Bayesian deep neural network for low-carbon scheduling strategy planning of V2G

With the increasing adoption of electric vehicles, the Vehicle-to-Grid (V2G) model has become crucial in integrating renewable energy generation. However, challenges exist in developing grid scheduling strategies that are tailored to different regions, as well as in quantifying the economic benefits and carbon emissions associated with such scheduling. To address these issues, our study proposed a novel V2G low-carbon scheduling strategy planning method based on Bayesian neural networks. Initially, we established a stochastic V2G model that incorporated grid and electric vehicle scheduling, along with a mathematical model that captured the random behavior of EV users, enabling us to capture the essential characteristics of scheduling planning. Subsequently, we employed an enhanced Bayesian deep neural network to learn and assimilate these scheduling planning characteristics, allowing for the provision of a grid scheduling strategy that ranks economic benefits based on weighted priorities. Furthermore, we conducted simulation experiments within the coverage area of the Internet of Things in Energy (IoTE) to gather scheduling characteristics. The experimental results demonstrated that our method outperformed other deep learning models in terms of voltage amplitude stability during the grid scheduling process, showcasing superior robustness and accuracy. Moreover, we evaluated the economic benefits of the scheduling model and compared it with the original V2G scheduling model. The findings revealed that our model exhibited higher economic benefits and lower carbon emissions. Considering the future challenges of low-carbon urban planning, our method holds significant potential in terms of grid scheduling economic benefits and carbon emission indicators.