An opportunity to construct micro-grids and smart communities through the energy internet has emerged due to the ever-increasing demand for electric vehicles throughout the world. These systems rely on the effective switching and routing of large amounts of electricity. Smart EV charging relies on accurate projections of future EV charging requests to effectively manage EV charging loads. Electric vehicle charging fleets are anticipated to cause a surge in demand for power networks. Maintaining a secure and reliable electrical system depends on accurate demand forecasts from EV fleets. A new hybrid deep learning model for charge prediction is presented in this work. In a hybrid model, we optimise it using a Novel Black Widow Optimisation Algorithm (NBWOA) after integrating a convolutional neural network (CNN) with an RNN. By doing so, we may adjust the hyper-parameters of the RNN against ensure the model’s accuracy, we compare it against datasets that provide actual charging times for EV fleets. We may use a variety of statistical criteria, such as comparisons to existing models, to assess the projected model’s forecast presentation. A deep learning strategy without an optimisation model is less successful than the hybrid approach.