SUMMARY

In the context of rising power system complexity and ongoing workforce transitions, the power sector struggles to manage, preserve, and transfer technical knowledge in Protection, Automation, and Control (PAC) systems. The retirement of experienced professionals, combined with the onboarding of younger engineers, highlights gaps in traditional knowledge management practices that rely heavily on file-based documentation and tacit expertise.

This paper proposes a paradigm shift from conventional document-centric engineering toward a data-centric, object-oriented methodology based on Engineering Information Modeling (EIM). Inspired by principles widely adopted in Building Information Modeling (BIM) for civil and electromechanical engineering. EIM provides an integrated approach to structure, manage, and share PAC engineering information consistently across the entire project and asset lifecycle.

EIM is grounded in three pillars: process, people, and technology. It emphasizes cultural transformation and governance to ensure consistent and reliable use of structured data, enabling organizations to manage not just documents, but interconnected, traceable engineering objects. The technology and process pillars jointly support a top-down engineering methodology in which standardized and validated templates enable the automated generation of PAC projects, thereby improving consistency and significantly reducing engineering time and error rates.

The paper highlights the role of structured taxonomies and engineering ontologies implemented within database-driven platforms. These semantic structures enable the formal representation of PAC-specific concepts, such as protection functions, logical schemes, communication architectures, and device settings, ensuring consistency, traceability, and intelligent cross-referencing throughout the lifecycle.

To overcome the burden of legacy data, the paper presents the application of artificial intelligence techniques for the automated extraction of engineering metadata from unstructured sources, including legacy documents and engineering diagrams. By identifying elements such as equipment models, revision history, and diagram descriptions, AI-assisted extraction accelerates the population of structured databases and substantially reduces the effort required to establish a usable and centralized digital knowledge base. Extracted data undergoes human-in-the-loop validation to ensure accuracy, mitigating error risks.

The case studies demonstrate that the adoption of a centralized, object-oriented engineering platform leads to measurable improvements in efficiency, traceability, and reliability of PAC engineering activities. Tasks such as information retrieval, change identification, and project generation were executed significantly faster and with greater consistency when compared to traditional document-centric workflows. Results show substantial reductions in engineering time, while confirms improved accuracy, reduced ambiguity, and faster access to critical technical information, particularly in fault analysis and substation modernization contexts.

Beyond immediate productivity gains, the results indicate that EIM provides a robust foundation for systematic knowledge preservation and workforce transition. By combining structured data models, transparent object histories, guided engineering templates, and AI-assisted tools, engineering knowledge becomes explicit, auditable, and reusable across projects and generations. This integrated paradigm supports scalable management of both legacy and new PAC data, enables reliable AI-assisted interaction with engineering information, and contributes to long-term consistency, traceability, and resilience in highly digitalized power system environments.

1 Introduction

Increasing power system complexity and workforce transitions challenge the management and transfer of Protection, Automation, and Control (PAC) engineering knowledge. This knowledge spans several information categories, including manufacturer data and diagrams for primary and secondary equipment, protection, control, and supervision schematics describing electrical circuits and logic schemes, communication architecture documentation, configuration files such as IED, SCADA, and IEC 61850 System Configuration Language (SCL) artifacts, as well as protection setting tables and coordination reports. Despite the availability of standards-based information models for certain categories of PAC information, the heterogeneous information elements described are still predominantly handled as static and loosely coupled knowledge containers. In typical organizational practices, schematics are maintained as PDF or DWG files and spreadsheets, configuration backups are stored as proprietary files, reports are archived in document formats, and SCL files remain isolated XML artifacts. These materials are commonly stored in shared network drives or document management systems (DMS) and organized according to pragmatic criteria, rather than semantic relationships.

