We have added a complete grid reduction module to VeraGrid that allows you to simplify large networks while preserving the electrical behaviour that matters at the boundary. The feature works directly from the schematic or the database view, and the API exposes the full process for scripted studies.
Grid reduction is essential when the area of interest is local and the rest of the system only needs to appear electrically consistent. A good equivalent keeps boundary voltages, currents and flows close to the original system while removing the internal details. This approach reduces model size, speeds up iterative studies and makes scenario analysis more manageable without compromising the physics.
The implementation in VeraGrid supports three established methods. Each method follows the published formulations and targets a slightly different need.
This is the classical method built around admittance matrix partitioning. The external buses are removed and their effect on the boundary is represented with equivalent admittances and complex injections. The algorithm slices the original Y matrix into internal, boundary and external blocks, computes the Schur complement Yeq and derives the equivalent injections Seq that reproduce the boundary behaviour. The outcome is a compact system that behaves consistently at the boundary under the same operating point. It is a solid choice for high level planning or contingency screening where preserving the base case behaviour is sufficient.
This method extends Ward reduction to handle generators and loads more realistically. Instead of collapsing everything, external generators are relocated to boundary buses based on electrical proximity using an auxiliary equivalent. Loads are then recalibrated through an inverse power flow so that the reduced system reproduces the original flow pattern even after generator relocation. The result keeps generator objects intact with their attributes, which is very practical for operational studies, dispatch analysis or investment workflows where generator identity matters. The method is described in Di Shi’s dissertation and has become a reference for planning oriented equivalents.
This method focuses on preserving the flow pattern rather than strictly preserving the non-linear boundary injections. Flows in the original system are computed using the PTDF and base injections. After aggregating injections to the boundary and removing the external buses, the PTDF of the reduced system is recomputed. A least squares solution then finds the injections that best reproduce the original flows. Any remaining mismatch is compensated with small correction loads. This approach aligns well with analysts who work primarily with linear sensitivities and want the reduced network to replicate the base case flows as closely as possible within a linearised framework.
Each reduction method modifies the network, so the software operates on a copy of the original system. Benchmarks on the IEEE 118 bus grid are included in the documentation and compare the flow errors for Ward, Di Shi and PTDF equivalents using both linear and non-linear power flows. These benchmarks show how each method behaves under different assumptions and how closely the reduced networks track the original flows.
The documentation covers the full theory, algorithmic details and API examples, including reproducible benchmark scripts. If you want to try the feature, you can download VeraGrid and explore the docs for a complete technical walkthrough. Links are in the comments.
%201.svg){kind=small}