DERMS Case Study

I Built a DERMS Controller That Eliminates Voltage Violations

An OpenDSS quasi-static time-series simulation of a high-PV IEEE 13-bus feeder. I implemented four control strategies—baseline, heuristic Volt-VAR, coordinated optimization, and battery storage—and measured their impact on voltage compliance and curtailment waste.

Explore Results ↓ View Simulation Repo

OpenDSS QSTS 288 timesteps / 24 h CVXPY optimization Real simulation data

Violation reduction

−42%

Heuristic vs. uncontrolled baseline

Best outcome

0 min

Optimization: zero violations in 24 h

Worst baseline

1.075

p.u. max voltage · ANSI limit 1.05

Curtailment gap

808×

Heuristic 970 kWh vs. optimization 1.2 kWh

What I Built

A Python simulation pipeline for grid-edge voltage control

⚡

Simulation Engine

Python wrapper around OpenDSS driving a quasi-static time-series power flow. Each 5-minute timestep scales load and PV profiles, solves the power flow, applies control setpoints, and re-solves—288 steps per 24-hour day on the IEEE 13-bus test feeder.

🗂️

Feeder Model

IEEE 13-bus test feeder with synthetic high-PV penetration that causes reverse power flow and midday overvoltage. DER placements are CSV-driven; the same code runs the larger 123-bus production feeder by changing a single config flag.

🎛️

Control Strategies

Baseline: unity power factor, no DERMS.
Heuristic: rule-based Volt-VAR — absorb Q when V > 1.03 p.u., curtail P when V > 1.05.
Optimization: CVXPY convex dispatch minimising curtailment subject to voltage bounds.
Battery: energy shifting to reduce peak-hour curtailment.

How I Evaluated It

Problem framing, control design, and measured outcomes

Problem and Objective

High PV penetration on the IEEE 13-bus feeder caused sustained midday overvoltage beyond ANSI limits. The objective was to bring voltages back into bounds while minimizing renewable energy curtailment.

IEEE 13-bus feeder 24h QSTS simulation 5-minute timesteps

Method

Ran the same load and PV profiles across four control strategies.
Compared baseline, heuristic Volt-VAR, optimization dispatch, and battery support.
Evaluated each run using the same feeder model and time horizon.

This keeps the comparison fair and isolates the effect of each control strategy.

Result Highlights

Violation Minutes

355 -> 0

Baseline vs optimization

Max Voltage

1.0747 -> 1.0499

Back inside ANSI upper limit

Curtailment Tradeoff

970 -> 1.2 kWh

Heuristic vs optimization

Technical Setup (Short Version)

Feeder: IEEE 13-bus OpenDSS model (IEEE13Nodeckt.dss) from the IEEE PES Test Feeder Repository or the OpenDSS package.
Inputs: 24-hour load and clear-sky PV profiles from CSV (load_24h.csv, pv_clear_sky_24h.csv) plus DER placement CSV (ders_ieee13.csv).
Compute stack: Python orchestration + OpenDSS power flow (OpenDSSDirect.py), with optimization dispatch solved in cvxpy using ECOS.
Run loop: every 5 minutes, update load/PV, solve power flow, compute DER commands, apply setpoints, re-solve, then log voltages and KPIs.

python -m src.sim.run_qsts --config config/study_optimization.yaml --pv-scale 4.0
python -m src.analysis.run_phase5_analysis --baseline-config config/study_mvp.yaml --heuristic-config config/study_heuristic.yaml --optimization-config config/study_optimization.yaml --battery-config config/study_battery.yaml --output results/phase5

Key Results

24-hour simulation across four control strategies

Strategy	Violation Minutes	Max Voltage (p.u.)	Curtailment (kWh)	Avg Q Dispatch (kvar)
Baseline	355	1.0747	0	0
Heuristic	205	1.0570	970	168
Optimization	0	1.0499	1.2	4.4
Battery	—	—	—	—

Interactive Playground

Switch strategy, watch the feeder react

Control strategy:

Violation Minutes

…

min above 1.05 p.u.

Max Voltage

…

feeder peak (p.u.)

Curtailment

…

kWh lost

Avg Q Dispatch

…

kvar average

Voltage Envelope — 24 hours

Feeder-wide min / max / mean. ANSI limits: 0.95–1.05 p.u.

Reactive Power Dispatch

Q absorbed by smart inverters (kvar)

Active Power Curtailment

PV generation withheld to limit voltage rise (kW)