Hybrid EV Energy Control Optimization Using Gradient-Boosted Surrogates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for optimizing the energy control management system of hybrid electric vehicles are time-consuming and fail to provide an optimal solution due to complex multi-objective optimization problems, particularly in quantifying vehicle speed and torque, and fitting complex physical models.
Innovation Solution
A multi-objective optimization method assisted by a gradient boosted neural network, utilizing a Kriging model to establish an approximate function and a NSGA-II algorithm to solve the approximate function, transforming the problem into an optimal solution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing modal analysis methods (finite element analysis, experimental modal analysis) are used to approximate real physical models, then model accuracy is improved, but model complexity and evaluation time increase significantly
Solution Approach 1:
The patent creates simplified proxy models (binomial approximation, support vector machine, gradient boosted neural network) that copy the essential input-output relationships of complex physical models without replicating their full complexity. These proxy models serve as lightweight replicas that can be evaluated rapidly while maintaining acceptable accuracy for optimization purposes.
Solution Approach 2:
The patent replaces traditional mechanical/modal analysis methods with machine learning-based proxy models. Instead of using computationally intensive finite element analysis or experimental modal analysis, the system substitutes these with data-driven models that can be trained once and then evaluated rapidly for optimization iterations.
2Measurement precision
If existing modal analysis methods are used for multi-objective optimization, then model accuracy is improved, but optimization time becomes excessively long
Solution Approach 1:
The patent performs preliminary training of proxy models using data from finite element analysis or experimental modal analysis before the optimization process. This preliminary action creates pre-trained models that can be rapidly evaluated during optimization iterations, avoiding the need to run complex simulations repeatedly.
Solution Approach 2:
The patent creates simplified proxy models that replicate the behavior of complex physical models. These copies can be evaluated rapidly during optimization iterations while maintaining sufficient accuracy, thus reducing evaluation time significantly compared to using the original complex models directly.
3Loss of time
If proxy models (binomial approximation, support vector machine) are used to fit complex functions, then evaluation time is reduced, but ability to fit complex multi-objective optimization functions is insufficient
Solution Approach 1:
The patent employs gradient boosted neural networks with adjustable hyperparameters (learning rate, number of estimators, max depth, subsampling ratio) that can be tuned to match the complexity of the underlying physical system. By changing these parameters, the model can adapt to fit complex non-linear relationships while maintaining computational efficiency.
Solution Approach 2:
The patent uses composite modeling approaches by combining gradient boosted decision trees with neural network elements, creating a hybrid model that leverages the strengths of both approaches. This composite structure enables the model to capture complex non-linear relationships better than simple binomial approximation or basic support vector machines while remaining computationally efficient.
4Manufacturing precision
If complex physical models are used for energy control management optimization, then solution accuracy is improved, but computational complexity and resource requirements increase
Solution Approach 1:
The patent creates proxy models that copy the essential input-output behavior of complex physical models. These proxy models require minimal computational resources during optimization iterations while producing sufficiently accurate results for control parameter optimization in hybrid electric vehicles.
Solution Approach 2:
The patent substitutes computationally intensive physical models with machine learning-based proxy models. This substitution dramatically reduces computational complexity and resource requirements during optimization iterations while maintaining sufficient accuracy for practical control applications.
Data Source
AI summary
Provided are a multi-objective optimization method and system assisted by a gradient boosted neural network, and a device, relating to the technical field of objective optimization. The method includes: obtaining sample data of an energy control management system of a hybrid electric vehicle through design of experiments; determining an approximate function for a multi-objective optimization problem of the energy control management system based on a gradient boosted neural network algorithm; and solving the approximate function by using a multi-objective optimization algorithm to obtain an optimal solution. The present application can solve issues such as time-consuming function evaluation and inability to obtain an optimal solution of the problem.