AC Optimal Power Flow (ACOPF) — C++ (Exact Jacobian + Exact Hessian) with Ipopt / Knitro

This is self-contained ACOPF solver written in C++. It loads a MATPOWER .mat case file, builds the full nonlinear ACOPF with exact first- and second-order derivatives, and writes the solution out as simple CSV files.

---

Dependencies

You’ll need:

• Ipopt (headers + libraries)
• Matio (to read MATPOWER .mat files)
• A BLAS/LAPACK stack (usually pulled in via Ipopt)
• A linear solver for Ipopt
  • Default: MUMPS (linear_solver = mumps)
  • If unavailable, switch to another installed option (e.g., MA27/MA57, Pardiso, etc.)

Optional

• Knitro (requires a separate Knitro installation and license; not distributed with this repo)

---

Build

Default solver is Ipopt:

make

make knitro

make cmake

Then run

./acopf test_cases/pglib_opf_case30000_gocc.mat

./acopf_knitro test_cases/pglib_opf_case30000_gocc.mat

Benchmarks — Ipopt vs Knitro

Case	Obj (Ipopt)	Obj (Knitro)	Δ Obj (K − I)	Solve(s) Ipopt	Solve(s) Knitro	Speedup (I/K)	Total(s) Ipopt	Total(s) Knitro	Iter Ipopt	Iter Knitro
pglib_opf_case118_ieee	9.721361e+04	9.721361e+04	-3.41e-03	0.023	0.009	2.56×	0.025	0.009	18	9
pglib_opf_case300_ieee	5.652200e+05	5.652200e+05	-9.11e-03	0.046	0.025	1.84×	0.048	0.027	21	12
pglib_opf_case1354_ieee	1.258844e+06	1.258844e+06	-3.72e-03	0.280	0.148	1.89×	0.290	0.156	34	20
pglib_opf_case2000_ieee	1.228892e+06	1.228892e+06	+7.55e-02	0.560	0.264	2.12×	0.610	0.309	30	16
pglib_opf_case2848_rte	1.286608e+06	1.286608e+06	+1.85e-02	3.785	1.360	2.78×	3.800	1.374	163	69
pglib_opf_case2868_rte	-	2.009605e+06	—	—	1.311	—	—	1.326	—	70
pglib_opf_case2869_pegase	2.462790e+06	2.462790e+06	+2.04e-02	0.818	0.298	2.74×	0.834	0.315	43	23
pglib_opf_case3022_goc	6.013838e+05	6.013838e+05	+4.71e-02	0.741	0.260	2.85×	0.760	0.275	43	21
pglib_opf_case3120sp_k	2.147969e+06	2.147969e+06	+1.07e-01	0.779	0.246	3.17×	0.796	0.260	49	20
pglib_opf_case4020_goc	8.222473e+05	8.222473e+05	+1.31e-02	1.891	0.729	2.59×	1.917	0.752	37	25
pglib_opf_case4661_sdet	2.251344e+06	2.251344e+06	+8.38e-02	1.741	0.474	3.67×	1.765	0.497	44	22
pglib_opf_case5658_epigrids	1.207314e+06	1.207314e+06	-2.08e-01	1.723	0.526	3.28×	1.754	0.557	37	17
pglib_opf_case6468_rte	—	2.069730e+06	—	—	6.575	—	—	6.607	—	196
pglib_opf_case6495_rte	—	3.067826e+06	—	—	3.929	—	—	3.962	—	95
pglib_opf_case6515_rte	—	2.825500e+06	—	—	4.694	—	—	4.726	—	126
pglib_opf_case7336_epigrids	1.882390e+06	1.882390e+06	+2.38e-02	2.167	0.787	2.75×	2.205	0.823	38	20
pglib_opf_case9241_pegase	6.243090e+06	6.242484e+06	-6.06e+02	10.831	7.823	1.38×	10.883	7.873	141	55
pglib_opf_case9591_goc	1.061684e+06	1.061684e+06	-4.27e-01	3.898	1.521	2.56×	3.944	1.564	42	22
pglib_opf_case10000_goc	1.354031e+06	1.354031e+06	+3.46e-01	11.208	6.615	1.69×	11.267	6.660	165	83
pglib_opf_case10192_epigrids	1.686923e+06	1.686923e+06	+1.46e-02	4.700	1.351	3.48×	4.748	1.403	53	20
pglib_opf_case10480_goc	2.314648e+06	2.314648e+06	+2.35e-02	5.965	1.819	3.28×	6.022	1.870	50	24
pglib_opf_case13659_pegase	8.948049e+06	8.948048e+06	-1.29e+00	7.373	5.734	1.29×	7.443	5.799	71	48
pglib_opf_case19402_goc	1.977354e+06	1.977354e+06	+2.40e-02	10.464	4.095	2.56×	10.558	4.188	48	27
pglib_opf_case20758_epigrids	2.617867e+06	2.617867e+06	+7.47e-03	11.886	2.606	4.56×	11.981	2.699	72	20
pglib_opf_case24464_goc	2.629531e+06	2.629531e+06	-1.29e-01	12.228	5.633	2.17×	12.336	5.743	63	34
pglib_opf_case30000_goc	1.142332e+06	1.142332e+06	-3.89e-01	21.827	30.646	0.71×	21.940	30.757	131	192
pglib_opf_case30000_goc__api	1.776875e+06	1.776875e+06	+1.04e-01	91.831	29.108	3.15×	91.939	29.215	405	176
pglib_opf_case30000_goc__sad	1.142332e+06	1.142332e+06	-7.16e-02	22.875	30.652	0.75×	22.987	30.767	131	—

Notes

• Speedup = Solve time (Ipopt) / Solve time (Knitro)
• Speedup > 1 ⇒ Knitro is faster

Experimental Setup

• Platform: Apple MacBook (Apple M4 CPU, 16 GB RAM, macOS)