Within this document-centric context, revision and change management is usually implemented through a heterogeneous combination of local or cloud-based repositories and document-oriented practices, and, in some cases, supported by dedicated DMS platforms. In simpler scenarios, versioning relies primarily on file naming conventions (e.g., “REV_A”, “REV_B”), email exchanges, or manually maintained change logs, offering limited support for systematic traceability, semantic consistency, and cross-artifact impact analysis. More mature organizations may adopt cloud-based DMS solutions that store revision codes as document attributes and enforce formal workflows for document approval and coordination, representing a clear improvement over local and non-standardized practices. Even when supported by DMS solutions, revision control often remains limited to file-level versioning, offering little insight into what exactly changed, when and by whom the changes were performed, or how modifications propagate across dependent drawings and equipment configurations. Over time, document-centric practices lead to fragmented knowledge structures, increased dependence on tacit expertise, and heightened organizational vulnerability to workforce turnover [1], [2]. This fragmentation directly impacts engineering efficiency, reuse of validated solutions, onboarding of new professionals, and long-term system resilience [3], [4].

To mitigate these challenges, this paper proposes a paradigm shift from conventional document-centric engineering toward a data-centric, object-oriented approach based on the Engineering Information Modeling (EIM) methodology. Inspired by principles widely adopted in Building Information Modeling (BIM) for civil and electromechanical engineering, EIM adapts these concepts to the PAC domain. Although the term BIM can, in principle, be applied to structured information management even when not associated with civil and electromechanical engineering, this paper adopts the term EIM to clearly denote the broader engineering scope addressed in PAC systems. While EIM is less widely established than BIM, its use is consistent with terminology and concepts already discussed in the literature [5], [6]. This shift aligns with recent CIGRÉ work highlighting the evolution of PAC architectures toward increased abstraction, virtualization, and lifecycle awareness [3], [2].

While BIM initially emerged with a strong emphasis on 3D modeling and spatial coordination among civil construction disciplines [7], its most significant foundations go beyond geometry and align closely with the needs of disciplines involved in PAC systems. In PAC systems, the primary challenges relate to functional architecture, electrical topology, communication models, control logic, and configuration consistency rather than spatial representation [5] [8]. The first BIM principle incorporated into the EIM methodology is data-driven modeling, where engineering elements, such as IEDs, transformers, panels, networks, IEC 61850 functions, and control logics signals, are represented as objects within a unified database carrying attributes, relationships, and governing rules. Changes to any object attribute automatically propagate across dependent artifacts, including diagrams, SCL files, cable lists, and reports, ensuring consistency through a Single Source of Truth (SSoT).

Interoperability is another shared foundation. While BIM relies on standards such as IFC, PAC systems depend on IEC 61850 and CIM. In both domains, standardized, machine-readable representations of objects and relationships are essential to support multi-vendor environments and complex system integration. Multidisciplinary collaboration is equally critical. A centralized information model enables coordinated work across disciplines, reducing inconsistencies and rework. The concept of a Common Data Environment (CDE) supports version control, structured change tracking, and traceability, key requirements in PAC systems, where unintended configuration changes can directly affect operational safety. Just as BIM prevents conflicts between architectural and structural designs, EIM applied to PAC systems prevents inconsistencies between logic, wiring, communication, and IED parameterization. Reuse and typification, widely applied in BIM through standardized libraries, are also fundamental to EIM. Typical IED configurations, protection schemes, and logic blocks reduce engineering effort and error rates. Finally, lifecycle management, strongly emphasized in BIM dimensions 6D and 7D, is equally relevant in PAC systems, as engineering data must remain accessible, traceable, and valid throughout decades of operation, maintenance, and system evolution [7].

2 EIM Pillars

2.1.1 Processes

EIM replaces fragmented, document-based practices with structured, data-centric engineering workflows built on a shared information model. Engineering activities are governed through explicitly defined tasks, roles, validation steps, and revision rules. Workflow execution, revision control, access rights, and status transitions are managed directly within the platform, ensuring that changes follow standardized procedures and that only validated data is approved.

Object-level revision history links each change to its affected elements, enabling transparent impact analysis, audits, and reliable engineering reviews. Because workflow events and revision metadata are stored in a structured database, the same data can be reused for monitoring and analytics, improving project visibility and decision-making without additional reporting effort.

