We propose a methodology and power models for an accurate high-level power estimation of physically partitioned and power-gated SRAM arrays. The models offer accurate estimation of both dynamic and leakage power, including the power dissipation due to emerging leakage mechanisms such as gate oxide tunneling, for partitioned arrays that deploy data-retaining sleep techniques for leakage reduction. Using the proposed methodology, dynamic, leakage and total power of partitioned SRAM arrays can be estimated with a 97% accuracy in comparison to the power obtained by running circuit-level simulations.