On 17 August 2017, the advanced LIGO and Virgo collaboration detected the first gravitational wave event produced by the merger of a pair of inspiraling binary neutron stars (BNS). This event, named GW170817, was associated with a short gamma-ray burst and the kilonova (KN) AT2017gfo that was followed-up over the following days and weeks over the electromagnetic spectrum from ultraviolet to near infrared frequencies. The merger was localized in a nearby galaxy, NGC 4993 at a distance of 40 Mpc. The close proximity and the orientation of the merger was favourable for studying its properties and evolution with time.

This unique discovery gave rise to a new era in multimessenger astronomy and showed the relevance of these events in astrophysics, cosmology, and nuclear physics. Today, we aim at detecting samples of kilonovae to answer the many open questions related to BNS mergers and their use as physics laboratories. However, the observations of kilonovae are generally challenging as these are rare, fast and faint transients.

In this licentiate thesis, I present the observational difficulties and study the detectability of kilonovae using survey simulations. I quantify the effect of survey depth and choice of filters to optimize detection probability. I highlight the importance of accounting for the asymmetries expected for kilonovae (KNe). One conclusion of this licentiate thesis is that it is unlikely to detect a new KN under the same conditions as AT2017gfo. Even if the intrinsic properties were similar, the orientation of the KN would significantly affect the detection feasibility.

We have performed several search campaigns for KN candidates within the Zwicky Transient Facility (ZTF),  the Global Relay of Observatories Watching Transients Happen (GROWTH) and the Electromagnetic counterparts of gravitational wave sources at the Very Large Telescope (ENGRAVE) collaborations. We can distinguish targeted searches after GW triggers and archival searches for serendipitous KN detections. Both kinds of investigations have so far failed to detect KNe with ZTF. Nevertheless, non-detection studies provided meaningful constraints on the luminosity function and the rates of KNe.