The power grid is rapidly experiencing a transformation driven by renewable and climate targets which pose a huge challenge to maintain the system stability and reliability, thus balancing services become more crucial now than ever. To provide balancing services in a cost-efficient way, it is necessary to develop predictive models which can optimize power portfolio in an online manner. This paper presents the development of an Online Predictive Dispatch Optimizer and its connection with a grid model that simulates power and frequency control within interconnected power systems. The performance of the dispatch optimizer and its connection with the grid model is tested by simulating several cases where the adequacy of the model is confirmed, especially regarding its ability to manage energy imbalance in real-time. The outcomes and flexibility of this framework can be used to quantitatively evaluate the operation of power portfolio owners in the power grid.