In Intensive Care Unit (ICU) patients with acute kidney injury (AKI) and treated with renal replacement therapy (RRT) often present a fluid overload which is associated with morbidity (mechanical ventilation duration increase, kidney recovery decrease) and mortality. Patients' prognostic could be improved by correcting the fluid overload with net ultrafiltration (UFnet) however it may lead to harmful iatrogenic hypovolemia responsible of deleterious ischemic lesions. In usual practice, UF net prescription are variable and there are different international recommendations. Some observational studies suggest that using a UFnet between 1 et 1.75 mL/kg/h in fluid overloaded patient decrease mortality. Fluid overload increases morbidity and mortality, particularly in RRT. Studies without RRT argue for an efficacy of management by decreasing the fluid overload .Cohort studies suggest to use a moderate UFnet instead of a low UFnet. Some data from studies on early versus late RRT that relate the fluid balance or correct the fluid overload during the early strategy argue for a beneficial effect of an early deresuscitation strategy Consequently, the impact of a moderate UFnet (to decrease the fluid overload) compared to a low UFnet (to stabilize the fluid overload) in a randomized interventional study could be assessed. The study hypothesis is that : an early fluid overload deresuscitation protocol with a high UFnet (2 ml/kg/h) targeting both the negativation of cumulated fluid balance to reach a dry weight and the maintenance of tissue perfusion. Compared to fluid overload deresuscitation protocol with a low UFnet (between 0 and 1 ml/kg/h) to reach a stabilization of cumulated fluid balance without monitoring the tissue perfusion. could improve overall, renal, hemodynamic and respiratory prognosis in fluid overloaded patients with renal replacement therapy in ICU