Materials & Energy
A Storage Strategy That Wins Only With Tomorrow's Prices
2026-09-10
An optimized storage schedule can look compelling because the model already knows the future. A useful computational investigation asks how much of that advantage survives realistic information timing and forecast error. Intuition may propose a scheduling pattern, but its value must be tested without access to tomorrow's answers.
The computational starting point
Start with a transparent offline storage model, documented efficiency and capacity assumptions, and timestamped historical inputs. PyPSA supports intertemporal storage constraints. Preserve the distinction between forecast issue time and the eventual observed value. Revised data can leak future information even when a spreadsheet appears to be ordered chronologically.
Sources: PyPSA storage units.
Where intuition enters
The hypothesis might be that preserving flexibility during uncertain periods matters more than aggressively optimizing the expected schedule. Compare that with a simple rule and a perfect-foresight ceiling. The intuition should specify when it abstains or reserves capacity, not just describe successful historical choices.
A test that can disagree
Run rolling decisions using only information available at each decision time. Compare the proposed policy, a declared simple baseline and the hindsight optimum under identical physical constraints. If weather-sensitive demand is modeled, archived forecast availability must be distinguished from reanalysis, which is not an original real-time forecast.
Perturb forecast bias, timing and tail errors while preserving plausible temporal structure. Report unmet constraints, cycling and losses alongside the chosen objective. A strategy that wins only when transaction costs, degradation or uncertainty are omitted has a restricted result. Do not convert hypothetical optimization output into realized revenue.
Sources: Copernicus climate reanalysis.
An illustrative decision
Suppose a sophisticated policy captures most hindsight benefit on calm days but repeatedly exhausts capacity before a forecast error arrives. A more conservative rule might perform better under the declared uncertainty scenario. That is an illustrative model comparison, not investment advice or a prediction of actual electricity-market returns.
What the research would deliver
The buyer receives a leakage audit, uncertainty stress test and a defensible comparison of offline strategies. This can prevent funding an advantage that comes from information unavailable in practice. Live dispatch, financial projections and operational integration remain separate scopes with their own validation and approval requirements.
Questions this raises
Is perfect foresight ever useful?
Yes, as a clearly labeled upper benchmark under specified assumptions, not as an achievable operating result.
What should the client provide first?
The decision timing, permitted information, physical constraints and comparison baseline, without confidential operational material in the initial enquiry.
Sources and their limits
- PyPSA storage units. Storage constraints in a model, not realized project returns.
- Copernicus climate reanalysis. Reanalysis combines observations with models; it is not independent ground truth.
Prepared with AI assistance. The linked sources support the specified technical points; they do not validate applied psionics as a whole or guarantee a result for a client.
Read the editorial and evidence standard.
Continue reading
Explore Scientific Oracle consultingfor a scoped review of an existing-data research decision. Start with a non-confidential outline of the question, available evidence and the decision it needs to inform.