Abstract: A system, controller, and method for predictive control of energy management for a segmented load centre. The method including: determining a load prediction for a predetermined prediction timeframe, the load prediction using a machine learning model taking load conditions of the power flow data as input features, the machine learning model trained using historical load conditions; determining a generation prediction for the predetermined prediction timeframe using a renewable generation model, the generation prediction using input features including solar radiation related weather data; determining an energy balance over the prediction timeframe using the load prediction and the generation prediction; modulating a minimum state of charge setpoint based on whether a current value of the setpoint results in curtailed generation in the determined energy balance over the prediction timeframe.