Utility Identification, Sensing and State Estimation Research | UtilityGrid
UtilityGrid Research

Identification, Sensing and State Estimation Research for Utilities

Reliable utility operations depend on knowing which asset is being observed, where it is located, and what its current condition may be. This research explores asset identity, sensing reliability and state estimation across electric and water infrastructure.

Research focus

Reliable observations begin with reliable asset context.

Utility field environments create practical challenges for identification and sensing. RFID missed reads, location conflicts, sensor drift and delayed inspection records can make it difficult to determine which asset is being observed and whether the information reflects its current condition.

The research considers electric and water infrastructure, including transformers, switchgear, reclosers, meters, pumps, valves, distribution equipment, field instruments, tools and spare parts.

The central concern is distinguishing trusted observations from uncertain observations, and determining when additional evidence is needed before an operational decision is made.

01 / Core principle

Why identity and state must be evaluated together

Asset identity, sensing and state estimation are related but distinct parts of a dependable utility information workflow.

01

Asset identity

Establishing which physical asset is associated with an observation, record, inspection or maintenance event.

02

Sensing reliability

Evaluating whether a sensor observation is current, correctly calibrated, sufficiently reliable and associated with the right asset.

03

State estimation

Combining available observations and asset context to estimate a condition while preserving uncertainty and evidence quality.

Important distinction: A state estimate is not automatically a measured or independently confirmed fact. For example, a transformer temperature trend, a meter replacement record or a pump condition trend must be interpreted in the context of the asset identity, source and observation history.
02 / Identification methods

Multimodal identification, location and state estimation

The research considers how different identification and location technologies can contribute evidence about an asset’s identity, position and movement.

Candidate technologies and information sources include:

  • RFID and sensor-enabled RFID
  • Bluetooth Low Energy (BLE)
  • Ultra-Wideband (UWB)
  • Real-Time Location Systems (RTLS)
  • Global Navigation Satellite Systems (GNSS)
  • Computer vision and equipment signals

Combining these sources requires timestamp alignment, coordinate transformation, probabilistic association, confidence assessment and uncertainty handling. The original source and age of each observation should remain available.

Utility engineer reviews transformer asset information on a tablet during a substation field inspection.
Research visual 01

Utility Substation Asset Identification and State Estimation

An engineer reviews a transformer using a rugged tablet, connecting equipment identity, field observations and condition information to the relevant asset record.

A measurement becomes more useful when its asset association and operational context are dependable.

Alt text: Utility engineer reviews transformer asset information on a tablet during a substation field inspection.
03 / Traceability

Asset movement, maintenance records and traceability

Asset movement and maintenance information should be interpreted as a sequence of observations and recorded events, rather than as automatically complete proof of physical custody.

Connected records

Work orders, inspection records, inventory, meter exchanges, maintenance activities and field observations can contribute to an asset’s operational history.

Tracked workflows

Candidate workflows include meters, tools, spare parts, portable instruments and fixed infrastructure, with attention to how observations are recorded across locations.

Evidence quality

Event time, observation time, source, identifier, quality flag and unresolved association should remain distinguishable in the traceability record.

RFID or RTLS evidence is an observation. It is not, by itself, proof that custody changed or that an asset completed a physical movement. Uncertainty and audit history should be preserved.
04 / Sensor reliability

Sensor reliability, adaptive placement and sampling

The value of sensing depends on the quality of the measurement, the operating environment and whether the observation can support the decision being considered.

Candidate measurement types

Research may consider vibration, temperature, pressure, flow, level, electrical, position and environmental measurements.

Factors affecting reliability

  • Measurement range and sampling frequency
  • Sensor calibration and placement
  • Time synchronization and data quality
  • Communication gaps and stale observations
  • Outliers, sensor drift and unit inconsistencies
  • Sensor substitution or incomplete measurement history

Distinguishing asset change from measurement issues

A change in a sensor reading does not necessarily indicate a change in the physical asset. It may reflect drift, calibration problems, timing differences, missing data or another measurement issue.

Adaptive sensing

Redundant measurement, adaptive sensor placement and adaptive sampling may be evaluated as ways to improve observation quality.

An important research question is whether additional data actually changes the decision, rather than simply increasing the volume of information collected.

05 / Field perception

Perception when field conditions degrade

Utility environments can make visual and other sensing modalities less reliable. Research should consider how systems respond when observations become incomplete or uncertain.

Environmental challenges

Conditions that can affect field observations include:

  • Low light and glare
  • Weather, dust and vegetation
  • Vibration and equipment movement
  • Occlusion and difficult viewing angles
  • Restricted access and underground vaults

Candidate sensing approaches

Candidate modalities may include radar, thermal sensing, event cameras and computer vision, depending on the observation task and environment.

