⚡ electrical engineering

GeoDistNet: An Open-Source Tool for Synthetic Distribution Network Generation

This paper introduces GeoDistNet, an open-source tool that generates electrically usable and geographically representative synthetic distribution network models from public GIS data to address the scarcity of detailed utility feeder information for power-flow studies.

Original authors: Yunqi Wang, Xinghuo Yu, Mahdi Jalili

Published 2026-04-01

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Yunqi Wang, Xinghuo Yu, Mahdi Jalili

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are an urban planner trying to design a new neighborhood. You have a map of the streets, but you need to figure out how to wire the electricity to every house.

The problem? Real utility companies keep their actual wiring diagrams secret (like a secret recipe). Without these real maps, researchers often have to use tiny, made-up test models that look nothing like real cities. They are too simple to teach us how to handle electric cars, solar panels, or sudden power surges.

Enter GeoDistNet. Think of it as a "Smart Architect" that builds a realistic, working electricity model out of thin air, using only public street maps.

Here is how it works, broken down into simple steps:

1. The Raw Material: The "Street Skeleton"

Imagine you take a giant, high-definition map of a city (like Google Maps or OpenStreetMap). It shows every single street, alley, and driveway.

The Problem: This map is too messy for an electrician. It has too many tiny details, dead ends, and loops that don't make sense for a power grid.
The Fix: GeoDistNet looks at this messy map and says, "Okay, let's keep the main roads and the big intersections, but let's ignore the tiny alleys." It creates a clean "skeleton" of the area.

2. The Brain: The "Mixed-Integer Puzzle"

Now, the tool has to decide how to connect the power. It can't just draw lines randomly; electricity grids have strict rules:

Rule 1: It must be a tree, not a loop (like a family tree, not a circle). If a wire breaks, power shouldn't circle back around; it should just stop.
Rule 2: It must start from a power source and reach every house.
Rule 3: It shouldn't be too long or too expensive.

GeoDistNet solves a giant math puzzle (called Mixed-Integer Programming) to find the perfect path. It's like playing a game of "connect the dots" where the computer tries a million different ways to draw lines, but only keeps the one that looks most like a real city's power grid and follows all the safety rules.

3. The Dress-Up: Adding "Electric Clothes"

Once the lines are drawn, the model is just a drawing. It needs to come to life.

The Wires: The tool guesses what kind of wires to use based on the road type (e.g., a big highway gets thick wires, a small street gets thinner ones).
The Load: It figures out where the houses are and how much electricity they might need. It doesn't need secret customer data; it just uses logic: "This area has a park and houses, so let's put a big electricity demand here."

4. The Test Drive: The "Simulation"

Finally, the tool hands this new model to a simulator (called pandapower). It's like putting a new car on a test track.

They turn the power on.
They add more and more "traffic" (electricity demand) to see if the voltage drops too low or if the wires get too hot.
The Result: In the paper, they tested this in a real suburb in Melbourne. The model worked perfectly! It handled light traffic, normal traffic, and even heavy "rush hour" traffic without breaking.

Why Does This Matter?

Think of it like this:

Before: Researchers were trying to design a Ferrari using a toy car blueprint. It was okay for basic ideas, but it couldn't handle real speed.
Now: GeoDistNet builds a full-scale, realistic blueprint using public maps. Researchers can now test how to add thousands of electric vehicles or solar panels to a grid before they actually build them.

In a nutshell: GeoDistNet is a free, open-source tool that turns public street maps into realistic, testable electricity grids. It bridges the gap between "what the city looks like" and "how the electricity flows," helping engineers build a smarter, safer energy future without needing secret company data.

1. Problem Statement

Distribution system research increasingly requires feeder models that are both electrically analyzable and structurally representative of real-world urban and suburban networks. However, researchers face two major hurdles:

Data Scarcity: Detailed utility feeder data is rarely accessible due to confidentiality, security, and commercial constraints.
Benchmark Limitations: Existing benchmark systems (e.g., IEEE 33-bus) are often too small, lack regional diversity, and fail to capture the geographic organization, corridor layouts, and spatial heterogeneity of actual service areas.

While synthetic grid generation has advanced, many existing methods rely on proprietary data, statistical training sets, or graph-learning models that are not open-source. Furthermore, a significant gap exists between raw geographic data (e.g., road maps) and simulation-ready electrical models; raw maps lack radiality, source definitions, and electrical parameters.

