The aim of this thesis was to better understand the cosmic history of galaxy formation thanks to the largest cosmological survey with the large ALMA interferometer. This observation was preformed in a region of the sky, the GOODS-South field, which benefits from the deepest observations of the Hubble Space Telescope, the Spitzer and Herschel infrared space observatories, as well as data from the Chandra X-ray telescope and the VLA Radio Interferometer. These observations at 1.1mm with ALMA therefore complete this multi-wavelength panorama, and make this region of the sky a leading laboratory for the study of the evolution of galaxies. The observation with ALMA allows us to observe this region of the sky without being affected by the confusion limit that affected Herschel, and to search for more distant galaxies. For the first time, we can study dust-obscured star formation at z > 2 over a large enough area to reduce different observational biases.A large part of this thesis was devoted to the scientific exploitation of this 1.1mm cosmological image. This was done by analysing the image from the interferometric data, precisely characterising the survey, defining detectability thresholds and indicators that can quantify the credibility of detections, and carrying out simulations on these images. We then extracted and identified the galaxies present in the image.This analysis shows that ALMA surveys can reveal new galaxies that are not detected by the deepest surveys conducted with the Hubble Space Telescope. These ``dark" galaxies are among the most massive and distant galaxies in this region of the sky. The discovery of these new dark galaxies, which represent of the order of 10 - 20% of the ALMA detections, suggests that the number of massive star-forming galaxies in the distant universe may be much larger than previously expected. This work has also made it possible to determine properties of the galaxies detected by ALMA, through the modelling of their spectral energy distributions: the star formation rate, the masses of gas and dust, the dust temperature, the time required for a galaxy to consume its gas, the relationship between the infrared luminosity and the radio luminosity, and the excess of the infrared component in the spectrum of a galaxy. Analysis of these findings suggests that massive galaxies at high redshift consume their gas slowly to form stars and indicate that these galaxies are the ideal progenitors of passive galaxies at z∼2.