Skip to content
RT
Skip to main content
Provenance:Dataset:TEIAS v2025.1Model:PPO v3.2Run:1 Oct 2025rl_ppo_instability_test_…Seeded

PPO v3.2 — Run Detail

Seeded Stochastic

Random seed fixed; reproducible on same hardware. Seed: 99. PPO instability test — training oscillation when renewable > 85% without storage. Used in failure-modes documentation.

Interpretation Guardrail

These results are scenario-based projections, not forecasts. They reflect model assumptions about technology costs, policy choices, and demand trajectories. Actual outcomes will differ. Always consult the full methodology and assumption registry before drawing policy conclusions.

Frozen Metrics Snapshot

READ-ONLY · Not recomputed
Cost Reduction
−84.5%
Convergence Score
2100 / 100
Policy Stability
32 / 100
Final Episode Reward
2100.0
Final Training Loss
0.0412
Training Steps
1,200

These values are stored in frozen_metrics_snapshot and written at import time. They are never recalculated from raw artifacts.

Run Metadata

Run ID
rl_ppo_instability_test_01
Group ID
validation_runs
Source File
ppo_instability_test_snapshot.json
Content Hash
a11b22c33d44e55f…
Config Hash
cfghash_ppo_instability_01
Created At
01/10/2025, 09:00:00 UTC
Immutable
✓ Yes (DB enforced)
Dataset Version
TEIAS v2025.1

Artifact Manifest (1)

frozen_artifact_manifest · Integrity verified
📝
Instability training log
artifacts/ppo_instability_test/training.log
logtext/plain24 KBsha256: aa11bb22cc33

Same Group: Validation & Stress Tests

rl_a3c_highre_20251103A3CCost ↓91.2%View →
forecast_lstm_v2_validation_20250815LSTMCost ↓0%View →
rl_dqn_discrete_5lvl_20251127DQNCost ↓97.69%View →

Traceability Chain

📦 Dataset: TEIAS v2025.1→🧪 Run: rl_ppo_instability_test_01→📊 1 Artifacts→📰 Insights

📎 Cite this page

Önderol, S. (2025). Experiment Run: PPO v3.2 — rl_ppo_instability_test_01. Retrieved from https://renewabletransition.org/transition/experiments/rl_ppo_instability_test_01
PPO — rl_ppo_instability_test_01 — Experiment | Renewable Transition