2.1.2 People, Governance, and Organizational Adoption

EIM adoption depends on organizational alignment in addition to technology. PAC engineering involves multiple stakeholders across long lifecycles, requiring coordinated creation, exchange, and governance of engineering data. Within this collaborative ecosystem, equipment manufacturers (OEMs) play a key enabling role. By providing standardized and structured device data models, OEMs facilitate integration into engineering platforms, reduce manual data handling, and improve consistency across engineering, commissioning, and maintenance activities. From a governance perspective, EIM requires the formal definition of roles, responsibilities, access rights, and approval authority throughout the engineering and operational lifecycle. These governance rules are enforced through technology, ensuring that engineering information is updated in a controlled, auditable, and traceable manner, and that a single approved version of the data is maintained as the authoritative reference. Organizationally, EIM shifts the focus from document ownership to data stewardship, supported by training and clear procedures. This socio-technical alignment enables consistent engineering practices, improved collaboration, and sustainable knowledge management across the PAC lifecycle.

2.1.3 Technology

The technology used to support the EIM methodology presented in this paper is a collaborative Engineering Platform (EP). Built on a relational database and based on an object-oriented modeling approach, the EP consolidates technical data, diagrams, and functional representations into a cohesive digital engineering model. Assets such as protection devices, panels, cables, and control logic signals are modeled as interconnected objects with defined attributes and change-tracking. Beyond data storage, the platform actively manages engineering information, supporting structured queries, standardized templates management, and automatic generation of artifacts such as electrical diagrams and IEC 61850 SCL files. Change tracking and cascading updates preserve consistency across revisions and disciplines, while improving traceability, impact analysis, and long-term knowledge preservation.

Within this technological context, PAC engineering information is represented through an object-oriented data model implemented through an EP based on a centralized relational database. Within this platform, engineering information is organized as an object tree composed of several folders, among which three represent the core modeling domains: Equipment, Systems, and Documents. Each of these areas allows the creation of different types of objects, each associated with a specific set of attributes and master forms. The Equipment folder is intended for modeling physical assets and devices, while the Systems folder is used to define equipment functions, as well as to build the data model in accordance with the IEC 61850 standard. Finally, the Documents folder centralizes all engineering documentation linked to the modeled objects.

Within this context, it is possible to create and define any type of equipment directly in the Equipment folder. Depending on the defined equipment type, the attributes associated with the object may vary, enabling the construction of a data structure compatible with the technical and functional characteristics of each asset. Instead of treating elements as purely graphical entities, this approach considers each asset as an object with properties, relationships, and rules, allowing the developed graphical representations to be traceable anywhere in the design. Figure 1 shows the hierarchical structure of the database and the objects created related to the drawing.

In practice, this is implemented as a functional hierarchy that links each diagram element to an intelligent object model. Figure 2 illustrates the modeling of a Merging Unit, including its attribute set, its position in the parent–child object tree, and the semantic links that connect this secondary device to the transformer-bay supervision functions.

It is possible to observe that the objects are organized in parent-child relationships that represent both physical composition and functional linkage. Thus, objects can be defined to represent a substation, voltage levels, bays, panels, and other devices. In addition, it is noted that each modeled object has functions and attributes, such as type, operational data, specifications, etc.

In this way, equipment ceases to be merely a symbol on a drawing sheet and becomes an instance of a class, which may contain sub-objects corresponding to logical nodes. This perspective enables the object inserted on a drawing sheet to be traced in any drawing or sheet where it is located, in addition to enabling version control, since automatic change markers are inserted into the drawings when an object is added, removed, or modified.

To ensure consistency and scalability across projects and suppliers, the platform manages structured taxonomies and engineering ontologies to support a comprehensive PAC information model. This standardized data management can be grounded in international standards such as IEC 61850 and CIM, which provide complementary reference models for PAC systems. IEC 61850 defines standardized representations for substation topology, communication, and functional allocation of IED capabilities [9], while CIM addresses system-level electrical topology and asset relationships across the power system [10]. Despite their relevance, both standards allow flexibility in implementation depth and interpretation, so the EP can support utilities by providing tools for the governance of standardized implementation profiles. In addition, a significant portion of PAC engineering information, including cabinet wiring, secondary circuits, and detailed control logic, is not fully covered by either standard. With EIM and an EP, a wide range of engineering data can be modeled in accordance with international standards or, when necessary, through the definition of domain-specific taxonomies and ontologies, enabling utilities to consistently model and govern PAC information across projects, suppliers, and the system lifecycle.

