Objective In advanced adiabatic compressed air energy storage (CAES) systems, heat exchangers recover thermal energy. Prolonged high-frequency cyclic storage and release at varying pressures can cause cumulative damage to the surrounding rocks of underground caverns, compromising their stability. Improper cavern layout and operating pressures will accelerate this damage. Therefore, evaluating the stability of surrounding rocks in advanced adiabatic CAES power plants, considering rock damage characteristics, is crucial for ensuring the long-term safe operation of energy storage caverns.

Methods The damage stress level λ was introduced to characterize the damage-driven strength. Triaxial loading-unloading tests were conducted on granite under a confining pressure of 10 MPa with λ values ranging from 0.5 to 1.0 to analyze damage evolution characteristics. A compressive strength–cycle number equation was derived through nonlinear fitting to enhance the granite damage model under cyclic loading and unloading. This equation was applied to a project converting an abandoned mine into an underground cavern for an advanced adiabatic CAES power plant. Using the orthogonal experiment method, the Mohr-Coulomb strength element safety factor (F_s) was introduced to evaluate the significance and optimal combination of cavern shape, burial depth, and cross-sectional area after 31, 000 storage-release cycles. With a minimum F_s of 2.0 set as the stability safety threshold, the operating pressure range was optimized by energy density.

Results Results indicated that when λ ＜ 0.6, damage increased slowly; between 0.6 and 1.1, damage accelerated rapidly; and above 1.1, damage growth slowed again, following a power-exponential relationship with λ. The element safety factor–cycle number curve exhibited three stages: rapid decline, slow decline, and stabilization. Specimens with higher λ experienced rapid damage, while those with lower λ showed gradual damage. The influence of cavern shape, cross-sectional area, and burial depth on surrounding rock stability decreased in that order. The optimal configuration for stability was a burial depth of 300 m and a circular cross-sectional area of 400 m. Considering project conditions, the highest energy density (29.56–42.74 MJ/m³) was achieved within a pressure range of 4–12 MPa.

Conclusion This study did not consider heat exchanger efficiency. The Mohr-Coulomb model was employed as the rock constitutive model, with material parameter reduction linked solely to the mapping between λ and compressive strength. Within an acceptable error range, this approach reduced rock damage, increased the minimum element safety factor, and optimized the cavern’s operating range. Future research should incorporate temperature effects and investigate granite’s damage mechanisms and model improvements under temperature-stress coupling using visco-elastoplastic constitutive models.