A battery has a finite number of useful cycles. Everyone building a storage project knows this, and almost every project treats it as a procurement fact: a warranty clause, a guaranteed capacity curve, a number in the financial model. Then the asset is commissioned, and the number stops being used.
The operating question is different. Every time the system decides to charge or discharge, it spends a fraction of a finite resource. The only sensible way to make that decision is to ask what the cycle costs and what this particular use of it is worth.
A cycle spent on a two-rupee spread and a cycle spent on a twenty-rupee spread cost the same. Only one of them was worth it.
Degradation as a price, not a constraint
Most control strategies handle degradation as a boundary: stay within a state-of-charge band, limit depth of discharge, cap cycles per day. Boundaries are easy to implement and easy to defend. They are also blunt. A hard cycle limit stops the asset from acting on the one interval in the month when acting would have paid for the year.
The alternative is to model degradation as a cost per unit of throughput, informed by depth, rate, temperature and state of charge, and then let the optimiser compare that cost against the value of the action being considered. The battery is then allowed to work hard when the economics justify it and to sit still when they do not — which is what an owner actually wants.
What this changes in practice
- Marginal decisions get made on economics rather than on habit or a fixed schedule.
- Reserve policy becomes explicit: capacity held back has an opportunity cost, and that cost can be compared against the risk it covers.
- Value streams stop being modes. Arbitrage, peak shaving, firming and reserve become competing bids for the same cycle, evaluated interval by interval.
- Warranty conversations get easier, because the operating record shows why each cycle was spent.
The uncomfortable part
Pricing a cycle requires a degradation model that is honest about its own uncertainty. Cell behaviour varies by chemistry, by manufacturer, by pack design and by how the asset has been used so far. A model that is precise but wrong is worse than a conservative boundary, because it will confidently spend cycles it should not.
The practical answer is to start conservative, keep the model explicit, and let measured behaviour tighten it. That is also why degradation modelling belongs in the same system as dispatch: the only way it improves is by comparing what was predicted against what the pack actually did.
Why it matters now
As storage moves from a handful of showcase projects to a normal part of a portfolio, the difference between operators will not be the cells they bought. Identical hardware, run under different dispatch policies, produces materially different lifetime economics. The policy is the asset.