MultiPeriodDistOPF

Scalable multi-period optimal power flow (MPOPF) for distribution feeders with battery storage and solar PV. The headline method is temporal ADMM (tADMM), which decomposes the multi-period problem along time into small per-period branch-flow OPFs and coordinates them through an augmented-Lagrangian consensus — solving horizons that the monolithic formulation cannot. A monolithic Branch-Flow (BF) solver is included as the baseline.

Companion code for the IAS-Trans paper "Scalable Multi-Period Optimal Power Flow in Distribution Systems via Temporal Decomposition."

Install

Requires Julia ≥ 1.10. The project environment lives in envs/tadmm/:

git clone https://github.com/Realife-Brahmin/MultiPeriodDistOPF.git
cd MultiPeriodDistOPF
julia --project=envs/tadmm -e 'using Pkg; Pkg.instantiate()'

Key dependencies: JuMP, Ipopt (default solver), OpenDSSDirect, Plots. Gurobi is optional (needs a license). See docs/first-time-installation-guide.md for a step-by-step VS Code walkthrough.

Tip for a fast clone: this repo has a long history. To grab just the latest snapshot, git clone --depth 1 https://github.com/Realife-Brahmin/MultiPeriodDistOPF.git (or --filter=blob:none for a blobless partial clone).

Quickstart

The run scripts auto-activate envs/tadmm, so after instantiating you can just:

julia run_bf.jl                       # monolithic Branch-Flow baseline
JULIA_NUM_THREADS=16 julia run_tadmm.jl   # temporal ADMM (parallel over periods)

Configure via config.jl, or override per run with environment variables:

# system + horizon
SYSTEM_OVERRIDE=ieee123C_1ph T_OVERRIDE=24 julia run_tadmm.jl
# tADMM penalty / tolerances
RHO_OVERRIDE=15000 EPS_PRI_OVERRIDE=1e-4 julia run_tadmm.jl

Outputs are written to envs/tadmm/processedData/<system>_T<T>/ (gitignored — regenerated on each run). The curated final-winner results behind the paper live in results/.

Test systems

Balanced single-phase OpenDSS feeders in rawData/:

Name	Buses	Notes
ieee123C_1ph	128	IEEE 123-node, small (solution-quality reference)
ieee2552C_1ph	2,552	synthetic medium feeder
large10kC_1ph	10,321	synthetic large feeder (primary scalability benchmark)

Repository layout

config.jl, run_bf.jl, run_tadmm.jl   core entry points (edit config, then run)
run_rho_sweep.jl                     penalty (rho0) tuning sweep
tadmm_socp.jl                        single-file interactive runner (VS Code)
envs/tadmm/                          Julia project + parser/validators/logger/plotter
rawData/                             OpenDSS feeder models
results/                            curated final-winner results (summary.csv + trajectories)
plots/                               figure-generation scripts
scripts/                             helpers; scripts/probes/ = one-off experiment scripts
docs/                                installation guide

Reproducing paper results

Winning rho0 per (system, T) is listed in results/summary.csv. To regenerate a cell's full penalty sweep:

SYSTEM_OVERRIDE=ieee2552C_1ph T_OVERRIDE=144 julia run_rho_sweep.jl

Notes

• Windows "filename too long" error? Run, in an elevated Git Bash: git config --global core.longpaths true (reference).