Classical dynamic programming on a disclosed 1 kWh battery-state grid; no quantum execution.
In this article
LuniQ Energy: Understand Battery Scheduling Before Trusting a Savings Number
A cheap hour and an expensive hour do not automatically create a profitable battery schedule. Capacity, power, losses, demand and the amount of energy left at the end all affect the answer. LuniQ Energy is a small planning calculator that makes those assumptions explicit. Despite its research-family name, the current application uses classical dynamic programming, not QAOA or a quantum circuit.
What the model schedules
Each input row is one hour with an energy price and demand in kWh. The battery has a capacity, an initial charge, maximum AC charging and discharging power, and a one-way efficiency. The result is a sequence of charge, discharge, stored energy, grid import and hourly cost values.
The battery's stored-energy states are whole kWh. A capacity of 10 kWh therefore creates states from zero through ten, not an infinitely adjustable battery. Dynamic programming finds the least-cost path through those states for the entered window and constraints. It is an exact optimum on that discrete grid; it is not a claim of the best continuous physical dispatch.
Why the efficiency convention matters
Suppose one-way efficiency is 0.9. Adding 1 kWh to stored energy requires about 1.111 kWh from the AC side. Removing 1 kWh from storage delivers 0.9 kWh to the load. A complete charge/discharge cycle therefore has a modeled round-trip efficiency of 0.81.
The power limits are AC-side limits over one-hour slots. A battery cannot charge with more AC energy than its charging power permits in that hour, and discharge cannot exceed either the discharge limit or the entered demand. The model does not export to the grid and does not charge and discharge simultaneously.
The end of the window is part of the problem
A battery that starts partly charged can appear to save money simply by emptying itself. LuniQ Energy avoids granting that free depletion by requiring final charge to be at least initial charge. That condition makes a short comparison more meaningful, although it is still a chosen planning policy rather than a universal operating rule.
The reference cost is the sum of demand multiplied by price with no battery. Modeled savings are the difference between that reference and the scheduled import cost. They exclude purchase cost, degradation, financing, maintenance, demand charges and other bill components. A positive result is not a payback forecast or a guarantee of a lower actual bill.
When this is useful
Use the tool to learn how a tariff shape interacts with a small battery, to compare hypothetical power limits, or to check whether a large apparent saving disappears after including losses. It can also expose a modeling mistake: changing one-way efficiency from 0.9 to an assumed round-trip value changes the interpretation of both legs.
The tool accepts manually supplied hourly prices, including bounded negative values. It does not retrieve your utility tariff, inspect a meter, forecast demand or dispatch a real inverter. The timeline is a sequence of equal one-hour slots, not a timezone-aware billing calendar. Daylight-saving changes, seasonal tariff rules and site-specific contracts need separate treatment.
Where the evidence is
Read the output row by row. Stored energy must remain between zero and capacity. Grid import must stay nonnegative. Charge/discharge values should respect their limits, and the final stored-energy value must meet the terminal rule. Summing row costs should reproduce the total shown above the table.
Inputs live in page memory and the schedule CSV is an explicit download. Static page requests still reach hosting; the calculator does not upload a demand profile. Save the report if you need it, because this app does not offer a reusable project-file workflow like the engineering planners. Start with the synthetic sample and the Energy guide, then compare assumptions rather than treating one attractive number as a purchasing recommendation.