Rheora ETRM turns half-hourly forecast error into a managed P&L line — before MHHS makes it a daily cash-out bill. Forecast half-hourly demand and price, hedge the shape, trade it, settle it, and prove the imbalance cost you avoided, on one governed platform built for the UK power-and-gas supply market.
From October 2026, Market-wide Half-Hourly Settlement (MHHS) settles every meter on its actual reading. Forecast error stops being socialised across load-profile classes and becomes each supplier's own daily, sometimes violent, cash-out line.
The same disciplined Amplinth loop, tuned to UK energy trading and risk — every forecast is persisted and reconciled, so the next call is sharper.
Metered volumes, Elexon system & imbalance prices (BMRS), gas, market prices (streaming), weather & REMIT/news; broker import & tenant connectors.
Six production ML models blended by a learned stacking ensemble, retrained nightly by a walk-forward harness; a 3-agent market-intelligence pipeline. No black box.
Hedge ratio, open exposure & shape risk per settlement period; price-trigger alerts; imbalance-cost forecasts; VaR/sensitivity; risk-limit-bounded suggestions.
Trade capture, OMS & paper trading execute or dry-run hedges; settlement, ECVN & exception queues drive the middle office — human-in-the-loop, permissioned.
Every forecast reconciled against actuals; P&L and imbalance attributed per period — imbalance cost avoided in EVA terms.
Forecast → hedge → trade → confirm → settle → report — UK market structure (48 half-hourly periods, EFA blocks, GSP groups, Elexon cash-out, MHHS migration, REMIT) is a first-class concept, not a configuration afterthought.
Per-segment half-hourly demand forecasting (domestic, SME, EV-heavy, heat-pump, legacy NHH) with P10/P50/P90 bands; live hedge positions, unhedged exposure, hedge ratio and shape/baseload mismatch per settlement period — the demand forecast and live position share one book.
Daily metered-volume and Elexon system-price ingestion; MHHS migration tracking by MPAN measurement class (A–G) and GSP group; ECVN notification flows; load profiling and seeding for new tenants and segments.
An XGBoost day-ahead price model, an LSTM base forecaster and quantile-regression confidence bands, fused by a learned ElasticNetCV stacking ensemble. A walk-forward backtest harness retrains nightly across ~4 years and is judged on the worst slice — season, day-of-week, period, segment — not the headline MAPE.
Daily P&L attribution and expected imbalance cash-out cost or benefit per period, from forecast-vs-actual and Elexon imbalance prices; system buy/sell (SBP/SSP) imbalance forecasting; configurable price-trigger and exposure rules that raise hedge-timing and volume alerts.
Value at Risk on open positions; price and fundamental sensitivities (gas, carbon, generation mix, cross-border) feeding both the risk view and the sensitivity-baseline forecaster; collateral and counterparty exposure; a master-agreement (GTMA/EFET-style) registry tied to counterparties and trades.
Hourly REMIT inside-information and market-abuse analysis; trade surveillance over abusive or anomalous patterns; a compliance-obligations workflow; periodic access reviews (SoD/recertification); and a centralised operational exception queue with triage.
Forward, day-ahead and intraday power trade capture and blotter; order management with venue-routing concepts; a full paper-trading simulation to dry-run strategy without market impact; and a separate gas demand-and-price stack with its own models, positions and analytics.
Live and historic market data with a streaming channel, forward-curve construction and weather feature engineering; a 3-agent market-intelligence pipeline that turns outages, news and weather into quantified price-impact scenarios; and an AI Trading Copilot for natural-language Q&A grounded in your trading and market data.
The breadth is genuinely in production: ~45 capability modules wired, ~25 UI pages and six ML models retrained nightly, multi-tenant. Every number below carries a maturity label and a source.
Maturity matters here. Capability breadth and forecast-vs-actual provenance are realized — every forecast is persisted at generation and reconciled against actuals back-filled from BMRS, gas and metered tables. Forecast-accuracy and imbalance-saving figures are modelled (walk-forward backtest over ~4 years) or target outcomes in EVA terms, to be validated against each customer's baseline. Energy-company names are market context, not customers; any reference logos are illustrative.
Connect Rheora to your meter, Elexon/BMRS, gas, market-data, weather and broker feeds. Customize the segments, hedge rules, risk limits and REMIT obligations to your portfolio — the shape is reusable; the market logic is what we tailor. Operate the closed loop in production, where every forecast is reconciled and every hedge is human-gated.
UK-native and MHHS-ready by design — 48-period shape, EFA blocks, GSP groups, ECVN, MHHS migration, Elexon cash-out and REMIT are first-class, not bolted onto a commodity-agnostic core.
Honest, retraining-backtested forecasting: six split models, a learned stacking ensemble, quantile confidence bands, every forecast persisted and reconciled. The AI copilot and intelligence pipeline stay advisory — human-in-the-loop on every hedge.
One system for demand, hedging, trading, settlement and imbalance/P&L — so exposure-per-period is live, not re-keyed. Value framed in EVA: imbalance cost avoided vs the un-hedged baseline.
Book a 30-minute demo and we'll walk your demand-to-hedge loop — the live exposure-per-period view, the forecasting discipline behind it, and the EVA case for imbalance cost avoided.