQ-tool where the reasonable choice for recommender systems is the knowledge resulting values trigger rules and constraints to decide which circuit based recommendation technique since electric power transmission breaker has to be chosen. substations do not rely on choices or preferences of single users but on very detailed requirement specifications delivered by the customers (utilities). Even if the requirement specifications of utilities do not change frequently and a content-based recommendation could be possible in terms of parameter input to a configuration application an important circumstance, a unique feature of substations has to be taken into account – the environmental conditions. Unlike other typically configurable products like cars or kitchens substations are crucially impacted by the locations they are assembled. For example, the Russian ministry of electricity announces in a tender a high voltage transmission substation with the same topology (electrical circuit diagram) and electrical requirements as five years before. One might conclude that the same electrical requirements lead to the same bill of material and same physical topology of the substation. But taking the new location near the Baltic Sea compared to the previous one in Novosibirsk into account this paradigm is false. The environmental conditions Figure 8. CPQ-tool – expert system architecture including temperature, air pollution, earthquake requirements and altitude of side have an essential impact on the physical These two simplified examples are only supposed to give an arrangement of the substation. This set of conditions to be overview about the working-principle to be established with the considered can be depicted as a very complex conglomeration of introduction of a GKM and need to be further researched in constraints and is in conclusion applicable for recommendation following publications. But since the vision formulated at the systems. However, the complexity of the configuration task is ∂ Knowledge integrity – how should knowledge probably too high to be provided with all features in one product normalization methods look like to build the base for a configuration application which is why the example in the previous global knowledge manager and a framework for rules chapter has been chosen. and constraints? To achieve the best configuration solution many specialized ∂ System architecture – which system architecture allows systems have to give input to the configuration process. the highest performance, best maintainability, and Mechanical calculations, waste heat, ferroresonance and further security with respect to the requirements given by a fully calculation applications are delivering valuable information and integrated knowledge system? need to be addressed to receive a feasible solution. The expert ∂ Global knowledge manager – how should the global systems outcome is involved as input for the existing knowledge manager be designed? recommendation constraints to create a solution for the given ∂ Universally understandable format – how should a configuration task. The user of CPQ-tool 1 in the example format look like which is processible by all databases and illustrated for Figure 7 as a non-expert gets one feasible solution programs in the system, including expert systems and recommended without any repair mechanism. The solution will be product configurators? rather conservative, but is in line with the given rules and ∂ Recommender system – how can rules and constraints be constraints. On the other hand the expert using CPQ-tool 2 in the improved by analyzing decisions of experts in a product example has the possibility to neglect the recommended solution configurator to improve the recommended solutions to and choose a more efficient one. The changes are recognized by non-experts? the system and adapted to the existing constraints in the GKM. A self-learning system as formulated in the vision at the beginning of this paper is created using expert knowledge to continuously REFERENCES improve the general knowledge manager as a basis for all product [1] European Commission, Energy - Country datasheets, configurators. Basis for this mechanism is the system architecture https://ec.europa.eu/energy/en/data-analysis/country, (2018) and universally understandable format to provide communication [2] Katrin Schaber, Florian Steinke, Pascal Mühlich and Thomas and non-redundancy between all parts of the system as proposed in Hamacher, Parametric study of variable renewable energy this paper. integration in Europe: Advantages and costs of transmission grid extensions, Energy Policy, Volume 42, 498 – 508, (2012) [3] Mathias