Modeling of Battery Degradation for Integrated Design of Storage Systems

Abstract

Batteries in hybrid power plants lose capacity at a rate that depends on how they are dispatched. Design studies often omit this coupling or represent it through energy throughput, which treats shallow and deep cycles alike. This thesis quantifies how accounting for cycle-based degradation changes the operation and sizing of a wind-battery plant at the IEA Wind Task 50 site in western Denmark, trading in the DK1 day-ahead market. Dispatch is solved as a linear program for each year of a 20-year horizon, with capacity updated annually and replacement at 70% state of health. Three approaches of increasing integration are examined: a parameter sweep, a single nonlinear program, and a nested architecture linked by a degradation sub-gradient. Calendar aging accounts for 53 to 54% of first-year degradation. Cycle amplitude produces the largest change in degradation but cannot be reduced without giving up revenue, whereas lowering the center of the state-of-charge window reduces degradation at no revenue cost. Accounting for degradation lowers the optimal storage size by about 30%, to near 550 MWh and 175 MW. The nonlinear program does not improve on the linear program, and it fails to converge on 48% of active days, which indicates that a smooth-gradient solver is not suited to a rainflow objective. The sub-gradient was implemented and verified at fixed dispatch; the outer loop remains open. Results are conditioned on one price year and perfect price foresight.