The discovery of a kilonova (KN) associated with the Advanced LIGO (aLIGO)/Virgo event GW170817 opens up new avenues of multi-messenger astrophysics. Here, using realistic simulations, we provide estimates of the number of KNe that could be found in data from past, present, and future surveys without a gravitational-wave trigger. For the simulation, we construct a spectral time-series model based on the DES-GW multi-band light curve from the single known KN event, and we use an average of BNS rates from past studies of 103Gpc-3 yr-1, consistent with the one event found so far. Examining past and current data sets from transient surveys, the number of KNe we expect to find for ASAS-SN, SDSS, PS1, SNLS, DES, and SMT is between 0 and 0.3. We predict the number of detections per future survey to be 8.3 from ATLAS, 10.6 from ZTF, 5.5/69 from LSST (the Deep Drilling/Wide Fast Deep), and 16.0 from WFIRST. The maximum redshift of KNe discovered for each survey is z=0.8 for WFIRST, z=0.8 for LSST, and z=0.8 for ZTF and ATLAS. This maximum redshift for WFIRST is well beyond the sensitivity of aLIGO and some future GW missions. For the LSST survey, we also provide contamination estimates from Type Ia and core-collapse supernovae: after light curve and template-matching requirements, we estimate a background of just two events. More broadly, we stress that future transient surveys should consider how to optimize their search strategies to improve their detection efficiency and to consider similar analyses for GW follow-up programs.