2. Methodology

The paper proposes GeoDistNet, an open-source workflow that generates synthetic distribution feeders from publicly available geographic information (specifically OpenStreetMap data). The process is divided into four main stages:

A. Candidate Graph Extraction

Input: Raw road layout data from OpenStreetMap.
Preprocessing: The raw road graph is cleaned to remove duplicates and geometric artifacts.
Simplification: A candidate graph ( $G_{cand}$ ) is derived by filtering out unnecessary geometric details while preserving major intersections, branching points, and corridor continuity.
Edge Scoring: A composite weight is assigned to each edge based on segment length, road class, and geometric complexity to quantify suitability for feeder construction.

B. Mixed-Integer Feeder Synthesis (The Core)

The synthesis problem is formulated as a Mixed-Integer Programming (MIP) optimization problem to select a connected, radial subgraph from the candidate graph.

Variables: Binary variables determine whether nodes ( $y_i$ ) and edges ( $z_{ij}$ ) are retained.
Objective Function: Minimizes a weighted sum of geographic penalties, topological compactness (discouraging dense branching), and total route length.
Constraints:
- Node-Edge Coupling: An edge can only be selected if both incident nodes are retained.
- Required Nodes: Specific nodes (e.g., source-adjacent or demand anchors) must be included.
- Radiality: Enforced via a tree-cardinality condition ( $|E| = |N| - 1$ ).
- Connectivity: Enforced using a single-commodity flow formulation to ensure all retained nodes are connected to a fixed source node.
Solver: The problem is solved using Gurobi.

C. Electrical Parameter Assignment & Load Allocation

Once the topology ( $G_{feed}$ ) is synthesized:

Parameters: Line resistance and reactance are assigned using class-dependent templates based on road/line class and segment length.
Load Allocation: Demand is distributed across buses using a spatial weighting proxy. The weight for each node depends on local activity intensity and distance to demand centroids. This allows for spatially differentiated load patterns without requiring confidential customer data.
Output: The final model includes buses, lines, loads, and slack bus definitions.

D. Export and Validation

The generated network is exported directly to pandapower (an open-source Python-based power system analysis tool) for steady-state power-flow analysis. Validation checks include radiality, voltage deviation, and thermal loading.

3. Key Contributions

Open-Source Workflow: Introduction of GeoDistNet, a reproducible tool that bridges publicly accessible GIS data and simulation-ready feeder models without relying on proprietary utility data.
Optimization-Based Synthesis: Formulation of feeder synthesis as a mixed-integer optimization problem. This ensures the generated network is not just a direct copy of a street map but a geographically constrained, radial, and source-connected topology suitable for engineering analysis.
End-to-End Usability: The workflow integrates geographic data processing, structural synthesis, electrical parameterization, and direct export to standard simulation environments (pandapower), facilitating immediate validation.

4. Results (Melbourne Case Study)

The framework was tested on a residential area in Melbourne, Australia, representing 360 households aggregated into 82 load points with a peak active load of 1.44 MW.

Structural Characteristics:
- The generated feeder consisted of 95 buses and 94 branches.
- It exhibited a realistic hierarchy with a maximum depth of 22 hops and a mean depth of 11.20 hops, avoiding trivial star or chain structures.
- The total route length was 9.68 km, effectively balancing the retention of spatial structure with the removal of excessive geometric detail.
Geographic Interpretability: The topology followed major spatial corridors and branched toward demand locations, remaining geographically interpretable while abstracting the raw street graph.
Electrical Validation:
- Power-flow analysis was conducted under three scenarios: Sanity (light load), Representative (normal load), and Stressed (heavy load).
- The system responded physically correctly: voltage drops increased with load levels, and the network maintained voltage stability (above $V_{min}$ ) even under stressed conditions.
- The results demonstrated that the synthetic feeder behaves like a structured radial network with spatially differentiated demand.

5. Significance

Bridging the Data Gap: GeoDistNet provides a practical solution for researchers who cannot access real utility data, enabling studies on hosting capacity, voltage regulation, and network reinforcement using realistic, geographically grounded models.
Reproducibility: By using open-source tools (OSMnx, Gurobi, pandapower) and public data (OpenStreetMap), the workflow ensures that synthetic feeder generation is transparent and reproducible.
Scalability and Flexibility: The framework is adaptable to different geographic fidelities and can be extended to include time-varying demands and distributed energy resources (DERs) in future work.
Intermediate Abstraction: It successfully occupies a "middle ground" between raw geographic data and highly compressed benchmarks, offering a level of detail sufficient for feeder-level studies without the computational burden of modeling individual customer connections.