Taxonomies ensure consistent classification and hierarchy, while ontologies define semantic links between assets, functions, and signals. For example, an IED object is not only associated with its graphical representation defined by templates, but also with its hardware structures (physical terminals, communication ports, binary inputs (BIs), and binary outputs (BOs)), and with its logical functions, IEC 61850 data model elements, protection functions, GOOSE publisher and subscriber signals, and Sampled Value measurements. These relationships link cabinet wiring details (e.g., pins, terminals, jumpers), functional logic representations (e.g., interlocking and protection schemes), and communication architecture elements (e.g., network topology, VLAN configuration, redundancy schemes).

Data requirements are enforced through mandatory attribute sets associated with each object type, ensuring that engineering information is complete. For instance, an IED object must include attributes such as manufacturer, catalog number, operational identifier, protection-related parameters, and communication settings, while communication infrastructure elements require attributes such as IP address, VLAN assignment, port speed, and connector type. These structured requirements enable the automated generation of downstream engineering artifacts, including cable lists, IP and VLAN lists, wiring connections list, and IEC 61850 SCL files, directly from the underlying database. Together, structured taxonomies, engineering ontologies, and explicit data requirements transform PAC engineering from a document-centric activity into a coherent, database-driven information model that supports scalability, traceability, and lifecycle governance in large and highly digitalized utility environments.

Based on this object model, engineering information can be accessed through semantic, object-based querying rather than document-oriented searches. Semantic queries operate directly on the underlying object model, allowing engineers to retrieve and correlate PAC information based on attributes and relationships. These queries are typically exposed through customizable tables, where selected object attributes are presented as columns and object instances as rows. Within these dynamic tables, users can apply attribute-based filters and navigate directly to related objects and graphical representations. Typical use cases include retrieving terminals and wiring connections, listing devices and signals by bay or panel, identifying GOOSE publishers and subscribers, and opening all diagrams associated with a selected object. For example, a communication ports table can display all network interfaces with attributes such as IP address, VLAN, port type, and redundancy role. By filtering specific IP or VLAN values, users can navigate directly from a selected port to the diagrams representing its physical connections and communication topology. Similarly, pin and terminal tables allow filtering by a specific identifier and navigating to all associated wiring connections and diagrams, enabling fast and unambiguous verification.

Figure 3 illustrates the modeling of bay 04T8 and its associated panels and equipment in the EP. From panel “2UA3S”, the Devices table provides a consolidated view of all device attributes linked to that panel. In this example, a semantic filter was applied to identify the lockout relay (86), which was located within seconds. From the selected object, users can navigate to its representation in functional, layout, topographic, and logical diagrams, ensuring consistent and synchronized access to engineering information across multiple views.

Beyond information modeling and querying, the platform also supports top-down engineering through guided, template-based project generation. In this approach, corporate standards and best practices are formalized as structured engineering models that actively guide project development, reducing reliance on individual experience and minimizing recurrent errors. Engineering decisions are embedded in templates, ensuring consistency and regulatory compliance while supporting knowledge transfer, particularly for less experienced engineers.

Within this context, the EP assistants Advanced Typical Manager (ATM) and Project Builder (PB) structure validated engineering knowledge into parameterized and governed typicals. These typicals encapsulate protection logics, automation schemes, naming conventions, and IEC 61850 modeling rules as reusable and version-controlled components. Instead of copying existing designs, engineers configure projects by selecting governed alternatives explicitly modeled within the typicals. For example, different functional roles of a current transformer core are represented as parameterized configurations associated with explicit engineering rules. Once selected, the platform automatically propagates the corresponding changes across diagrams, circuits, functional associations, and technical lists, preserving consistency across all dependent artifacts. This technology supports full top-down project generation by guiding engineers through a structured sequence of design decisions. Starting from high-level definitions, such as substation topology, voltage level, and bay type, PB progressively constrains compatible equipment, protection philosophies, and communication architectures. These choices directly shape the associated IEC 61850 model and related engineering artifacts. By embedding tacit engineering knowledge into guided and traceable workflows, PB reduces dependency on individual expertise and enables consistent, standards-compliant PAC project development across different experience levels.

