Why Grid Operators' AI Investments Keep Hitting the Same Data Wall

In February 2026, Schneider Electric and ETAP announced a physics-based digital twin for utilities and critical infrastructure — engineering-grade simulation running against real-time grid data. The announcement was well-received. Physics-based modeling at scale, with operational data flowing in live, represents a genuine leap from static single-line diagrams.

But there is a structural question the announcement raised without answering: whose asset data, exactly, is feeding that twin?

Two asset worlds, one grid

For a grid operator managing transmission infrastructure, the asset reality is split. The transmission network — lines, towers, cables, substations, switching yards — is fundamentally a spatial graph. Its topology determines how load flows, how faults propagate, and how a switching operation ripples through the entire connected system. Managing it well requires GIS-native data models, network graph reasoning, and cascade failure simulation. That is the domain of linear asset management: assets whose risk is defined by their position in a network, not just their own condition.

The generation assets and major substations on that same grid are fundamentally different. A combined-cycle plant, a battery storage facility, or a large substation is a hierarchy of discrete assets — turbines, transformers, switchgear, control systems, cooling circuits — where failure propagates down the asset tree rather than across a network. Tracking their reliability means OEE-grade telemetry, ISO 14224 taxonomies, and parent-child failure mode analysis. That is the domain of discrete asset management: assets whose risk is defined by their position in a maintenance hierarchy.

Most grid operators manage these two worlds in two separate systems. The GIS platform holds the network. The CMMS or legacy EAM holds the substations and generation assets. The digital twin sees whichever data it was integrated with — which usually means control system data, not the full maintenance and inspection history that the EAM platform holds.

What the physics-based twin actually requires

A physics-based digital twin simulates what happens in the network under different operating conditions. That simulation is only as accurate as the asset data backing it. And asset data is not just nameplate ratings and circuit diagrams — it is the actual condition of each asset, its maintenance history, its known degradation profile, and the outstanding work orders that have not yet been executed.

A transformer operating at 75% of rated condition behaves differently from one at 95%. A circuit breaker with an overdue maintenance cycle carries a different failure probability than one inspected last month. These are not refinements to the physics model — they change what the simulation predicts about real risk.

When the maintenance data for a critical substation transformer lives in a CMMS that the digital twin has no integration with, the twin is simulating the network as it was designed, not as it is. That is useful for planning new configurations. It is considerably less useful for understanding today's operational risk.

The broader EAM market confirmed this gap in 2025. Machine learning models can now forecast equipment failures six to twelve months in advance with accuracy rates exceeding 85% — but only when they have access to both the network context and the maintenance state of the assets within it. A model that sees transmission line sensor data but not substation maintenance records is making predictions with half the relevant picture.

The energy transition makes this worse

The grid's asset composition is changing faster than at any point in the past century. Utility-scale solar arrays, battery storage systems, and wind farms are being interconnected to existing transmission infrastructure at high speed. Each new interconnection creates a new dependency: the transmission path the renewable asset relies on, and the substations that step its output up for transmission.

From an asset management perspective, a utility-scale battery storage system is a discrete asset: it has a hierarchy of modules, inverters, thermal management systems, and control electronics, each with its own maintenance and degradation profile. But its interconnection point — the substation transformer and breaker that connects it to the transmission network — is a node in a linear system. The risk associated with that battery asset cannot be fully understood without modeling both.

With GIS and EAM tools siloed, operators are doing this interconnection planning with incomplete asset intelligence. They can model the electrical consequences of a new interconnection. They cannot easily model the combined maintenance risk of the generation asset and its interconnection infrastructure together. That gap is exactly where unforeseen failures concentrate.

For a deeper explanation of how network-graph reasoning for linear infrastructure differs from hierarchy-based reasoning for discrete assets — and why AI models require both — see our guide to linear asset management.

What a unified platform actually enables

An EAM platform that holds both asset topologies natively — network graph for linear infrastructure, asset hierarchy for discrete — allows an AI agent to reason across the full picture without a data reconciliation step.

When an anomaly surfaces in a transmission segment, the system can immediately check the maintenance state of the substations at either end, model the network exposure if the segment degrades further, cross-reference any outstanding inspection deferrals on adjacent assets, and propose a prioritized intervention — all in one workflow. Not a report generated for a weekly review meeting. A structured action, proposed by an AI agent working continuously across both asset topologies, with a complete decision log attached.

Every recommendation includes the data inputs, the model version, and the decision path that produced it. When a FERC reliability auditor asks why a specific maintenance deferral was allowed on a low-impact substation, the answer is in the system — timestamped, traceable, and not dependent on anyone's memory.

FERC's updated NERC CIP framework, expanded in April 2026, extended mandatory cybersecurity controls to thousands of previously lightly-governed facilities. Producing a defensible, unified OT asset register from two siloed platforms on a compliance deadline is a different kind of problem than running a physics-based simulation — but it requires the same underlying thing: one coherent asset model that knows what every physical asset in the grid actually is, what condition it is in, and what work is planned for it.

The digital twin is a window into the grid's electrical behavior. The EAM is the record of the physical assets that produce that behavior. Closing the gap between them is where grid operators' AI investments go from impressive demonstrations to operational results.

Curious how UniEAM handles both? Request a demo.