ABSTRACT We perform a simulation with Galacticus, a semi-analytical galaxy formation model, to predict the number counts of H α and [O iii] emitting galaxies. With a state-of-the-art N-body simulation, UNIT, we first calibrate Galacticus with the current observation of H α luminosity function. The resulting model coupled with a dust attenuation model, can reproduce the current observations, including the H α luminosity function from HiZELS and number density from WISP. We extrapolate the model prediction to higher redshift and the result is found to be consistent with previous investigations. We then use the same galaxy formation model to predict the number counts for [O iii] emitting galaxies. The result provides further validation of our galaxy formation model and dust model. We present number counts of H α and [O iii] emission line galaxies for three different line flux limits: 5 × 10−17erg s−1 cm−2, 1 × 10−16 erg s−1 cm−2 (6.5σ nominal depth for WFIRST GRS), and 2 × 10−16 erg s−1 cm−2 (3.5σ depth of Euclid GRS). At redshift 2 < z < 3, our model predicts that WFIRST can observe hundreds of [O iii] emission line galaxies per square degree with a line flux limit of 1 × 10−16 erg s−1 cm−2. This will provide accurate measurement of large-scale structure to probe dark energy over a huge cosmic volume to an unprecedented high redshift. Finally, we compare the flux ratio of H α/[O iii] within the redshift range of 0 < z < 3. Our results show the known trend of increasing H α/[O iii] flux ratio with H α flux at low redshift, which becomes a weaker trend at higher redshifts.