3 Case Studies

3.1 Semantic and Object-Based Querying in PAC Engineering Databases

To quantitatively evaluate the benefits of object-oriented queries, a case study was conducted with 93 participants, including protection, automation, and commissioning engineers, all with prior experience in substation projects. Participants were asked to perform a standardized set of information-retrieval tasks representative of daily PAC engineering activities, such as: “How many spare BIs exist in the F1 protection IED?”, “Which project page presents the breaker failure logic (50BF)?”, “What is the destination of the wire connected to pin 1C4 of F1?”, etc. In the first scenario, searches were performed exclusively using traditional documentation, including PDF files and printed drawings. In the second scenario, the same tasks were executed using an EP based on a centralized, object-oriented database. The results show a clear quantitative difference. In the document-based scenario, the average time required to locate the requested information was 76 seconds, with variations of up to 50 seconds among participants, reflecting differences in individual experience and familiarity with the project. In several cases, incorrect answers were observed, mainly due to misinterpretation of diagrams. When using the EP, the average time to complete the same tasks was reduced to 22 seconds, with a variation of approximately 10 seconds. Beyond time reduction, the results demonstrated higher consistency and reliability. This case confirms that semantic, object-based querying substantially improves both efficiency and correctness in PAC engineering information retrieval.

3.2 Transparent Object Histories and Change Traceability

In traditional document-centric workflows, tracking modifications typically requires manual comparison of multiple document versions. This process is time-consuming, error-prone, and highly dependent on the memory of senior engineers, especially when changes span several revisions or long time intervals. In the evaluated EP, change management is supported by dedicated assistants for data tracking and revision control. All modifications are automatically recorded at object and attribute level, including the modified parameter, previous and new values, author, and timestamp. Instead of searching and comparing documents, engineers can directly query the object history to identify what changed.

From a time-comparison perspective, the difference is substantial. In document-based workflows, engineers often require several minutes, and in complex cases even hours, to locate, compare, and interpret changes across revisions. With object history and data tracking, the same analysis is completed in seconds, with changes explicitly highlighted at parameter level. This eliminates ambiguity by clearly identifying which parameters were modified, when the change occurred, and how the current configuration differs from previous states. Beyond object-level history, revision assistants support graphical comparison of engineering diagrams, highlighting graphical additions, removals, and modifications between revisions. This enables consistent tracking of both data-centric changes and their graphical representations, ensuring coherence between diagrams, configurations, and underlying objects throughout iterative project cycles, commissioning, retrofitting, and maintenance activities.

Transparent object histories transform tacit engineering knowledge into organizational memory by preserving technical decisions over time, enabling junior engineers to understand past changes and reducing dependency on individual experts. Overall, the case study shows that object-level data tracking and revision assistants substantially enhance traceability, enabling auditable and queryable PAC data that supports knowledge transfer, safer decision-making, and long-term system maintainability.

3.3 Top-Down Project Generation Using ATM and Project Builder

This case study evaluates the practical application of top-down, template-based engineering through the combined use of the ATM and the PB within an EP. The assessment starts with an empty project, containing only the default folder structure provided by the platform. With the execution of the ATM and PB assistant, the engineer initiates the project by selecting the bay (cubicle) type. Once the parameters are selected, the assistant automatically instantiate devices, generate functional diagrams and create object relationships. As a result, complete technical data containing approximately ten pages of functional diagrams was generated automatically. The entire project generation process was completed in less than three minutes, with full adherence to the predefined engineering standards embedded in the typicals. In conventional document-centric workflows, producing an equivalent set of diagrams and ensuring internal consistency would typically require several hours of manual work, iterative reviews, and cross-checks across multiple tools. This case study demonstrates that the combined use of ATM and PB effectively operationalizes top-down engineering by embedding expert knowledge, standards, and design rules directly into guided templates. The approach significantly reduces engineering time, minimizes dependency on individual expertise, and enables scalable, repeatable, and standards-compliant generation of PAC projects, particularly in large utility environments and programs with high project replication.

4 AI-assisted Legacy Data Structuring

Although object-oriented methodologies and international standards support the generation of structured data in greenfield projects, most operational power systems remain dominated by brownfield infrastructures, embedded in heterogeneous and unstructured legacy documents. To address this challenge, this paper presents an AI-assisted methodology for extracting and structuring engineering metadata from legacy PAC documentation. This controlled AI-assisted approach aligns with findings reported in the literature across multiple engineering domains. Prior studies demonstrate the effectiveness of deep learning and graph-based reasoning for digitizing engineering diagrams [11], and named-entity recognition techniques for transforming unstructured technical texts into machine-interpretable data [12]. While recent power system studies emphasize legacy documentation as a persistent barrier to interoperability and advanced analytics [13].