No single modality should be assumed to work through every obstruction or environmental condition.

Degraded-perception response

When perception quality falls below an acceptable level, the system may need to flag the observation as degraded, request additional evidence or defer interpretation.

Operational boundaries

These research directions do not replace protective sensing, established utility safety procedures or required operational safeguards.

Research visual 02

Utility Grid Sensor Fusion and Condition Monitoring

A field engineer reviews a grid-oriented monitoring display beside substation equipment. The interpretation layer combines asset context with available sensor observations.

Evidence quality and observation currency remain important: an inferred condition should not automatically be treated as independently confirmed.

Alt text: Utility engineer reviews grid monitoring data on a tablet beside substation equipment and sensors.
Utility engineer reviews grid monitoring data on a tablet beside substation equipment and sensors.
06 / Emerging sensing

Battery-free RFID and backscatter sensing

Battery-free RFID and backscatter approaches may offer additional ways to collect information in selected utility and inventory environments.

01

Sensor-enabled tags

Research should distinguish dedicated sensor tags from attempts to infer environmental or physical conditions indirectly through radio behaviour.

02

Candidate applications

Potential areas for investigation include containers, stored materials, sealed enclosures and fixtures, with comparison against an independent reference.

03

Feasibility conditions

Reader distance, orientation, metal, surrounding objects, tag density, available energy and quantity of observations can affect practical feasibility.

Research consideration: Battery-free does not mean energy-free. Energy availability and interruption can create observation gaps. Evaluation should compare conventional sensing and battery-free approaches using measurement error, read reliability, calibration needs, missed events and installation requirements.
07 / Evaluation framework

Research questions and proposed evaluation

The proposed evaluation considers how identity, location, sensor observations and estimated asset state can be assessed under both normal and degraded operating conditions.

Research questions

01 How should conflicting sources and late-arriving data be reconciled?
02 How can an unobserved asset be distinguished from an asset that has moved?
03 How do sensor placement and sampling choices affect state estimation?
04 How should drift, timestamp errors and communication gaps be handled?
05 Which sensing modalities remain useful under degraded field conditions?
06 Under what conditions are battery-free sensing approaches feasible?

Proposed evaluation metrics

Identity association error and identity switching

Location error and observation age

Event completeness, precision and recall

Unresolved handovers and inventory accuracy

State estimation error and confidence calibration

Detection delay, false alarms and recovery

Latency, communications, energy and compute requirements

Manual reconciliation, installation and calibration effort

Evaluation discipline: Synthetic data should be separated from hidden ground truth. Assumptions, runs, uncertainty and adverse cases should be documented. Simulation results do not establish field performance. Physical validation requires representative utility equipment, authorized data, independent references and review under actual operating conditions.

Baselines and test conditions

Candidate baselines include a single sensor, fixed thresholds, last-known location and current reconciliation processes. Tests should include both normal and degraded cases so that performance is not assessed only under favourable conditions.

08 / Research scope

Research scope, status and limitations

This page describes a shared research agenda and proposed evaluation directions. It does not claim completed demonstrations or validated utility deployments.

Research context and status

The research agenda is attributed to Aperture Venture Studio, GAO RFID Inc. and GAO Tek Inc., and is associated with Aperture Physical AI Research Labs.

This attribution is not evidence of a UtilityGrid research team, facility, completed demonstration or customer deployment.

The scenarios on this page represent proposed research directions. No UtilityGrid performance figures, outcomes, configurations or field trials are asserted.

Each topic requires a defined asset class, data plan, baseline, test environment and review process.

Limitations and assumptions

  • Inconsistent asset registers and incomplete maintenance history
  • Clock differences and timestamp uncertainty
  • Sensor placement and coverage limitations
  • Environmental variation and installation costs
  • Ongoing maintenance and calibration requirements
  • Transferability between asset types is not assumed

Identity evidence is not authentication. An inferred location is not necessarily current, and an estimated state is not necessarily independently verified.

09 / Technology architecture

How this research connects to the UtilityGrid technology architecture

AIoT configuration connects identification and sensing with AI-supported decisions. Physical AI adds an authorized action path where appropriate.

This research page focuses on the quality of identity and observation inputs. An inferred asset state does not, by itself, authorize a control action.

Continue the conversation

Discuss a utility identification or sensing research question

Start with an asset class, a recurring observation gap or a field workflow. UtilityGrid can discuss the available information, operating conditions and a suitable evaluation approach.