As part of the OptiEnR research project, the present paper deals with optimizing the multi-energy district boiler of La Rochelle (France) adding to the plant a controlled thermal storage tank. This plant supplies domestic hot water and heats residential and public buildings, using renewable and fossil resources. Due to the complexity of the district boiler as a whole and the strong interactions between the sub-systems, previous works focused first on a modular approach used for the modeling of the plant. Next, a methodology based on both a multi-resolution analysis and the use of artificial neural networks was proposed to forecast the outdoor temperature and the thermal power consumption of the hot water distribution network. The present paper deals first with the modeling of a stratified thermal storage tank. Next, a basic and easy-to-implement controller was developed. Finally, using the global model of the district boiler, a model predictive controller generated optimal command sequences dealing with the flow of the water passing through the storage tank and the wood boiler set-point temperature. As a result, the consumption of fossil fuels, CO2 emissions and functioning cost were significantly reduced. Energy is stored during low-demand periods and used when demand is high, instead of engaging the gas-fuel oil boiler.