Generative AI and its Impact on PACMM Knowledge Management

Intelligence (GenAI) technologies offers a paradigm shift in how knowledge is created, stored, shared, and applied across the PACMM domains.

This paper analyzes the potential of various Generative AI models and architectures to support and enhance knowledge management within the electric power utility sector, focusing specifically on PACMM-related activities. The paper classifies GenAI technologies into three broad categories—Large Language Models (LLMs), Generative Pretrained Transformers for

Code (CodeGen models), and Multimodal Generative AI—and explores their individual capabilities, limitations, and relevance to the domain.

LLMs, such as GPT-based models, are evaluated for their ability to process and generate natural language content, enabling intelligent document retrieval, technical report generation, compliance verification, and natural language interface development for engineering tools.

The paper presents use cases in which LLMs automatically extract information from IEC standards, summarize disturbance reports, and generate protection scheme explanations for both experts and trainees. The role of fine-tuning and retrieval-augmented generation (RAG) is discussed in the context of maintaining domain specificity and regulatory alignment.

CodeGen models, trained on technical codebases including engineering configuration languages such as Substation Configuration Language (SCL), are assessed for their potential in automating repetitive and complex engineering tasks. These models can support the creation, validation, and optimization of IEC 61850 SCL files, generate configuration scripts for IEDs, and assist in simulation model creation. The paper highlights the promise of integrating these models with utility engineering platforms to reduce the burden on skilled engineers and enhance configuration quality and consistency. Multimodal GenAI models, capable of interpreting and generating a combination of text, diagrams, and data visualizations, are analyzed for their ability to support situational awareness, fault analysis, and training. These models can generate annotated diagrams of protection schemes, create instructional material for substation operation, or synthesize event timelines by combining SCADA logs, protection relay records, and equipment diagnostics.

The paper discusses the integration of such capabilities into centralized knowledge repositories and digital twin environments, transforming them into active knowledge systems capable of interaction and adaptation.

A central theme of the paper is the shift from passive knowledge storage to active knowledge systems—platforms that can reason, contextualize, and generate domain-specific insights on demand. The role of GenAI in transforming documentation libraries, procedure manuals, and historical event records into conversational and generative knowledge assets is emphasized.

The paper also evaluates the implications of GenAI technologies for knowledge transfer, and the need for continuous learning in a rapidly evolving technological landscape.

Challenges associated with the adoption of GenAI technologies in PACMM are critically analyzed, including data quality, explainability, hallucination risk, cybersecurity concerns, and the need for regulatory oversight. The paper advocates for a hybrid approach that combines human oversight with GenAI capabilities to ensure reliability, accountability, and safety in knowledge-driven operations.