هومن گرشاسبي

In this research, a method for estimating the charge level of the lithium-ion battery of electric vehicles has been introduced. The advantage of this method is to have low computational load at the same time with high accuracy with the help of radial basis neural network. The proposed method, with the help of machine learning methods, designs a neural network that is competitive with more complex methods and networks in terms of performance accuracy, but the point that makes this design attractive is the low computational load of the network, which makes its implementation efficient in online applications. does In this research, at first, the data is cleaned with the isolated tree method, in the next step, a set wavelet transform is applied to the data, then it enters the neural network and is optimized with Adam's algorithm. Then, the accuracy of the model is measured with the help of the average absolute error cost and root mean square functions. In this research, the US standard motion cycles, namely UDDS, HWFET, LA92, US06, and the combination of these cycles at a temperature of 25 degrees Celsius for training and a variable temperature from 10 degrees to 25 degrees for eva‎luation were used to obtain more valid and comprehensive results. It is clear that the proposed architecture has an average absolute error of about 1.8% at constant temperature and 1.9% at variable temperature. that the proposed architecture, compared to a long-short-term memory neural network, has a lower sensory load with high accuracy, which makes it possible to implement this architecture in online applications.

شبكه عصبي پايه شعاعي , يادگيري ماشين , باتري ليتيوم-يون , باتري پاناسونيكPF18650 , تبديل موجك

RBF neural network , machine learning , lithium-ion battery , Panasonic PF18650 battery , wavelet transform

Houman Garshasbi

Behrooz mashadi- Abdollah amirkhani