New energy vehicle fast charging reservation algorithm based on internet of things coordination

Abstract To meet the rising demand for electric vehicles (EVs), effective and dependable fast-charging reservation systems are required. Conventional charging reservation systems frequently lack coordination between user preferences, real-time station status, and environmental factors, leading to poor user experiences and station ineffectiveness. Existing methods to EV charging reservation systems fail to account for dynamic real-world conditions such as changing traffic patterns, station uptime, and IoT sensor inputs, resulting in suboptimal station allocation and failed reservations. This study fills a gap by proposing an IoT-Coordinated EV Fast Charging Reservation Approach (IoT-CFCRA), which uses real-time data to predict reservation success and suggest the best charging stations under different conditions. The IoT-CFCRA uses the IoT-Enhanced EV Charging Reservation Dataset, which contains user attributes, vehicle data, and IoT-enhanced station data like vehicle type, battery level, distance to station, sensor status, and real-time traffic. Data preprocessing entails normalization, encoding, and feature selection to find important features. A Support Vector Machine (SVM) model is trained to predict reservation success through hyperparameter tuning and 80−20 data splitting. The algorithm also includes a station scoring method that considers IoT uptime, distance, traffic conditions, and membership status to provide ranked station suggestions. Users receive real-time notifications to help them adapt to traffic conditions and reservation results. The experimental results show that the proposed IoT-CFCRA approach outperforms other methods. It achieved an accuracy of 87%, outperforming the best baseline (Gradient Boosting) by 4% and improving on Logistic Regression by 10%. The AUC score of 0.92 indicates excellent discriminative capability, a 5-point improvement over Random Forest. The F1-Score of 0.84 demonstrates a strong balance of precision and recall, outperforming SVM by 8%. Furthermore, RMSE was reduced to 0.25, indicating a 19.4% decrease in prediction error when compared to KNN (RMSE = 0.36). The cross-validation score of 88% confirms the model’s robustness, outperforming the next-best performing model by 4%. These metrics highlight the resilience of the IoT-CFCRA in creating precise reservations and suggesting optimum charging stations. The IoT-CFCRA seamlessly combines IoT capacities with machine learning to tackle dynamic factors that influence EV charging reservations. The proposed approach encourages user-centered decision-making and effective resource allocation, laying the groundwork for future advances in IoT-driven EV infrastructure.