mrvictoru/energydecision-dt-v2-sdp

AEMO SDP-Teacher Decision Transformer (planner-distilled, physics-v2)

The shipped / preferred model of the energydecision benchmark. A standalone Decision Transformer distilled from an honest (non-clairvoyant), degradation-aware SDP-planning teacher — no solver at inference — that beats the PPO reference on all four identity surfaces and remains strong under the market-impact benchmark.

2026-09 update (physics-v2): this model was retrained on the corrected degradation/efficiency physics (round-trip efficiency 0.80, calendar aging, corrected rainflow C-rate units) and a wear-calibrated teacher (--deg-calibration 0.12, aligning the planner's per-step wear with the environment's realized per-cycle wear). See Provenance below. It supersedes the earlier aemo_dt_sdp_jtsoc_fullcorpus.pt.

Files in this repo

File	Model	Status
aemo_dt_sdp_jtsoc_v2cal.pt (+ .pt.meta.json)	Physics-v2 Stage C — J_t(soc) RTG, wear-calibrated SDP teacher (320 eps)	✅ Shipped — use this one
aemo_dt_sdp_jtsoc_fullcorpus.pt (+ .pt.meta.json)	Stage C v1 — J_t(soc) RTG, combined corpus (640 eps), pre-physics-fix	Historical (superseded; kept for the v1 comparison)

(Removed 2026-09: the behaviour-cloning-era files aemo_dt_fcas_model.pt and aemo_dt_fcas_best_checkpoint.pt.)

Results (physics-v2 shipped model, aemo_dt_sdp_jtsoc_v2cal.pt, rtg_mode="auto")

Surface	This model	Prev. shipped (v1)	PPO	Ratio vs PPO
Standard Oct	$16,209/ep	$11,573	$2,353	6.9×
Dispatch-matched	$40,039/ep	$35,320	$22,530	1.78×
2025 OOD	$30,791/ep	$25,862	$6,498	4.74×
Expanded broad-2024	$32,146/ep	$34,761	$19,504	1.65×
Market impact (piecewise merit-order)	DT well above PPO (e.g. Hornsdale-class SA1 Oct ≈ $362k vs PPO ≈ $75k)	—	—	passes

All four identity surfaces beat PPO; the model improves on the previous shipped checkpoint on three of four surfaces. Methodology, the physics fixes, and the full experiment ladder: report.md §8.2.10 and docs/known_issues.md.

Provenance: what changed in physics-v2

Fix	Effect
A3/A4 — rainflow inferred current was %SoC/hour instead of a C-rate (~100× too large, then clamped); reset() dropped the C-rate cap	corrected cycle-aging magnitude globally
A5 — planner degradation estimator always returned 0 (a 3-point rainflow sequence can never close a cycle)	planners are now wear-aware and state-dependent
A6 / B4 — the planner's per-step wear over-counted the env's per-closed-cycle wear by 8.4× on identical actions	added an explicit deg_calibration factor; the teacher corpus was regenerated with --deg-calibration 0.12
B1/B8 — impact-aware J_t(soc) dispatch-sign bug and a return_scale sidecar bug in the impact-eval runner	evaluation correctness (the previous "explicit j_t_soc collapses under impact" claim was an artifact)

The model card's numbers above are from the corrected pipeline. Prior household and AEMO numbers produced before these fixes are historical.

Architecture

Parameter	Value
Blocks / hidden dim	8 / 768
Attention	GQA (12 Q / 6 KV), QK-Norm
RoPE	enabled (rope_max_position=630, max_timestep=2016)
FFN	SwiGLU, dropout 0.15
Weight tying	Enabled
State / action dim	18 / 9 (full_fcas)
Action head	Mixed (Tanh energy, Sigmoid FCAS)
return_scale	23,284.83 (auto-calibrated)

Action space: dim 0 energy dispatch in [-1, 1]; dims 1–8 FCAS bids in [0, 1] (RAISEREG, LOWERREG, RAISE6SEC, LOWER6SEC, RAISE60SEC, LOWER60SEC, RAISE5MIN, LOWER5MIN). State space: 18-dim normalized market observations (time features, RRP, demand, 8 FCAS prices, generation mix, SOC).

Training data

• Shipped: dt_trajectories_jtsoc_v2cal_conservative.parquet — 320 episodes, 3.13M rows, honest SDP teacher with wear calibration and per-episode J_t(soc) cost-to-go RTG. Hosted at mrvictoru/AEMO_simulated_trade_sdp.
• Generated by scripts/generate_sdp_dt_trajectories.py (--deg-calibration 0.12, --rtg-mode j_t_soc); method in report.md §4.3/§5.4.

Usage

import json
import torch
from huggingface_hub import hf_hub_download
from decision_transformer import DecisionTransformer
from decision import AEMOAgent

repo = "mrvictoru/energydecision-dt-v2-sdp"
ckpt = hf_hub_download(repo, "aemo_dt_sdp_jtsoc_v2cal.pt")
meta = json.loads(hf_hub_download(repo, "aemo_dt_sdp_jtsoc_v2cal.pt.meta.json"))

model_kwargs = dict(
    state_dim=meta["model"]["state_dim"], act_dim=meta["model"]["act_dim"],
    h_dim=meta["model"]["h_dim"], context_len=meta["model"]["context_len"],
    n_block=8, n_heads=12, n_kv_heads=6, drop_p=0.15,
    max_timestep=2016, rope_enabled=True, qk_norm=True, tie_weights=True,
    action_head_mode="mixed",
)
model = DecisionTransformer(**model_kwargs)
model.load_from_checkpoint(ckpt)             # raw state_dict
model.return_scale = meta["return_scale"]    # 23284.83
model.eval()

agent = AEMOAgent(env, algorithm="dt", model=model, rtg_mode="auto")

rtg_mode="auto" resolves to state-dependent j_t_soc prompting on price-taking (identity) surfaces and a conservative constant-RTG fallback under a market-impact model.

Inference prerequisites (what else you need)

1. Source code. The DecisionTransformer model, AEMOAgent, and the AEMOBatteryTradingEnv live in the energydecision repo, not in this model repo.
2. Seasonal RRP profiles (bundled here). For rtg_mode="j_t_soc" / "auto" on identity surfaces, the agent builds a J_t(soc) cost-to-go table from a seasonal price forecast, cached to data/aemo_sdp/seasonal_rrp_<REGION>.json. The five profiles are included in this repo under data/aemo_sdp/ — copy them to data/aemo_sdp/ in the repo checkout (otherwise the j_t_soc path has to rebuild them from raw AEMO parquets, which the simulation environment fetches separately).
3. Alternative without profiles. rtg_mode="constant" (with a prompt value) needs no seasonal profile, but it is not the shipped setting and under-performs on identity surfaces.

Intended use

Research into offline RL / sequence modeling for energy markets; simulation of multi-market BESS dispatch in Australia's NEM; baseline for planner-distilled offline RL. Not intended for live trading without further validation (sim-to-real is open), risk management, and regulatory compliance.

Citation

@misc{energydecision-dt-v2-sdp,
  title = {AEMO SDP-Teacher Decision Transformer (planner-distilled, physics-v2)},
  author = {mrvictoru},
  year = {2026},
  url = {https://huggingface.co/mrvictoru/energydecision-dt-v2-sdp}
}

Related

• Dataset: mrvictoru/AEMO_simulated_trade_sdp
• Modern v2 (behaviour-cloning era, superseded): energydecision-dt-v2
• Impact-aware variant: energydecision-dt-v2-impact
• Source repo: mrvictoru/energydecision