Modelling and simulation of a predictive BESS controller based on load forecasting in a South European island power system
Modern isolated power grids are constantly evolving to adopt smart grid concepts that can permit higher renewable energy penetration and energy management optimization, in the view of a sustainable RES based energy production EU policy with reduced pollutant emissions. Nevertheless, many islandic power systems like the islands in Southern Europe are still depending on oil-fired diesel engines, while the renewable energy production is limited due to financial, technical and environmental reasons. In this study, the power system of a typical non-interconnected South European island consisting of diesel generators and a PV farm is modelled and simulated. Scope of this paper is to examine the ability of a Battery Energy Storage System (BESS) to achieve load peak shaving combined with maximization of the PV power penetration into the grid leading to pre-planned zero curtailment. For this purpose, a novel peak shaving algorithm is developed and implemented into an Energy Management System (EMS), for optimal scheduling of the diesel engines. Thereinafter, dynamic simulations of the island's power system are carried out employing a predictive control strategy for different time scales, ranging from a supervisor BESS controller based on load forecasting, to a real-time battery power regulation. The predictive BESS controller is based on future consumption values forecasting, which in turn result from an Artificial Neural Network (ANN) and an optimization procedure taking into account PV power generation and a peak shaving threshold. Thus, a new diesel engine scheduling is obtained capable of replacing the maximum peak power demand with renewable power while at the same time load curve smoothening and reduced diesel generators ramps-up are achieved. The simulations are executed in APROS (Advanced Process Simulator) dynamic simulation platform, using built-in components for the BESS modelling, an external model for load forecasting and a user-developed EMS structure.