The approach presented targets key information categories, including substation and bay identification, voltage levels, equipment models and manufacturers, document identifiers, and revision histories. The proposed solution adopts a hybrid AI architecture designed to handle the variability and degradation typical of legacy documentation. The pipeline is structured into preprocessing, processing and inference, post-processing, and validation stages. It combines OCR, supervised machine learning, natural language processing, rule-based logic, and Large Language Models (LLM) based contextual interpretation to extract metadata from scanned drawings and technical documents. Given the safety-critical nature of PAC systems, the approach explicitly avoids fully autonomous data integration. Instead, a human-in-the-loop (HITL) validation layer is applied prior to database registration, following established principles of interactive machine learning. Expert review is supported by an interactive validation interface and complemented by statistical sampling procedures in accordance with NBR 5426 and NBR 5427, ensuring reliability, transparency, and auditability.

Within a Transmission System Operator (TSO) implementation, the methodology was evaluated across multiple PAC-relevant metadata categories using similarity, accuracy, precision, and recall metrics. The results demonstrate robust performance across metadata types. Substation name extraction achieved a similarity of 0.99 and accuracy of 0.96, while voltage level identification reached similarity and accuracy values of 0.97 and 0.95, respectively. Equipment models and manufacturers were extracted with an accuracy of 0.86 and a recall of 0.96. Controlled benchmarks showed that the AI-assisted workflow reduced the average metadata registration time from 96 seconds to 10 seconds per page. When applied to high-volume datasets, this reduction has a substantial impact on the total man-hours required to populate databases. Access to critical technical information, which previously required hours or days in modernization projects or fault investigations, is now typically achieved in seconds or minutes. The results confirm that combining AI-assisted legacy data extraction with object-oriented engineering platforms effectively bridges brownfield and greenfield contexts, enabling faster analysis, improved fault investigation.

5 AI-Powered Engineering Advisors

The evolution of grid digitalization reaches a critical inflection point with the use of LLMs as interfaces for engineering intelligence. In this study, an AI Engineering Advisor integrated into an EP was evaluated. Rather than acting as a generic language model, the Advisor functions as a domain-specific expert with direct access to the structured engineering database, enabling reasoning across equipment attributes, functional relationships, protection settings, communication parameters, and project revision history. This allows engineers to perform semantic searches and validation tasks using natural-language queries.

The prototype was assessed across a range of PAC scenarios, including queries for concrete technical artifacts and configuration data, for example: “Display the interlocking logic for circuit breaker QA1”; “Retrieve the trip wiring diagram for circuit breaker QB3”; “Which elements were modified in the last approved project revision?”; “Which equipment is configured with the IP address 192.168.200.10?”; and “What is the pickup setting of the 50 overcurrent protection function for IED F1 in substation SME?”. In all cases, the Advisor produced context-aware responses by correlating information across multiple layers of the engineering model. Beyond productivity and quality control, the AI Engineering Advisor supports intergenerational knowledge transfer by providing less experienced engineers with intuitive access to structured engineering information.

6 Conclusion

The findings of this study demonstrate that the transition from document-centric PAC engineering toward a data-centric, object-oriented paradigm produces effects that extend beyond improvements in efficiency and information consistency. By representing engineering information as structured, governed, and semantically linked data, EIM enables the systematic preservation, traceability, and reuse of technical knowledge across the full lifecycle of PAC systems.

A relevant implication of this paradigm shift concerns education and lifelong learning in utilities, system integrators, and equipment manufacturers. When engineering platforms enforce standardized data models, workflows, and IEC 61850-aligned representations, learning processes can be embedded directly into routine engineering activities. Less experienced engineers interact with guided templates, object histories, and structured queries that explicitly expose design intent and change rationale, reducing dependence on informal knowledge transfer mechanisms and heterogeneous document repositories.

