Abstract: Battery energy storage systems participating in both energy arbitrage and frequency regulation face a core contradiction—different services have distinct depth-of-discharge, C-rate, and thermal stress profiles, causing 3-5x differences in degradation rates. This paper constructs a “electrochemical degradation→multi-market clearing→three-layer optimization” progressive framework.Methods A single-particle model with thermal balance establishes an SEI-dominated capacity fade model explicitly coupling temperature, C-rate, and DOD. The degradation model is embedded in a three-layer stochastic optimization: upper layer optimizes storage sizing for lifecycle NPV, middle layer allocates energy and regulation capacities across day-ahead/intraday/real-time markets, and lower layer handles minute-level real-time power dispatch. An SQP-DP hybrid algorithm is designed for the nonlinear high-dimensional model.Results For a 100MW/200MWh LFP plant with two years of PJM data, optimal sizing is 18% lower than the linear depreciation model, with 23% higher lifecycle NPV. Optimal service mix is 65-70% regulation and 30-35% arbitrage. When ambient temperature rises from 25°C to 40°C, optimal DOD should decrease from 80% to 55%.Conclusions The electrochemical-thermal coupled degradation model significantly improves storage sizing economics and operational lifetime. The differentiated service mix strategy provides storage investors with a more accurate lifecycle value assessment tool.