From my understanding the deterministic forecast (i.e. operational run) calculates one particular outcome/forecast at high resolution based on one set of initial weather conditions (which are estimated based on observations).
However, the initial weather conditions can never be perfectly known due to incomplete geographic coverage of weather observation networks and uncertainty in measurements etc.
To account for this uncertainty, ensemble predictions are created by running the same model multiple times at a lower resolution using a range of slightly perturbed/adjusted (but still plausible) initial weather conditions as input data. Examining the range of different forecasts that result from the set of slightly different weather conditions gives an idea of the uncertainty in the forecast and allows the deterministic run to be considered in the context of these possible outcomes (e.g. mild outlier etc.). However the deterministic run is still valuable because its higher resolution resolves weather features better, particularly small scale features that may be critical for how a forecast develops.