Prediction-based optimal system design of an improved building-to-vehicle-to-building energy community under different operation strategies concerning three PV installation types

In alignment with carbon-neutral targets, the growing reliance on renewable energy sources increases utility grid burden and emphasizes techno-economic requirements, necessitating enhanced design and operational strategies for distributed energy systems. This study firstly introduces an improved grid-connected building-to-vehicle-to-building system with onsite battery and photovoltaic paver, where bi-directional electric vehicles serve as energy vectors connecting residential and office building clusters. Various photovoltaic installations, including rooftops, façades, and pavers, are considered for higher system flexibility potential, with their respective annual outputs in Hong Kong at 1.34, 0.81, and 0.97 kWh/Wp. Better flexible energy management is shown with the proposed system prototype, with the system self-consumption rate, self-sufficiency rate increases, and CO2 emission reduction by 27.22 %, 5.54 %, and 1234.01 tons. The electric vehicle share increase influence is assessed and resulted in better economic and environmental performances, with increased 60.11 % carbon emission reduction for the basic strategy. To address demand and renewable generation uncertainty, a 2-layer Long Short-Term Memory model and a pseudo-3D room model are developed, achieving a photovoltaic generation root mean square error of 0.052 kWh (paver), 0.059 kWh (rooftop), and 0.042 kWh (façade). Thus, two predictive-control strategies are proposed based on the maximum self-consumption and time-of-use strategies through the heuristic algorithm to optimize both system capacity and operation control. The proposed predictive-control strategy is shown to decrease the electricity bill by up to 8.56 %, cutting down on the initial investment, meanwhile sacrificing the technical performance compared to the basic maximum self-consumption strategy.