From GridLAB-D Wiki
As of February 20, 2008, these are the known issues or incomplete items relating to the Gauss-Seidel Distribution Powerflow Solver.
Current Issues with Gauss-Seidel Solver
- Large systems
- Small discrepancies between Gauss-Seidel and FBS results. Voltages are within ±10 volts on 332 bus system.
- Convergence times are excessive and solver shouldn't take as long as is does. Consider whether we need a critical node solution implemented?
- Small impedance/large admittance systems an issue still. Low length lines (<50 ft) have very large admittance values. The adversely affects convergence rate (4 node system – 1698 iterations for length 20, 731 iterations for length 50, 267 iterations for length 150, 32 iterations for length 2500). Consider whether we want to use a threshold admittance before nodes are merged?
- Multithreading issue. Consider whether additional locks are required or if something else needed for multithreading with acceleration > 1.0.
- Triplex lines/nodes/meters – Appear to work, but further testing will be required once triplex/split-phase transformers are operational
- Meters – work fine from solver perspective. Untested as to if current accumulation (and reporting) is correct.
- Power in/out measurements for transformers – initial investigations seem to attribute this with delta-based windings. Possible line vs. phase current issue?
Unimplemented devices with Gauss-Seidel Solver
- Split phase (triplex) transformer implementation – (in progress)
- Regulators – Almost works in one form…more generic fix will be available from triplex implementation solution (when working)?
- Switches/fuses – implementation idea in progress, but untested and uncoded
- Relays – related to switches/fuses. Idea how to implement, untested and uncoded.
Other PowerFlow Issues
- Need a method to output GS Acceleration factor and/or iteration counts to XML output (extremely low importance, if at all)