Personalized electric vehicle energy consumption estimation framework that integrates driver behavior with map data

Sreechakra Vasudeva Raju Rachavelpula; Sangwhan Cha

✨ TL;DR

This paper presents a personalized BEV energy consumption framework that combines driver behavior prediction using Bidirectional LSTM with physics-based energy modeling and map contextual features. The system accurately estimates State-of-Charge depletion by capturing individual driving patterns across diverse road conditions.

01 · Problem

Accurate energy consumption estimation for Battery Electric Vehicles is challenging because it depends on both vehicle physics and highly variable driver behavior. Existing approaches often fail to capture personalized driving patterns such as acceleration/deceleration habits, speed-limit adherence, and responses to road conditions, leading to inaccurate range predictions and suboptimal route planning.

02 · Approach

The framework integrates multiple components: route selection with map-based feature extraction, a rule-based reference velocity generator, a PID controller-based vehicle dynamics simulator, and a Bidirectional LSTM model trained on individual driver data to predict personalized velocity profiles. These predicted velocity profiles are then coupled with a quasi-steady backward energy consumption model to compute tractive power, regenerative braking, and SOC evolution.

03 · Key insights

What the paper shows.

01Combining learned driver behavior with physics-based energy models produces more accurate SOC predictions than either approach alone

02Bidirectional LSTM effectively captures driver-specific patterns including deceleration at intersections, speed-limit tracking, and road grade-dependent responses

03Map-based contextual features (road grades, speed limits, intersection locations) are critical for accurate velocity and energy prediction

04The framework successfully generalizes across diverse driving scenarios including urban, freeway, and hilly routes

04 · Results

The proposed approach demonstrates accurate power and SOC trajectory estimation across multiple route types. The system successfully captures key behavioral patterns such as intersection deceleration, speed-limit adherence, and grade-dependent driving responses, producing personalized SOC depletion profiles that reflect individual driver characteristics.

05 · Limitations

The paper does not provide specific quantitative metrics (e.g., RMSE, MAE) for SOC prediction accuracy or comparative benchmarks against baseline methods. The evaluation scope and dataset size are not clearly specified, and generalization to drivers not represented in the training data is not discussed. Computational complexity and real-time feasibility for in-vehicle deployment are not addressed.

✨ Generated by Claude · Apr 25, 2026 · Read the PDF for authoritative content.

What the paper shows.

↘ Related papers