Scholarlink

Electric vehicle(EV) range prediction is pivotal for adoption, affecting route planning, charging, and user confidence. A specifically curated dataset, formed via surveys and simulations, would provide in-depth specifics about EV usage customized for India. To broaden its scope, it could employ statistical techniques alongside a distinctive fusion of primary data and Probabilistic Conditional Generative Adversarial Networks (PCGANs). Despite this, traditional models like Artificial Neural Networks (ANN) struggle with inherent EV range uncertainty. Our research introduces Multi-Modal PCGANs (MM-PCGANs) to forecast EV range, offering a range of predictions with associated probabilities, expressing the model’s uncertainty based on input features. Through rigorous experimentation with our enhanced dataset, the MM-PCGANs model demonstrated superior performance, showcasing a 0.73% higher R2 score and a significant 63.89% lower Mean Squared Error (MSE) compared to the baseline ANN model. These findings highlight the accuracy of our approach in predicting EV range, emphasizing the efficacy of probabilistic generative models in alleviating range anxiety and delivering reliable predictions.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Multi-modal Probabilistic Conditional Generative Adversarial Networks for Electric Vehicles Range Prediction

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Prajna Dora,
  • N. Harini

摘要

Electric vehicle(EV) range prediction is pivotal for adoption, affecting route planning, charging, and user confidence. A specifically curated dataset, formed via surveys and simulations, would provide in-depth specifics about EV usage customized for India. To broaden its scope, it could employ statistical techniques alongside a distinctive fusion of primary data and Probabilistic Conditional Generative Adversarial Networks (PCGANs). Despite this, traditional models like Artificial Neural Networks (ANN) struggle with inherent EV range uncertainty. Our research introduces Multi-Modal PCGANs (MM-PCGANs) to forecast EV range, offering a range of predictions with associated probabilities, expressing the model’s uncertainty based on input features. Through rigorous experimentation with our enhanced dataset, the MM-PCGANs model demonstrated superior performance, showcasing a 0.73% higher R2 score and a significant 63.89% lower Mean Squared Error (MSE) compared to the baseline ANN model. These findings highlight the accuracy of our approach in predicting EV range, emphasizing the efficacy of probabilistic generative models in alleviating range anxiety and delivering reliable predictions.