MultiPeriodDistOPF

Julia research code for multi-period optimal power flow (MPOPF) on distribution feeders with batteries and solar PV. Batteries tie the periods together through their state of charge, so the horizon cannot be solved one period at a time. This repository holds three lines of work on that problem. Each has its own code, Julia environment, branches and README.

Workstream	What it is	Status	Start here
DDP	Differential dynamic programming with DDP4OPF, our fork of FilterDDP.jl, benchmarked against centralized Ipopt on identical instances	Active, main focus	ddp/README.md
tADMM	Temporal ADMM: splits the horizon into per-period OPFs coordinated by an augmented-Lagrangian consensus	Companion code for the IAS-Trans paper; parked	envs/tadmm/README.md
MSOPF	Multi-source OPF: feeders supplied by several substations	Active	MSOPF section below

All three read the same OpenDSS feeder models in rawData/.

Getting started

Requires Julia 1.10 or newer (the DDP work runs on 1.12).

git clone https://github.com/Realife-Brahmin/MultiPeriodDistOPF.git
cd MultiPeriodDistOPF

Then instantiate the environment for the workstream you need. The workstream READMEs give the exact commands:

Environment	Used for
envs/ddp2026	all DDP work: DDP4OPF, JuMP, Ipopt
envs/tadmm	tADMM, exporting feeder instances for DDP, and the MSOPF angle-sweep drivers
envs/multi_poi	the other MSOPF scripts

Ipopt is the default solver everywhere. Gurobi is optional and needs a license. For a step-by-step VS Code setup, see docs/first-time-installation-guide.md.

Fast clone. The history is long. git clone --depth 1 <url> fetches only the latest snapshot.

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

Test systems

Balanced single-phase OpenDSS feeders in rawData/. Diagrams of the main three are in assets/networks/.

Directory	Paper name	Size	Used by
ieee123C_1ph	ieee123	128 buses	DDP, tADMM
ieee2522C_1ph	med2522	2,522 buses	DDP, tADMM
large10kC_1ph	large10k	10,321 buses	DDP, tADMM
ieee123_5poi_1ph		IEEE 123-node with five substations	MSOPF
small2poi_1ph		two-substation test feeder	MSOPF

MSOPF: multi-source OPF

Code in envs/multi_poi/, which is being renamed MultiSourceOPF. Earlier multi-source multi-period OPF results are archived in the PESGM 2026 release. Current work studies how the voltage-angle difference between two substations drives power flow, and how batteries widen the range of angles in which no substation back-feeds:

• full_angle_pf.jl: AC power flow and OPF with exact bus-voltage angles, verified point by point against OpenDSS.
• root_level/two_source_angle_sweep.jl: sweeps the angle difference between two substations.
• root_level/battery_angle_sweep.jl: the same sweep with battery dispatch.

MSOPF does not have its own README yet. Until it does, the header comment of each driver documents its usage.

Repository layout

ddp/                DDP workstream: solver, drivers, results, notes
envs/
  ddp2026/          Julia environment for all DDP work
  tadmm/            tADMM code and environment
  multi_poi/        MSOPF code and environment
  ddp/              the original first-order DDP scheme (2025), kept as read-only reference
rawData/            OpenDSS feeder models
assets/networks/    feeder diagrams
results/            curated tADMM paper results
plots/              tADMM figure scripts
scripts/            helpers and run queues; scripts/probes/ holds one-off experiments
docs/               installation guide

Reference papers

Each resources/ folder (ddp/resources/, envs/ddp/resources/, envs/multi_poi/resources/) has a MANIFEST.txt listing the papers it relies on, with source links. Other authors' PDFs are not committed. Fetch the openly available ones with:

bash scripts/fetch_resources.sh

Branches and contributor notes

master holds merged work. New work goes on a dated branch named for its workstream, e.g. ddp-hessian-rewrites-sep20 or msopf-initial-testing-sep17.

CLAUDE.md and AGENTS.md record standing rules and context for AI coding agents, and are worth reading before producing DDP numbers. They include the settings every reported comparison must share.