The promise of the transactive energy concept to develop an economic-based control mechanism for power system applications has received significant attention. Transactive energy can enable the system operator to exploit supply and demand-side flexibility at their full potentials. This paper presents strategies for energy flexibility of a solar net-zero energy institutional public building, located in Varennes (Quebec, Canada). The aim of this study is to assess the energy flexibility in this building, as a fully electric public unit, by optimizing the relationship between building-integrated photovoltaic system (BIPV) capacity, geothermal heat pump size, and radiant floors with thermal storage. A Machine Learning algorithm with an adaptive window learns the thermal dynamics of the building using historical data, and a Model Predictive Control (MPC) optimizes the building’s energy management. The MPC considers Fixed, Rolling, and Receding horizons with one, three, and five days ahead forecasts. The study quantifies the impact of the horizon selection over the available flexibility of the building during the two critical periods of the grid. Numerical simulation results demonstrate that the performance of the three MPC horizon types achieves reductions in energy consumption ranging from 20.1% to 48.6% and 9.5% to 27.5% in morning and evening peak periods.