A random growth model for power grids and other spatially embedded infrastructure networks

Paul Schultz, Jobst Heitzig, Jürgen KurthsThe European Physical Journal Special Topics 2014 · DOI · arXiv

Abstract

Synthetic power grid models are essential for infrastructure resilience research, yet standard network models often fail to capture the unique spatial and economic constraints of physical transmission lines. This paper presents a novel random growth model that balances cost minimization with redundancy requirements. By incorporating a "line splitting" mechanism—where new nodes are inserted into existing connections—the model successfully replicates the exponential degree distributions and stable topological properties observed in real-world national grids. Validation against Western US and European (ENTSO-E) datasets demonstrates that the model maintains a stable mean degree and realistic structural scaling across varying network sizes.

Introduction

Analyzing the vulnerability of power grids to cascading failures or external attacks requires accurate topological maps. However, detailed data on high-voltage transmission networks is frequently classified or proprietary for national security reasons. Consequently, researchers often rely on synthetic networks.

Standard network models, such as "small-world" or "scale-free" (Barabási-Albert) models, are generally unsuitable for this purpose. They often ignore the physical and economic costs associated with building lines over geographic distances. Existing spatial models frequently produce networks that are either too tree-like (and thus overly fragile) or too regular. There is a critical need for a model that is spatially embedded, sparse, and tunable to match the statistical "fingerprint" of real-world infrastructure.

The Random Growth Model

The authors propose a two-phase growth model designed to mimic the historical evolution of infrastructure:

  1. Initialization: The network begins with a small set of nodes connected via a Minimum Spanning Tree (MST). This represents the early stage of infrastructure where the primary goal is to establish connectivity at the lowest possible cost.
  2. Growth Phase: New nodes are added sequentially. The attachment rule balances two competing forces: Cost Minimization (connecting to the nearest physical neighbor) and Redundancy (connecting to a second node to ensure the grid remains connected if a single line fails).
  3. Line Splitting: A key innovation of this model is the "line splitting" mechanism. Unlike traditional models where nodes only attach to existing vertices, this model allows a new node to be inserted into the middle of an existing edge. This reflects the real-world practice of building substations along existing long-distance transmission lines.

This trade-off is managed by a heuristic function $f$ that weighs "network distance" (topological hops) against "spatial distance" (physical cable length).

Validation and Key Results

The model was validated against two primary sources: the Western US power grid (~5,000 nodes) and the ENTSO-E dataset, which covers the power grids of over 35 European member states.

Degree Distribution and the "Degree Two Bump"

While many social and digital networks follow a power-law distribution, power grids exhibit an exponential decay. Physical and economic constraints prevent the emergence of massive "hubs" seen in scale-free networks. Notably, real grids show a high frequency of nodes with exactly degree two. The authors demonstrate that their line-splitting mechanism is the essential component for reproducing this specific topological feature.

Mean Degree Stability

A defining characteristic of real-world power grids is that the mean degree ($k$) remains remarkably stable regardless of the total network size ($N$). Empirical data shows that for most national grids, $k \approx 2.8$.

Mean degree $k$ as a function of network size $N$ for various real-world power grids, showing stability around $k \approx 2.8$.

As shown in the figure above, the model successfully maintains this sparsity across several orders of magnitude of $N$, aligning with the observed data from both small and large national networks.

Scaling Laws and Robustness

The model’s Average Shortest Path Length (ASPL) scales with the number of nodes as $N^{0.24}$, matching the scaling laws found in European power grids. Furthermore, using algebraic connectivity as a metric for structural integrity, the synthetic grids were found to match the robustness of real-world national infrastructure.

Conclusion and Future Work

The proposed model provides a robust framework for generating realistic synthetic grids, though some simplifications remain. The current redundancy heuristic does not fully capture the complex "N-1" engineering safety standards used in industry. Additionally, a slight systematic deviation exists between the theoretical mathematical predictions for the decay parameter and simulation results, likely due to the added complexity of spatial constraints. Future research will focus on using these realistic topologies to simulate cascading failures and grid stability during the integration of renewable energy sources.

Summary updated

BibTeX
@article{schultz2014random,
  author = {Paul Schultz and Jobst Heitzig and Jürgen Kurths},
  title = {{A random growth model for power grids and other spatially embedded infrastructure networks}},
  journal = {The European Physical Journal Special Topics},
  year = {2014},
  doi = {10.1140/epjst/e2014-02279-6},
}
Figure