Observability for Distributed Energy Resources in Power Grids

This direction is split into other directions: grid visibility gap. Its last version stays readable and citable here.

Version 1 of 1143 papers156 conceptsfrozen 2026-10-04

The modern power grid is undergoing a fundamental shift from a centralized "top-down" architecture to a decentralized "bottom-up" system. While high-voltage transmission lines are heavily instrumented and monitored, the local distribution networks—where residential rooftop solar, electric vehicles, and home batteries reside—remain largely a "black box" to utilities.

This lack of visibility, often called the observability gap, creates a significant challenge: grid operators cannot manage what they cannot see. Research in this field focuses on closing this gap by using data-driven methods to infer the hidden state of the distribution grid.

The Visibility Gap and the Masking Effect

The core problem is that traditional utility infrastructure was designed for one-way power flow. As noted in VADER: Visualization and Analytics for Distributed Energy Resources by Jiafan Yu and Ram Rajagopal, utilities often only see "net load" at the point of interconnection.

Net load is the difference between a household's total consumption and its local generation. If a homeowner turns on an air conditioner (increasing demand) at the same time the sun comes out (increasing solar generation), the net load recorded by the smart meter might remain unchanged. This "masking effect" hides the true stress on local transformers and makes it impossible to distinguish between a drop in demand and an increase in renewable production. As Benjamin F. Hobbs and colleagues argue in the Pro2R project, this invisibility complicates the procurement of reserves and threatens grid stability.

Mapping the Invisible from Space

One primary strategy for closing this gap is remote sensing. Since many small-scale solar installations are never centrally registered, researchers use computer vision to "find" them in satellite and aerial imagery.

However, simply identifying a panel is not enough for grid modeling. Gabriel Kasmi and Yves‐Marie Saint‐Drenan have developed tools like PyPVRoof to extract technical metadata—such as tilt, azimuth, and capacity—directly from images. This allows operators to estimate how much power a specific roof should be producing based on current weather.

The challenge in this area is "distribution shift." A model trained to find solar panels in California may fail in Denmark or Kuwait due to different architectural styles and lighting. Kyle Bradbury and Jordan M. Malof highlight these pitfalls in SolarMapper, emphasizing that real-world performance often lags behind lab results. To combat this, researchers like Muhao Guo are exploring the use of Multimodal LLMs to improve cross-domain generalization.

PV Disaggregation: Unmixing the Signal

While remote sensing tells us what could be happening, PV disaggregation tells us what is happening. This involves using machine learning to "unmix" the net load signal into its constituent parts: consumption and generation.

Xiaolu Chen and Hao Wang have proposed frameworks that use Transformer-based models and Federated Learning to perform this disaggregation while preserving consumer privacy. By fusing net load data with local weather factors, these models can accurately estimate real-time solar production without requiring a separate meter on the solar inverter itself.

Inferring the Grid’s Hidden Structure

The observability gap isn't just about generation; it's also about the physical wires. Many distribution utilities lack accurate, digital maps of their "secondary" networks—the final stretch of wire from the transformer to the home.

Researchers are treating the grid as a "hidden layer" that can be reconstructed from the data at its edges. Sejun Park, Deepjyoti Deka, and Michael Chertkov have demonstrated that it is possible to recover the exact topology and electrical parameters of a grid using only the voltage and power readings from smart meters.

This is mathematically possible because electrical properties like resistance act as a distance metric. By analyzing the correlations between meter readings, algorithms like PaToPa can map the physical connectivity of the network. This "Virtual-SCADA" approach allows utilities to detect outages and manage voltage fluctuations even in parts of the grid they have never manually mapped.

Towards a Transparent Grid

The ultimate goal of this research is to transform the "fog of war" in distribution networks into a clear, real-time stream of actionable data. Whether through the Carbon Monitor-Power project's global tracking or localized outage detection using micro-PMUs, closing the observability gap is the prerequisite for a stable, renewable-heavy energy future. By combining physics-based constraints with advanced machine learning, researchers are finally making the invisible parts of the grid visible.

Papers in this version

Showing 60 of 143 papers.

History & citationVersion 1 · 2026-10-04

Split into other directions · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_observability_for_distributed_energy_res_2fe49942,
  title = {Observability for Distributed Energy Resources in Power Grids},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/observability-for-distributed-energy-resources-in-power-grids/rdr_2fe4994250c347b091b7562ba3052915},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-20},
}

This exact version (version 1, frozen text):

@misc{lacuna_observability_for_distributed_energy_res_2fe49942_v1,
  title = {Observability for Distributed Energy Resources in Power Grids},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/observability-for-distributed-energy-resources-in-power-grids/rdr_2fe4994250c347b091b7562ba3052915/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-20},
}