TECH

Improving power communication systems with knowledge graphing

New research published in the International Journal of Information and Communication Technology suggests that so-called knowledge graphs, a form of AI-based data organization, could improve the reliability and maintenance of power communication systems that help keep the lights on and modern electricity grids running smoothly.

Why context is fundamental in complex systems...Context is what transforms raw information into actionable knowledge. Each data element can have significant applications in multiple overlapping contexts. In a system of systems, managing these interconnected contexts is crucial—the integrity, reliability, and usefulness of the data depend on it.

Consider a smart city, where transportation, energy, water, and emergency services are all interconnected. Each system generates its own data, but the real value emerges when you understand the relationships between them—how a power outage affects traffic lights or how emergency services reroute their routes in response to a road closure. Graph-based frameworks are particularly well-suited for modeling these complex and dynamic relationships.

The human factor: perception and cognitive limitations...While technology provides the tools to store and connect vast amounts of information, humans remain the ultimate consumers and decision-makers. Research shows clear limits to the complexity that people can handle when visualizing graphs—especially as the number of nodes and connections increases. Many existing studies are based on small, homogeneous groups (often university students), meaning that best practices for real-world use cases and diverse users are still under development.

With the explosion of sensors and data streams, navigating a massive information space becomes increasingly challenging. Users need clear starting points ("You are here") and intuitive navigation tools, or they risk getting lost in a dense and confusing data forest. A good system design should enhance the user's situational awareness and support effective guidance and decision-making, even under pressure.

Complexity factors: beyond volume...Complexity in systems of systems isn’t just about the sheer amount of data. Other factors, such as time, geography, and the layering of different concerns, add further dimensions. Each facet may require its own abstraction or visualization technique to make the data comprehensible to users. Often, generic methods of navigation in complex systems fall short of helping users navigate domain-specific knowledge spaces.

Graph Navigation Strategies: Top-Down, Bottom-Up, and Middle-Out...Navigating complex graphs can follow several strategies, each suited to different user needs:

Top-Down: Start with an overview of the entire system, then drill down to specifics. This is ideal for monitoring or data science tasks, where understanding the big picture is crucial before focusing on details.

Bottom-Up: Begin at a specific point of interest and explore outward. Useful for investigations or troubleshooting, such as tracing the source of a system failure.

Middle-Out: Start from an abstraction or cluster within the graph, then move to more detailed or broader views. This is common when browsing large systems where users may not know their endpoint in advance, such as navigating a wiki or exploring a knowledge graph.

The researchers report that such a system works better than a conventional database in query efficiency, fault diagnosis, and operational decision-making. They explain that this technology could be used to help utility operators anticipate equipment failures earlier and manage increasingly complex power networks more effectively.

Power communication equipment functions as the information backbone of electricity grids, enabling substations, sensors and control centers to exchange data in real time. However, as grids are becoming more digitalized through smart sensors, distributed energy systems and private 5G networks, operators are generating far larger volumes of interconnected data that somehow has to be managed.

The researchers argue that conventional relational databases struggle with this level of complex data. Relational databases organize information into rigid tables linked by predefined relationships. While suitable for simpler systems, the researchers say they create information silos in large infrastructure networks, where maintenance records, fault reports, environmental conditions, and operational data are fragmented across separate systems.

The proposed AI framework instead uses a knowledge graph, which represents devices, faults, maintenance activities, and communication links as interconnected nodes. By explicitly mapping relationships between all these different pieces of information, the system can identify dependencies and hidden correlations more effectively. In order to integrate this information from different sources, the researchers used natural language processing (NLP), an AI technique that extracts meaning from human language.

NLP enables the system to analyze unstructured materials such as maintenance reports and technical documents alongside structured operational data. The resulting information is stored in the graph database designed specifically for highly connected data. This approach allows the utility operator to have in place predictive infrastructure management. Now, instead of relying mainly on manual inspections and operator experience when faults occur, they can predict failures in advance and carry out preventative maintenance.

Provided by Inderscience