We present a proactive energy management framework that integrates predictive dynamic building models and day-ahead forecasts of disturbances affecting efficiency and costs. This enables an efficient management of resources and an accurate prediction of the daily electricity demand profile. The strategy is based on the on-line solution of mixed-integer nonlinear programming problems. The framework is able to integrate forecasts of weather conditions, fuel prices, heat gains, and utility demands. In addition, it can capture net-metering interactions using agent-based market models. We claim that a large adoption level of this proactive technology can improve the predictability of the overall electricity demand at high-level power grid operations such as unit commitment and economic dispatch which can be used to minimize the overall reserves.