Heterogeneous solid electrolytes integrating two or more constituents (viz., phases, domains, and grains) may display a global ionic conductivity that is significantly superior to those of their homogeneous constituents. Many computational models and theories have been developed to understand the influence of microstructural features on the global ionic conductivity of these heterogeneous electrolytes. These model or theories typically utilize randomly packed spheres to represent second-phase fillers, however, random phase distributions are rare in practice because correlations always occur during materials processing or microstructure evolution. In this work, we develop a phase-field method based computational model that can generate 3D microstructure forming by phase separation instead of random packing, and predict both the global and local ionic conductivities of these microstructure informed by first-principles density functional theory (DFT) calculations. Using nanoporous Li₃PS₄ solid electrolytes as an example, the calculated global ionic conductivity agrees well with the experiments. The present model can therefore be utilized to generate databases of microstructure and corresponding properties for the next-step data-driven mesoscale computational design of heterogeneous solid electrolytes.