Electric Vehicle Range Estimation Using Neural Network Compensation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for estimating the remaining range of electric vehicles based on the state of charge of power batteries are inaccurate due to the linear model's inability to consider the battery's attenuation and driving propensity, affecting energy management.
Innovation Solution
A method using a neural network, specifically a radial basis function neural network, to compute a range compensation coefficient by integrating the state of charge, number of charging-discharging cycles, and energy consumption, providing a non-linear model for precise range estimation and energy management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a linear model of state of charge is used to estimate remaining range, then the estimation method is simple, but the accuracy of remaining range estimation deteriorates
Solution Approach 1:
The patent transforms the estimation approach from using only state of charge (one parameter) to using multiple parameters including state of charge, driving propensity, and battery attenuation characteristics. This multi-parameter approach enables the neural network to capture the non-linear relationships between these factors and remaining range, significantly improving estimation accuracy while maintaining computational feasibility through the neural network architecture.
2Measurement precision
If multiple factors including battery attenuation and driving propensity are considered, then the remaining range estimation accuracy is improved, but the model complexity increases
Solution Approach 1:
The patent introduces a neural network as an intermediary computational layer that processes multiple input parameters (state of charge, driving propensity, battery attenuation) and transforms them into an accurate remaining range estimation. The neural network acts as a mediator that handles the complexity of multi-factor relationships internally, providing accurate results without requiring the external system to manage the computational complexity directly.
Solution Approach 2:
The patent employs a dynamic estimation model that adapts to changing conditions through the neural network's ability to process varying inputs of state of charge, driving propensity, and battery attenuation. Unlike static linear models, this dynamic approach continuously adjusts the estimation based on current operational conditions and battery health status, capturing the non-linear dynamics of battery performance degradation and usage patterns.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
A method and system for estimating a remaining range of an electric car, and electric car. The method comprises: acquiring a discharge current and a discharge voltage of a power battery, and computing a current released energy amount of the power battery according to the discharge current and the discharge voltage; acquiring a state of charge and the number of charge-discharge cycles of the power battery; inputting the state of charge, the number of charge-discharge cycles, and a total released energy per unit of time of the power battery into a neural network to acquire a range compensation coefficient; computing a travel distance per unit of time of a vehicle when the vehicle is in a moving state; and computing, according to the travel distance per unit of time, the range compensation coefficient, the state of charge of the power battery, and a difference value of the state of charge of the power battery in one time unit, a remaining range of the vehicle. The method can accurately compute a remaining range of an electric car.