The purpose of this paper is to simulate with ORIENTE® modelling tool, the operation of the MYRTE platform in a mode dedicated to PV output power fluctuations smoothing, and compare the impact of 5 meteorological prediction models on the results. These models are ARMA, MLP, CS (computed global radiation neglecting cloud cover), persistence, and a hybrid predictor developed by the laboratory SPE. The MYRTE platform is funded by the Corsican Authority, the French state and the FEDER funds of the European Union. The partners are CEA and AREVA SE. This platform consists in a PV array coupled to H2 Chain (electrolyser – gas tanks – fuel cell) and electricity converters to regulate the injection on the electrical grid of Corsica. The different indicators considered in the comparison of the different simulations are on the one hand linked to energy balance in the system and on the other hand statistical indicators (RMBE, rRMSE, and CC) of each subsystem. The main result shown in this paper is that none of the prediction models is clearly distinguishable from others on the basis of those indicators. Best results are obtained with the hybrid model, which has a lower rRMSE than persistence model of approximately 15%.