Batch Analysis

The gat batch command enables running power flow and OPF analysis across multiple scenarios in parallel, following the CANOS (Coordinated Automatic Network Operating System) framework for reliability studies.

Overview

Batch analysis is essential for:

• Reliability studies — Monte Carlo simulation across outage scenarios
• Sensitivity analysis — Varying load, generation, or topology
• Training data generation — Creating datasets for ML models

Power Flow Batch

gat batch pf

Run power flow for every scenario defined in a manifest:

gat batch pf grid.arrow \
  --scenarios scenarios.yaml \
  --method dc \
  --out batch_results/

Arguments:

• <GRID_FILE> — Base grid topology (Arrow format)
• --scenarios — Scenario manifest (YAML or JSON)
• --method — Power flow method: dc or ac
• --out — Output directory

Scenario Manifest Format:

# scenarios.yaml
scenarios:
  - id: base_case
    description: "Normal operating conditions"

  - id: high_load
    description: "Summer peak"
    load_scale: 1.2

  - id: gen_outage_1
    description: "Generator G1 offline"
    offline_generators: [G1]

  - id: line_outage_5_7
    description: "Line 5-7 outage"
    offline_branches: [branch_5_7]

Output Structure:

batch_results/
├── base_case/
│   └── flows.parquet
├── high_load/
│   └── flows.parquet
├── gen_outage_1/
│   └── flows.parquet
└── summary.parquet

OPF Batch

gat batch opf

Run optimal power flow for every scenario with reliability statistics:

gat batch opf grid.arrow \
  --scenarios scenarios.yaml \
  --method dc \
  --out batch_opf_results/

Additional Output:

• Generator dispatch per scenario
• LMPs (Locational Marginal Prices)
• Binding constraints
• Objective value (total cost)

CANOS Framework

GAT's batch analysis follows the CANOS framework for coordinated network analysis:

1. Scenario Definition — Define operating conditions, outages, load variations
2. Fan-Out — Distribute analysis across scenarios (parallelized)
3. Aggregation — Collect results and compute statistics
4. Reliability Metrics — Derive LOLE, EUE, and other indices

Reference: CANOS Framework

Example: N-1 Contingency Analysis

# Generate N-1 scenarios
cat > n1_scenarios.yaml << 'EOF'
scenarios:
{% for branch in branches %}
  - id: outage_{{ branch.id }}
    offline_branches: [{{ branch.id }}]
{% endfor %}
EOF

# Run batch power flow
gat batch pf grid.arrow \
  --scenarios n1_scenarios.yaml \
  --method dc \
  --out n1_results/

# Analyze results
gat analytics contingency n1_results/ --out contingency_report.parquet

Example: Monte Carlo Reliability

# Generate random outage scenarios
gat scenarios generate \
  --grid grid.arrow \
  --count 500 \
  --outage-rate 0.02 \
  --out mc_scenarios.yaml

# Run batch OPF
gat batch opf grid.arrow \
  --scenarios mc_scenarios.yaml \
  --method dc \
  --out mc_results/

# Compute reliability metrics
gat analytics reliability mc_results/ --out reliability.parquet

Performance

Batch analysis leverages Rayon for parallel execution:

Times on 16-core system. Actual performance varies by hardware.

Integration with ML Pipelines

Batch results feed into ML feature extraction:

# 1. Run batch analysis
gat batch pf grid.arrow --scenarios scenarios.yaml --out batch/

# 2. Extract GNN features
gat featurize gnn grid.arrow --flows batch/*/flows.parquet --out features/

# 3. Train model
python train_gnn.py --features features/

Related Commands

• Power Flow — Single power flow analysis
• OPF — Single OPF analysis
• Reliability — Reliability metrics
• ML Features — Feature extraction