In parallel, experienced professionals benefit from reduced effort in information retrieval and verification, allowing expertise to be applied more consistently at the level of design validation, coordination, and decision-making. From an organizational perspective, EIM-based platforms facilitate the formalization of engineering knowledge, enabling its dissemination and consistent application across teams, projects, and timeframes.

The combination of EIM and Artificial Intelligence provides a viable and scalable foundation for managing PAC engineering knowledge across the asset lifecycle. AI-assisted extraction techniques enable the efficient structuring of legacy documentation, significantly reducing the effort required to populate centralized engineering databases in brownfield environments. Furthermore, the evaluation of an AI Engineering Advisor shows that, when grounded in validated and semantically organized engineering data, natural-language interaction can be reliably applied to complex PAC use cases, including configuration analysis, change identification, and parameter verification.

In summary, the results indicate that EIM, supported by standardized data modeling, IEC 61850 integration, and AI-enabled tools, provides a robust methodological foundation for addressing complexity, knowledge preservation, and workforce transition challenges in PAC engineering. By aligning engineering processes, governance, and training around a shared and structured information backbone, the approach supports scalable knowledge management and contributes to long-term consistency, traceability, and resilience in highly digitalized power system environments.

BIBLIOGRAPHY

[1] CIGRE WG B5.87, “Digital Transformation of Protection, Automation and Control Systems: Expanding the Application of IEC 61850,” in Terms of Reference (TOR), 2025.

[2] CIGRE WG B5.88, “Implementation Guide for Fully Digital IEC 61850-Based Protection, Automation and Control Systems,” in Terms of Reference (TOR), 2025.

[3] CIGRE WG B5.69, “Protection, Automation and Control Architectures with Functionality Independent of Hardware,” in Technical Brochure No. 891, 2023.

[4] CIGRE WG B5.62, “Test Strategy for Protection, Automation and Control (PAC) Functions in a Fully Digital Substation,” in Technical Brochure No. 760, 2019.

[5] Z. M. Ma, “Engineering information modeling in databases: needs and constructions,” Industrial Management & Data Systems, 105(7), pp. 900-918, 2005.

[6] J. Wu, J. Chen, G. Chen, Z. Wu, Y. Zhong, B. Chen, W. Ke and J. Huang, “Development of data integration and sharing for geotechnical engineering information modeling based on IFC.,” Advances in Civil Engineering, 8884864., 2021.

[7] O. F. Yashchenko, D. A. Makatora, R. A. Kubanov, P. L. Zinych and D. E. Prusov, “Theoretical and Methodological Bases for Implementing BIM Technologies in Construction Companies: Essence. Characteristics. Economic Efficiency,” Business Inform, no. 1, pp. 167-177, 2024.

[8] J. Farooq and P. Sharma, “A BIM-based Detailed Electrical Load Estimation, Costing and Code Checking,” International Journal of Electrical & Computer Engineering 8(5), pp. 2088-8708, 2018.

[9] N. Kabbara, M. O. Nait Belaid, M. Gibescu, L. R. Camargo, J. Cantenot, T. Coste and H. Morais, “Towards software-defined protection, automation, and control in power systems: Concepts, state of the art, and future challenges.,” Energies, 15(24), p. 9362, 2022.

[10] X. Wang, N. N. Schulz and S. Neumann, “CIM extensions to electrical distribution and CIM XML for the IEEE radial test feeders,” IEEE transactions on Power Systems, 18(3),, pp. 1021-1028, 2003.

[11] S. Mani, M. A. Haddad, D. Constantini, W. Douhard, Q. Li and L. Poirier, “Automatic digitization of engineering diagrams using deep learning and graph search,” IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, pp. 176-177, 2020.

[12] S. Eltyeb and N. Salim, “Chemical named entities recognition: a review on approaches and applications,” Journal of cheminformatics, 6, pp. 1-12, 2014.

[13] R. Monaco, C. Bergaentzlé, J. A. L. Vilaplana, E. Ackom and P. S. Nielsen, “Digitalization of power distribution grids: Barrier analysis, ranking and policy recommendations,” Energy policy, vol. 188, p. 114083, 2024.

[14] R. Sulaiman, M. F. M. Amran and N. A. A. Majid, “A study on information extraction method of engineering drawing tables,” International Journal of Computer Applicationss, 50(16), 2012.