Artificial Intelligence-driven optimization of V2G and charging point selection en-route: A systematic literature review

• 114 research papers about V2G are analyzed. • Profit for electric vehicle owner can be increased by improving ML algorithms. • More datasets from Australian required to improve prediction of charging behavior. • Selecting charging stations along travel routes in Australia is not yet explored. The common implementation of electric vehicles (EVs) necessitates effective integration strategies to ensure their seamless interaction with the electrical grid while meeting private EV customer preferences and enhancing grid stability. This review aims to enhance comprehension of both industrial and academic applications of vehicle-to-grid (V2G) technology, with a specific emphasis on maximizing profits for private EV owners seeking to engage with this technology. Additionally, it provides an overview of prevalent machine learning (ML) algorithms suitable for addressing V2G-related challenges, along with discussions on predicting EV owners’ charging behavoir and challenges associated with Electric Vehicle Routing Problems (EVRP). Notably, this review highlights the distinctive requirements and characteristics of the Australian consumer base and grid infrastructure. It underscores a notable gap in understanding how private customers utilizing V2G-capable EVs can maximize benefits, particularly given Australia’s significant share of photovoltaic (PV) energy generation and its unique EVRP layout necessitating longer travel distances between major urban centres. Further investigation is warranted into the impact of high PV generation shares on charging behavoir and the optimization of charging strategies within the Australian context. For the industry, the primary advantage of this paper lies in its examination of existing profit maximization models and the consequent heightened interest from customers in this technology plus the strategic placement of charging stations. For academic researchers, this paper demonstrates that the prediction of charging behavoir can be effectively tackled through the appropriate selection of ML-algorithms.