Abstract: Automatically managing surplus and deficiencies on an electrical distribution network comprises sending and receiving information from a connected control point of a computer-based grid controller, and operating the control point, including: monitoring the energy surplus and deficiency state of the electrical distribution network to predict the energy surplus and deficiencies on the electrical distribution network. Locally trained machine learning models are run against collected data corresponding to an electrical distribution network to generate predictions of energy consumption and generation at one or more sites that each include at least one locally controlled energy management system.