"Electric Vehicle (EV) dataset"

"The rapid growth of electric vehicle (EV) adoption is re-shaping distribution networks frompassive infrastructures into stochastic cyber\u2013physical systems. Uncoordinated charging at public and semipubliccharging facilities can aggravate evening peaks, increase operating costs, and shift emissionsfrom tailpipes to carbon-intensive generators. This paper proposes a consumer-centric, carbon-awaresmart grid framework for dynamic EV charging management that tightly couples long-horizon demandforecasting, multi-objective charging schedule optimization, and a personalized recommendation engine.First, a hybrid convolutional neural network\u2013long short-term memory (CNN\u2013LSTM) model is developedto forecast multi-step charging demand using a fused dataset that combines a large-scale synthetic EVcharging corpus generated via conditional tabular generative adversarial networks with real chargingstationrecords. Temporal features (hours, weekday, season), electricity prices, and time-varying marginalemission factors are encoded to produce horizon-specific forecasts (7, 30, 60, and 90 days ahead). Acrossall horizons, the CNN\u2013LSTM architecture consistently outperforms benchmark models, including vanillaLSTM, bidirectional LSTM, gated recurrent unit (GRU), sequence-to-sequence LSTM, and a compactTransformer variant, achieving up to 15\u201320% lower root-mean-square error and 10\u201318% higher R2 on thetest set, while maintaining training times suitable for daily retraining. Inorder to coordinate EV charging ina grid constrained environment, the present study combines carbon-aware dynamic scheduling, user centricdecision support and deep learning based forecating strategies. To minimize a weighted combination ofenergy cost and feeder level carbon emissions while satisfying practical constraints, the EV chargingis formulated as mixed integer multi objective optimization problem thatincludes feder capacity limits,charger ratings, user arrival departure windows and required state of charge levels. A heuristic solve withCNN-LSTM demand and carbon forecasts, enables near-real time scheduling and reshapes aggregate loadprofiles, reducing daily peak demand by approximately 20\u201325% and charging related emissions by 30\u201340% compared with uncontrolled plug-in charging, with negligible violations of user time constraints.By considering carbon awareness, scoring candidate time windows against user specific price sensitivityand convenience preference further optimized schedules are then translated into personalized chargingrecommendations thereby showcasing proposed framework provides better operational optimization andhuman centered personalization can be co-designed to support scalable, low-carbon EV integration intofuture smart grids."