This study presents the implementation of a model predictive control strategy (MPC) applied to a liquid-based photovoltaic/thermal-thermal storage system for the offices of a school building near Montreal, Canada to enhance the building energy flexibility during the electrical grid peak demand events in winter. The methodology was applied for 3 days in January with different solar conditions, including a sunny day, a cloudy day, and a partly sunny/cloudy day, and 2 different configurations of peak demand events. The MPC strategy was compared to the same system installation with rule-based control (RBC) to show the advantages of the predictive controls. The MPC strategy compared to RBC reduced the event peak load by 21.5%, the energy consumption by 2.9%, and the energy cost by 14.0% in addition to improving the electrical generation self-consumption by 24.2%.