Uta, Development of a product configurator for highly 7 CONCLUSION AND OUTLOOK modular and modifiable products, Figure 1, (2017) [4] Siemens AG, High-Voltage Products, https://www.siemens.com/ The paper discussed the currently arising challenges and content/dam/internet/siemens-com/global/products- opportunities given in the high voltage transmission business and services/energy/high-voltage/high-voltage-switchgear-and-devices, highlighted the current situation of the information technology in 2016 the business. Insulated solutions due to complexity of the products [5] ABB, High Voltage Products – Business snapshot, and the world-wide distribution of the associated business http://new.abb.com/high-voltage, (2017) segments lead to redundant data storage and the development of [6] Graeme C. Simson and Graham C. Witt, Data Modeling Essentials, Third Edition, p. 10, (2005) several different product configuration solutions. On the basis of [7] E. F. Codd, A Relational Model of Data for Large Shared Data the state of the art methods to normalize data and to integrate Banks, Communications of the ACM., 13 (6), 377–387, (1970) master data solutions into the system, considerations concerning [8] Alan Radding, So what the Hell is ODBMS?, Computerworld. 29 the resulting system architecture have been made. Furthermore, the (45), p. 121–122, (1995) need to create a non-redundant knowledge set has been emphasized [9] Frank Manola, An Evaluation of Object-Oriented DBMS in addition to a normalized and integrated data set. Deducted from Developments, GTE Laboratories technical report TR-0263-08-94- this condition, the proposed system architecture was extended by 165, (1994) introducing a global knowledge manager on the same level as the [10] Ramaknth S. Devrakonda, Object-relational database systems – The master data system to create a framework for configuration, pricing road ahead, Crossroad magazine, Volume 7, p. 15-8, (2001) and quotation rules and constraints. Additionally, a universally [11] Michael Stonebraker, Object-Relational DBMSs: The Next Great Wave, Morgan Kaufmann Publishers, p. 12, (1996) understandable format was proposed to enable communication [12] Alejandro Vaisman and Esteban Zimányi, Data warehouse systems, between all parts of the system and to create the possibility to Springer, (2014) integrate expert systems via inquiries scheduled and processed by [13] ISO 8000-2:2012, Data Quality, Part 2: Vocabulary, (2012) the global knowledge manager. The working principle of this [14] Cornel Loser, Christine Legner and Dimitrios Gizanis, Master Data system was outlined by two examples. Finally, the opportunity to Management For Collaborative Service Processes, International extend the system by recommendation techniques and self-learning Conference on Service Systems and Service Management, (2004) algorithms was pointed out. [15] Habibullah Jamal and Kiran Sultan, Performance Analysis of TCP Since this paper was supposed to give only a first overview over Congestion Control Algorithms, International Journal of Computers the problem statement, several future prospects arise from the and Communications, Issue 1, Volume 2, (2008) evaluations presented. These future prospects are the following: [16] William H. Inmon, Building the Data Warehouse, John Wiley & Sons, p. 31, (1996) [17] Robert R. Hoffman, The Problem of Extracting the Knowledge of ∂ Data integrity – how can product configurators be Experts from the Perspective of Experimental Psychology, AI enabled to use data from centrally maintained databases Magazine, Summer Edition, p. 53-67, (1987) instead of using encapsulated exclusive data sets [18] S. M. Davis, Future Perfect: Mass customizing, Addison-Wesley, resulting in redundancy? (1987) [19] Aparajita Suman, From knowledge abstraction to management, Woodhead, p 87-109, (2014) [20] Alexander Felfernig, Lothar Hotz, Claire Bagley and Juha Tiihonen, Knowledge-Based Configuration – From Research to Business Cases, Elsevier Inc., Chapter 7 and 12, (2014) [21] Frank Manola and Michael L. Broadie, On Knowledge-Base System Architectures, On Knowledge Base Management Systems, Springer, p. 35-54, (1986) [22] Henning Grempel and Gerald Kopatsch, Schaltanlagen Handbuch, Cornelsen, 11th Edition, p. 603, (2008) [23] Michael D. Ekstrand, John T. Riedl and Joseph A. Konstan, Collaborative filtering recommender systems, Foundations and Trends in Human-Computer Interaction, Vol. 4, p.81-173, (2011) [24] A. Felfernig and R. Burke, Constraint-based recommender systems: technologies and research issues, In Proceedings of the 10 th International Conference of Persuasive Technologies, ICEC ’08, ACM, p.1, (2008)