Multi-objective battery pack structure optimization method based on neural network

Abstract

The invention discloses a multi-objective battery pack structure optimization method based on a neural network, and the method comprises the following steps: S1, converting a three-dimensional model of a battery pack into a two-dimensional model, carrying out the simulation in COMSOL, and generating a plurality of groups of structure parameters which affect the performance value of the battery pack; s2, taking the multiple groups of structural parameters as the input of a neural network, taking the performance value of the battery pack as the output of the neural network, and training the neural network by adopting a Bayesian regularization algorithm; and S3, carrying out structure parameter optimization on the trained neural network in the step S2 by adopting an NSGA2 algorithm to obtain multiple groups of Pareto optimal solutions, and selecting one group of the solutions with the lowest highest temperature in the Pareto optimal solutions as a final scheme of the battery pack. According to the invention, the optimization period of the battery pack structure is greatly shortened, and the performance of the battery pack is improved.

Description

technical field [0001] The invention relates to the technical field of battery systems, in particular to a neural network-based multi-objective optimization battery pack structure method. Background technique [0002] With the increasing problems of air pollution and resource shortage, more and more countries have begun to vigorously promote the development of electric vehicles. The core of electric vehicles is power batteries, and the performance of power batteries affects the quality and user experience of electric vehicles. The impact of temperature on the performance of power batteries is significant. During high-current discharge, if the heat generated by the battery is not dissipated in time, the battery will experience problems such as life decay, combustion, and even explosion. Therefore, in order to maintain the temperature within the optimal range and increase the cruising range of electric vehicles, designing a reasonable battery pack is of great significance for...

AI Technical Summary

Problems solved by technology

[0003] At present, three-dimensional battery packs are widely used for battery pack performance simulation, but each simulation of a three-dimensional battery pack takes a lot of time

When using big data to train the neural network, a large number of simulations are required to obtain the performance data of the battery pack. Therefore, the 3D simulation of the battery pack will consume a lot of time, and it may even take several months to obtain the trained neural network.

Moreover, the previous battery pack optimization only aimed at the highest temperature, without considering the power consumption required for the heat dissipation of the battery pack

Therefore, due to the long process time and poor consideration of factors in the current battery pack structure simulation, the structure setting of the battery pack will be unreasonable.

Images