SF2 Systems Brings State Fingerprints to the Energy Sector

Wind and solar output fluctuate, power plants change load, storage systems charge and discharge, and transformers respond to changing grid conditions. At the same time, vast amounts of technical data are generated across this entire chain — from conventional operational and SCADA data to high-resolution electrical, mechanical and vibration signals.

In many cases, the data is already there. The real challenge is turning it into a meaningful picture of the system’s actual state.

A System-Level View Instead of Sensor Silos
An energy system is more than a collection of individual measurements. Current, voltage, power, temperature, pressure, flow, rotational speed and vibration all need to be understood in the context of the system’s operating condition. A single measurement may appear perfectly normal even though the interaction between multiple signals has already begun to change.

SF2 combines relevant sensor and operational data into a shared state fingerprint. Similar operating conditions produce similar fingerprints, making it possible to compare whether an asset remains within a known stable state, moves into another normal operating condition, or begins to show unusual drift.

The difference is one of perspective: rather than looking only at individual thresholds, SF2 looks at the behaviour of the system as a whole.

This approach can be applied across the energy value chain — from generation assets and storage systems to power electronics, transformers, switchgear and distribution infrastructure. It is particularly relevant in environments where weather, feed-in, load, switching states and operating phases are constantly changing and technical systems rarely experience exactly the same conditions twice.

Processing High-Resolution Data Where It Is Generated
Energy assets produce data at very different time resolutions. Conventional operating data such as power, temperature or pressure is often sampled at comparatively long intervals or already aggregated before being transferred to SCADA or historian systems.

Applications such as electrical measurement, protection systems or vibration analysis may require much higher-resolution raw signals. These can generate very large data volumes within a short period of time, making continuous transmission, central storage and processing technically and economically demanding.

SF2 can move the processing of such raw data closer to the asset itself. Signals are transformed locally into compact state information, using sparse binary representations that can be compared through efficient bitwise operations.

Instead of continuously moving large raw-data streams across networks, central control, SCADA or automation systems can receive the information that matters, for example, similarity, drift or anomaly values.

The same principle applies to unusual vibration, disturbance or interference patterns. What matters is not only that a signal changes, but how that change relates to load, temperature, power and other operating variables.

“The energy sector rarely lacks data. The real challenge is turning very different signals into an understandable representation of the overall system state,” says Francisco Webber, CEO of SF2 Systems. “That’s where we come in: deploying locally, on existing infrastructure, without forcing operators into a new data or cloud ecosystem.”

Limited Fault Data Is Not a Barrier
Critical energy assets are designed to fail as rarely as possible. As a result, the large, accurately labelled fault datasets required for many supervised machine-learning methods often do not exist.

SF2 does not require large volumes of labelled failure data or conventional supervised fault training to get started. Existing operational data can be used to model relevant system states from the outset. Maintenance records, event logs or fault data provide valuable additional context, but are not a prerequisite.

Once a model has been fixed and versioned, runtime analysis is reproducible: the same input produces the same transformation, the same fingerprint and the same comparison result.

A deviation does not automatically constitute a fault diagnosis. It indicates that the current system state differs meaningfully from previously observed states, while technical interpretation remains with the people who understand the asset and its operation.

A Real-World Case Demonstrates the Principle
One practical reference case is a real-world water-pumping system comprising 52 sensor channels and more than 220,000 measurement points.

Although only a small number of documented fault and recovery phases were available, one analysed sequence showed a relevant change in system state approximately four days before a documented pump failure.

This is not a general promise of four-day failure prediction. It does, however, demonstrate a principle that is highly relevant to the energy sector: even where many signals are involved and documented failure cases are scarce, meaningful changes can become visible early in the overall system state.

From Individual Components to System-Level Insight
SF2 extends state analysis beyond individual sensors and components. State information can be organised hierarchically — from component to subsystem to entire asset and, ultimately, larger fleets or groups of infrastructure. Local states are generated where the data originates and then consolidated at higher levels to reveal broader system-wide patterns.

For distributed energy infrastructure and large asset fleets, this creates an additional perspective for condition monitoring and state intelligence.

Starting with Real-World Energy Data
Whether the asset is a wind turbine, hydropower plant, storage system, transformer, generator or auxiliary system, the essential question is not what the technology could achieve in theory, but what the operational data of a real asset can reveal in practice.

That is the purpose of the SF2 Edge Challenge.
Operators, energy companies, equipment manufacturers and technology partners can bring historical datasets containing complex operating conditions, drift, disturbances, recovery phases or high-frequency signals. Together with SF2, they can evaluate which system states and changes can be reliably identified.

The first step can be completed using historical or anonymised data — without interfering with protection or control systems or committing to a major integration project.

Teams can also explore their own historical sensor and process data independently using SF2 Suite SE, which is permanently available free of charge for local evaluation.

The approach is deliberately pragmatic: start with the data already available, identify relevant system states, and use the results to determine whether production deployment is warranted.

Try SF2

SF2 Suite SE:
https://sf2systems.com/produkte/software/studio-se

SF2 Shop:
www.sf2systems.com/shop