Risk-aware dispatch optimisation for renewable and battery assets (MILP + CVaR)

Powerlys offers a production-ready dispatch decision layer for renewable and battery assets, available to a consortium as a work package.

What it does: a mixed-integer linear programming planner solves the day-ahead optimum across price, weather and asset constraints, and a CVaR layer prices the tail explicitly instead of planning around an average. One core serves solar, wind, hybrid and standalone battery. It is software only, sitting above existing hardware and SCADA, and it requires no change to the asset.

Evidence: over 109 days of real Dutch day-ahead prices and real TenneT per-PTU imbalance settlement, the optimiser produced 14.0 percent more revenue than a rule-based baseline, roughly EUR 257,000 per year on a single 15 MW site. Both price legs are real market data; plant generation is still modelled, so this is a backtest and it becomes live validation the moment a real asset connects its telemetry. The method is published: SSRN preprint, DOI 10.2139/ssrn.7147659.

What we can contribute to a project: the dispatch and risk optimisation work package; live ENTSO-E and TenneT data pipelines running nightly; a shadow-mode evaluation environment that assesses a partner's asset on its own settlement data without touching production; and published methodology for the dissemination work package.

What we are looking for in return: one demonstration asset with real telemetry, in a liberalised market with day-ahead, intraday and imbalance and a BRP construct. Shadow mode is enough. The resulting report belongs to the asset owner, whether or not anything follows.

What we do not bring: no physical demonstration site, no laboratory infrastructure, and no capacity to coordinate a consortium. Powerlys contributes a work package; it does not lead one.