Vehicle Energy Consumption Estimation Using Offline-Trained CAE Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional energy consumption estimation techniques for electrified vehicles are computationally intensive and require substantial experimental validation, making them unsuitable for real-time applications, and they struggle with accurate determination of various vehicle and environmental parameters.
Innovation Solution
A machine-learning model, such as a recurrent neural network (RNN) or artificial neural network (ANN), trained offline using computer-aided engineering (CAE) software, is used to estimate energy consumption in real-time by integrating parameters like ambient temperature, wind speed, vehicle mass, and road grade.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If physics-based formulations are used for energy consumption estimation, then measurement precision is improved, but device complexity increases due to computational intensity
Solution Approach 1:
The patent pre-calculates energy consumption values using physics-based formulations during an offline training phase, storing results in lookup tables. During real-time operation, the controller simply retrieves pre-computed values based on current operating conditions, avoiding the need to execute complex physics calculations online while maintaining high estimation accuracy.
Solution Approach 2:
The patent creates a simplified digital copy of the complex physics-based energy consumption model by training a neural network on pre-computed physics model data. This neural network copy can be executed efficiently on resource-constrained controllers while reproducing the accuracy of the original physics-based formulation without requiring its computational complexity.
2Measurement precision
If physics-based formulations are used for energy consumption estimation, then measurement precision is improved, but loss of time increases due to computational intensity
Solution Approach 1:
The patent performs all computationally intensive physics-based calculations in advance during an offline training phase, storing the results in pre-computed lookup tables. During real-time vehicle operation, the system only needs to perform simple table lookups based on current operating parameters, reducing real-time processing time from seconds to milliseconds while maintaining the accuracy benefits of physics-based formulations.
3Measurement precision
If extensive experimental validation is conducted to provide empirical data, then measurement precision is improved, but loss of substance increases due to data requirements
Solution Approach 1:
The patent creates a virtual copy of extensive experimental test data by using physics-based simulations to generate synthetic training data that covers the full operating range of the vehicle. This simulated data serves as a proxy for actual experimental measurements, providing sufficient training data for machine learning models without requiring physical vehicle testing under all possible conditions.
Solution Approach 2:
The patent varies key operating parameters (speed, acceleration, road grade, temperature, wind conditions) in physics-based simulations to generate diverse training data across the entire vehicle operating envelope. By systematically changing these parameters in virtual environments, the system generates comprehensive training datasets without needing to conduct physical experiments for every parameter combination.
Data Source
AI summary
An energy consumption estimation and control system for an electrified vehicle includes a memory configured to store a trained energy consumption model configured to estimate an energy consumption of the electrified vehicle, wherein the trained energy consumption model was previously trained offline using training data generated by computer-aided engineering (CAE) software and a control system configured to access the trained energy consumption model from the memory, obtain a plurality of input parameters for the trained energy consumption model based on real-time parameters of the electrified vehicle, utilize the trained energy consumption model and the plurality of input parameters to estimate the energy consumption of the electrified vehicle in real-time, and generate an output based on the estimated energy consumption of the electrified vehicle.