AC load flow solver
Velra is a next-generation AC load flow solver that plugs into the tools grid operators already use, built for probabilistic risk assessment, real-time operation and planning at scale.
Proprietary solver. Designed to integrate with open-source tools.
Load flow solved 11.8 ms
Base case Illustrative network and timings, not a real grid.
*IEEE 14-bus, IEEE 118-bus and Polish 2383-bus test networks. Validation on operator data is the goal of our pilot programme.
The problem
Grids are shifting towards variable renewable generation. The operating point is no longer a handful of predictable cases, so operators are moving from a few deterministic studies towards probabilistic risk assessment.
That shift turns dozens of load flow runs into thousands. The same solver has to do far more work in the same time window.
The Newton-Raphson method sits at the core of the major tools, including PSS/E, PowerFactory and pandapower. It dates from the 1960s, it is sequential by nature, and it can fail to converge near stressed operating points.
When it fails, operators are left with a question the solver cannot answer: does this case have no solution, or did the solver simply not find one?
Dozens to thousands
load flow runs per assessment
1960s
origin of the method at the core of today’s tools
No clear answer
when a stressed case fails to converge
All 12,000 scenarios solved in the same window, assuming a solver twice as fast.
Technology
Velra is a novel iterative solver for the AC power flow equations. It is not a tuned version of Newton-Raphson: it takes a different route to the same answer. We describe what it does, and what it is designed to do, rather than how it works inside.
01Speed
On the IEEE 14-bus, IEEE 118-bus and Polish 2383-bus test networks, Velra reaches the same convergence tolerance in about half the time of Newton-Raphson. Speed matters most where load flow runs by the thousand.
Iteration 4 Relative time 1.00
Converged
Iteration 3 Relative time 0.50
Converged
Velra reached the tolerance in about 1/2 of the time (about 2x faster). Newton-Raphson total = 1.00
02Reliability
Newton-Raphson is most likely to fail where operators most need an answer: stressed operating points close to the edge of what the network can carry. Velra is designed to converge over a larger region, and to report early when a case has no solution, so that a failure is a result you can act on rather than a question mark.
Stress parameter
Loading level
Newton-Raphson
Did not converge: unclear whether a solution exists
Velra
Converged
Reliable answers, including a reliable no.
03Parallel by design
Newton-Raphson is a chain: each step waits for the one before it. Velra is structured so that work can be spread across many compute units at the same time. It runs on standard CPUs today. GPU acceleration is the optimisation path, with a projected 5 to 10x with full GPU optimisation.
A chain of dependent steps. Each waits for the one before.
Done
Independent work spread across many lanes, finishing together.
Done Projected: 5 to 10x with full GPU optimisation
Use cases
Pick your role. Each view shows what changes when load flow is faster and more reliable.
TSOs
From operational security to long-term planning with higher-fidelity models, from one solver.
SchematicIllustrative, not a real network.
RCCs
Large merged models, repeated again and again, on cycles that keep getting shorter.
SchematicIllustrative, not a real network.
DSOs
Connection and hosting-capacity questions multiply as distributed generation grows.
SchematicIllustrative, not a real network.
Integration and security
Velra is proprietary software with an open integration surface. It is designed to plug into open-source power system tools as a solver component, so there is no rip-and-replace.
Designed to integrate into open-source toolchains such as pandapower and PowSyBl as a solver component, through adapters.
Runs fully on-premises or behind your firewall. Your grid data never leaves your infrastructure.
Designed to work with CIM and PSS/E RAW. Status for each is shown below.
A mathematical method, not AI. The same input gives the same result, and the behaviour can be audited and explained.
NDA and cybersecurity agreements are welcome.
Each item carries its own status. Nothing is released unless it is marked Available.
import pandapower as pp
import pandapower.networks as pn
net = pn.case118()
pp.runpp(net) # default: Newton-Raphson
pp.runpp(net, algorithm="velra") # PLACEHOLDER API
import pypowsybl.network as pn
import pypowsybl.loadflow as lf
network = pn.create_ieee118()
lf.run_ac(network) # default solver
lf.run_ac(network, provider="velra") # PLACEHOLDER API
The solver argument is a placeholder, not a released interface.
Illustrative result. Newton-Raphson = 1.00.
Pilot programme
A free, scoped benchmark. We run Velra against your current solver, on your models, on your premises, and go through the results with you.
Agree the networks, cases and metrics that matter to you, and how data is handled. NDA and cybersecurity agreements are welcome.
We run Velra and your current solver on the same models, on your premises. Your data stays in your infrastructure.
We go through convergence behaviour and timing together, and decide what, if anything, comes next.
A few lines is enough. We reply by email.
Team
Founder and technical lead
Postdoctoral researcher at Lund University, Department of Automatic Control. Expert in power grids and networked dynamical systems.
Background in interaction design.
Background in computer science.
Business development
Background in interaction design.
FAQ
Velra is a different, novel iterative solver for the AC power flow equations. It is not a tuned version of Newton-Raphson. Newton-Raphson is sequential and can fail near stressed operating points. Velra is designed to converge over a larger region, to flag cases with no solution early, and to run in parallel. Currently it is about 2x faster on standard academic benchmarks.
No. It is designed to sit alongside your existing tools as a solver component, so your models, workflows and validation stay as they are.
No. Velra runs on-premises or behind your firewall. A pilot is run on your premises, on your models.
Three steps. We scope the cases with you, run Velra and your current solver on the same models on your premises, and review the results together. It is free and scoped, with no commitment. NDA and cybersecurity agreements are welcome.
Velra targets the same AC power flow equations and the same convergence tolerances as your current solver, so results can be compared case by case. On the benchmark networks we have tested, it converges over a larger region than Newton-Raphson. We make no unqualified guarantees: behaviour on real operator data is what the pilot is for.
No. The solver is proprietary. It is designed to plug into open-source tools like pandapower and PowSyBl through adapters, so you can use it inside workflows you already run.
Publication and IP strategy are currently being finalised. Ask us what we can share today.
Standard CPUs today. GPU acceleration is the optimisation path: we project 5 to 10x with full GPU optimisation, but that is a projection, not a measured result.