grid-feedback-optimizer 0.1.6

Optimization methods for voltage and congestion management in modern power distribution grids.

⚡ Grid Feedback Optimizer

A Python package for feedback-based optimization of generator and device setpoints in electrical distribution grids.

---

🧩 Overview

Grid Feedback Optimizer reads a JSON/Excel network description and iteratively computes optimal setpoints.
Ideal for experimenting with voltage regulation and congestion management.

📖 Full API Reference: ReadTheDocs

---

🔋 RenewGen

RenewGen models controllable generators and power-consuming devices:

Type	Condition
Generator	p_max > 0 and p_min >= 0
Load	p_max < 0 and p_min <= 0
Flexible	p_min < 0 < p_max (can generate or consume)

Key Attributes:

• index, bus – identifiers
• p_max, p_min – active power limits
• s_inv – apparent power rating
• p_norm – normal active power (auto-computed)
• q_norm – normal reactive power (default 0.0)
• c1_p, c2_p – linear/quadratic active power cost coefficients
• c1_q, c2_q – linear/quadratic reactive power cost coefficients

Cost Function:

Cost = c1_p * p + c2_p * (p - p_norm)^2 + c1_q * q + c2_q * (q - q_norm)^2

Automatic p_norm:

• Generator → p_norm = p_max
• Load → p_norm = p_min
• Flexible → p_norm = 0

---

🚀 Features

• Load networks from JSON/Excel files
• Iterative feedback optimization using:
  • Gradient Projection (GP) – fewer hyperparameters, convex conic problem solver
  • Primal-Dual (PD) – lightweight, more hyperparameters
• Structured input/output data
• Modular design (models, engine, utils) for extensions

---

🏃 Quick Start

Install:

pip install grid-feedback-optimizer

Example Notebooks:

• Load sample grid models
• Run optimization with GP or PD
• Visualize voltages, line loadings, and setpoints
• Time-series simulations

Folders:

• examples/ → JSON/Excel network files
• notebooks/ → example notebooks

---

🧑‍💻 Usage Example

from grid_feedback_optimizer.models.loader import load_network
from grid_feedback_optimizer.engine.solve import solve
from grid_feedback_optimizer.engine.powerflow import PowerFlowSolver

network = load_network("../examples/simple_example_with_transformer.json")
power_flow_solver = PowerFlowSolver(network)
print(power_flow_solver.is_congested)

res_gp = solve(network, algorithm="gp")
res_gp.print_summary()
res_gp.plot_iterations()
res_gp.save("gp_result.json")

---

🐳 Docker

docker pull senzhan2025/grid-feedback-optimizer:latest
docker run --rm -v $(pwd)/output:/app/output senzhan2025/grid-feedback-optimizer:latest examples/simple_example.json --save_path output/result.json --verbose

---

🔄 Conceptual Workflow

flowchart LR
    A[Power Flow Calculation] --> B[Optimization Step]
    B --> C[Update Generator Setpoints]
    C --> D{Convergence?}
    D -->|No| A
    D -->|Yes| E[Optimal Steady-State Achieved]

1. Power flow calculation → voltages, line/transformer loadings
2. Optimization step → update generator active/reactive power using feedback
3. Iteration → until convergence

---

📚 References

• Gradient Projection – V. Haberle et al., IEEE Control Systems Letters, 2021. DOI
• Primal-Dual – E. Dall’Anese & A. Simonetto, IEEE Transactions on Smart Grid, 2018. DOI

---

⚠️ Limitations

• Currently supports balanced systems only
• Currently only includes buses, lines, transformers, and sources from power-grid-model

---

📄 License

MIT License

---

👤 Author

Sen Zhan
📧 [e-mail]