Utilities have spent decades building reliable Geographic Information Systems (GIS) as systems of record to capture assets as they exist in the field. These GIS platforms are highly effective in supporting operational monitoring and day-to-day management of the network. However, modern grid management now requires grid-level analysis across connected assets, not just operational reporting. They now require the capability to compute, validate, and reason over the network at scale—covering its present state, how it has evolved over time, and how it may evolve in the future. This shift moves GIS beyond a static system of record, positioning it as the foundation for network intelligence within utilities.
From GIS Representation to Physical Reality
Traditionally, utilities have focused on storing and visualising assets in GIS. While effective for operations, this model is increasingly insufficient for analytical and planning use cases. Utilities need to continuously validate and reason over their networks so that digital representations stay consistent with physical reality and aligned across enterprise systems.
A key challenge becomes visible in how physical assets are represented. A single cable installed from a drum may bend, change installation method, and pass through ducts and pits along its route. Although physically continuous, it can be represented in the GIS as multiple segmented records, each capturing a local geometry and installation context. Over time, this results in a fragmented representation where one physical asset becomes multiple logical records within GIS.
For example, in the diagram below, three continuous joint-to-joint cables may be represented as nine GIS records due to changes in geometry, installation rules, or historical modelling decisions. This is because the GIS stores data at the lowest level of granularity, meaning a physically continuous cable can be split into multiple digital assets. Changes such as layout adjustments or direction shifts often trigger new GIS segments.
Figure 1: How cables are represented in GIS (segmented view)
From a network analysis perspective, however—such as in probability of failure forecasts—we are interested in the underlying physical reality: continuous cables connected end-to-end rather than segments.
In Example 2, those same nine GIS segments correspond to just three continuous physical cables.
Figure 2: How the same physical cables exist (continuous view)
From Fragmented Assets to Graph Network-Level Intelligence
As illustrated in the example above, what appears in GIS as nine separate cable segments represents just three continuous physical cables.
This creates a fundamental challenge for network analysis. In analytical use cases— such as probability of failure, load impact analysis, or asset-level risk assessment—this view becomes a limitation, because the unit of analysis is no longer aligned with how its stored within GIS.
To bridge this gap, utilities need a graph-based abstraction on top of GIS. A graph model allows these fragmented GIS features to be reconnected into meaningful end-to-end relationships, reconstructing the underlying physical cables from their segmented representations. In this context, GIS remains the system of record for spatial truth, while the graph becomes the analytical layer that reinterprets that data as a connected network.
This new analytical layer expands what the data can be used for. Instead of analysing only spatial features and assets, utilities can traverse complete paths to understand how assets are connected, how issues propagate through the network, and how different constraints affect connectivity. As a result, the combination of GIS data and graph processing shifts the problem from managing spatial records to enabling network-wide traversal and analysis. GIS continues to provide the granular truth of assets, while the graph layer enables path-based reasoning that reveals how the system behaves as a connected whole.
A Distributed Graph Intelligence Layer for Utilities
To handle this level of complexity, utilities need a distributed graph processing layer built on top of their system of record.
At one utility we used Databricks as the platform to bring the large volumes of energy,
network and asset data and process it efficiently at enterprise scale. On top of this, we utilised GraphFrames as a tool to build a graph that allows the GIS representation to be
reconnected into a single unified network structure, enabling efficient traversal across connected assets and relationships. When combined, these capabilities turn the GIS data into a distributed graph intelligence layer—one that supports full network traversal and analysis at scale.
Figure 3: A distributed graph intelligence layer
Once the network is represented as a connected graph, utilities can run path-based analysis across the system to validate and interrogate connectivity. This makes it possible to surface issues that are difficult to detect in traditional GIS views, such as failure propagations, outage customer impact analysis, incorrect phase detection, invalid switching configurations, investment planning and inconsistent asset groupings.
Figure 4: Applications of a graph intelligence layer
More importantly, it allows the network to be analysed as a whole system rather than a collection of isolated records. Utilities can traverse end-to-end paths, trace dependencies, and understand how local changes propagate through the wider network. This fundamentally shifts how the network is used. Instead of simply managing spatial representations of assets, utilities gain the ability to reason over the structure, behaviour, and integrity of the entire system.
Utility Network Intelligence with Databricks + GraphFrames
Graph Network Intelligence has been implemented for a DNSP’s cost-benefit analysis (CBA) program supporting asset investment decisions, which govern multi-million-dollar capital and maintenance planning across the electricity network. Within this context, a Databricks + GraphFrames-based graph layer delivered a step change in how GIS data is used. Rather than relying solely on GIS representations, it introduces a connected graph network model built on top of existing GIS data, enabling scalable end-to-end tracing, dependency analysis, and topology validation. For this DNSP, it is particularly critical in the CBA process, where decisions to repair or replace assets directly influence significant capital and operational expenditure. The outcome is a platform-wide graph intelligence layer that strengthens the consistency, traceability, of high-value asset investment decisions for a DNSP at scale.
Ready to transform your GIS data into a scalable graph intelligence platform?
Contact us to start the conversation.
