What a Grid Model Is and What It Is For

The problem a grid model solves

A naive view of an electricity system is a single balance equation: total generation equals total demand. That view is sufficient for a rough energy-balance study and is wrong for almost everything else.

The reason is that power does not flow where it is dispatched to flow. Once generators inject power at their terminals and loads withdraw it at theirs, the distribution of flow across the network is determined by physics — by the impedances of every line in the network — not by contract. This is a consequence of Kirchhoff’s laws, and it is why a bilateral trade between two parties on opposite sides of a country loads lines that neither party has any relationship with. These are loop flows or unscheduled flows, and they are a first-order effect in a meshed system such as Continental Europe.¹

A grid model is the artefact that lets you compute this. It is a structured representation of the network — its topology, its equipment parameters, and its injections — sufficient to solve for the voltages at every node and the flows on every branch.

What the model is used for

The fidelity required varies enormously by use case. Getting this match right is the single most consequential modelling decision, so it is worth being explicit about the landscape before going deeper.

Operational security assessment

The system operator must verify, continuously, that the network can survive the loss of any single significant element — the N-1 criterion. This requires solving a power flow for the base case and for hundreds to thousands of contingency cases, on a horizon from real-time out to day-ahead.² In the EU this obligation is codified: the System Operation Guideline requires TSOs to keep the system within operational security limits in the N-1 situation.³

Congestion management and redispatch

When the secure dispatch differs from the market outcome, the operator intervenes — curtailing here, ramping there. Quantifying the cost of that intervention, and forecasting it, requires a model that reproduces which constraints bind and how sensitive each is to each generator. This is the domain of PTDFs (power transfer distribution factors), covered in module 4.

Market design and price formation

Locational marginal pricing takes the network constraints directly into the market clearing: the price at a node is the marginal cost of supplying one more unit of demand there, which decomposes into an energy, a congestion, and a loss component.⁴ Zonal markets such as most of Europe’s aggregate the network into bidding zones and re-introduce a network representation through flow-based market coupling. Both designs require a grid model — they differ in how coarse it is and who runs it.

Network expansion planning

Where should the next line, transformer, or interconnector go, and is it worth building? Answering this means co-optimising generation and transmission investment over decades of operating conditions under scenario uncertainty. ENTSO-E’s Ten-Year Network Development Plan is the European instance of this exercise.⁵

Connection studies and hosting capacity

Can this 300 MW solar farm, data centre, or electrolyser connect at this substation without violating thermal or voltage limits? Distribution-level equivalents — how much rooftop PV a feeder can host before voltages rise out of band — use the same machinery at a different scale and with different assumptions.

Dynamic and stability studies

Everything above is a steady-state question: the system is assumed to be sitting at a stable operating point. A separate class of study asks what happens in the seconds after a disturbance — whether generators stay in synchronism, whether frequency falls far enough to trigger load shedding, whether voltages collapse. This needs a model with time-domain dynamics of machines and controls, and it has become considerably harder as synchronous generation is displaced by inverter-based resources.⁶

Resilience, restoration, and planning against extreme events

Black-start planning, cascading-failure analysis, and climate-driven risk studies use grid models to explore failure paths that are, by construction, outside normal operating experience. The Continental European system separation of 8 January 2021 and the Iberian event of 28 April 2025 are reminders that these paths are not hypothetical.⁷

What a grid model contains

Concretely, a steady-state grid model is a dataset with the following layers.

Topology

The connectivity graph. Two levels of detail are in common use:

• Node-breaker: every busbar section, circuit breaker, and disconnector is represented individually. This is what an operator’s real-time model looks like, because switching actions change the network, and it is what the CIM/CGMES exchange standard carries.
• Bus-branch: switching devices are collapsed, leaving electrical buses connected by branches. This is what most analysis tools consume, and converting from node-breaker to bus-branch (“topology processing”) is a routine and error-prone step.

Branches

Lines, cables, and transformers, each with electrical parameters — series resistance and reactance, shunt susceptance, transformer tap position and phase shift — and ratings. Ratings are usually a set, not a number: continuous, and one or more short-term emergency ratings, sometimes dynamically dependent on ambient conditions.

Nodes and injections

Buses with a nominal voltage, and attached to them: generators (with active power output, reactive capability limits, and a voltage setpoint), loads, shunt compensation, storage, and HVDC converter terminals.

Operating limits

Thermal ratings on branches, voltage bands at buses, and stability-derived limits on corridors — the last of which are not properties of any single piece of equipment but of the system’s behaviour, and are typically supplied as pre-computed constraints.

Scenario data

The model above is a network; it becomes a case only when combined with a specific set of injections: a demand profile, a generation dispatch, cross-border exchanges, and a topology state (which switches are open). Most of the effort in real grid modelling goes into producing credible scenario data, not into solving the equations.

The abstraction ladder

It helps to hold the following hierarchy in mind, because almost every practical question is really “how far down this ladder do I need to go?”

Each step down costs data, computation, and — most expensively — validation effort. A model one level too coarse gives confidently wrong answers; a model one level too fine will not be populated with trustworthy data and will fail to converge.

Where the data comes from

Grid data has an awkward status: it is operationally sensitive, commercially valuable, and increasingly required to be public. In Europe, the ENTSO-E Transparency Platform publishes generation, load, cross-border flows, and outage data under Regulation 543/2013.⁸ Network topology and parameters are far less open; researchers commonly work from reconstructed datasets such as SciGRID or PyPSA-Eur, which build a network from OpenStreetMap and public sources.⁹ Module 3 covers this in detail.

Summary

A grid model exists because flows are set by physics, not by contracts. It is used across a spectrum from real-time security to multi-decade planning, and the appropriate fidelity is dictated by the question — not by the data you happen to have. Its contents are topology, branch and node parameters, operating limits, and scenario injections, and the last of these is usually the hardest to get right.

The next module works through the physical system these datasets are describing.

References