Solid electrolyte ion conductivity prediction method based on a BP neural network

Abstract

The invention discloses a solid electrolyte ionic conductivity prediction method based on a BP neural network, and the method is characterized in that the method comprises the following steps: 1, collecting material data, carrying out the preprocessing of the material data, obtaining a total sample set, and enabling the material data to be divided into a test set sample and a training set sample;2, constructing a BP neural network model according to the step 1; and 3, selecting a transfer function, a training function and a learning function in the BP neural network model, and initializing each parameter. According to the method, the ionic conductivity performance of the solid electrolyte containing elements such as Li, Na, Mg and Al can be accurately predicted, corresponding components and structures of the solid electrolyte are prepared according to the prediction result, the blindness in the electrode material design process can be reduced, and a large amount of time and cost are saved.

Description

technical field [0001] The invention relates to the field of artificial intelligence, and specifically provides a BP neural network-based method for predicting the ion conductivity of solid electrolytes. Background technique [0002] Lithium-ion batteries are widely used because of their high energy density and good rate characteristics. However, due to the potential safety hazards of the electrolyte, the scope of use is limited. Solid electrolyte materials can inhibit the growth of lithium dendrites at the negative electrode, and have non-combustible properties, which can fundamentally solve the problem of lithium battery safety, and are key materials for the next generation of lithium batteries. Traditional material development mainly adopts the "trial and error" experimental method, and a large number of repeated iterative experiments are carried out in the way of "hypothesis-experimental verification", so that the experimental material is constantly approaching the targ...

AI Technical Summary

Problems solved by technology

But this method is inefficient, and the speed of new material development lags far behind the speed of new product development

Since the 20th century, it takes about 10 years for a new material to go from research and development to practical application, which has been unable to meet the demand for new materials in fields such as lithium batteries

However, the above-mentioned technology still has the following defects: 1. When the amount of data is very large, the demand for hardware equipment for support vector machine training is often too large

2. The logistic regression model is prone to underfitting, and the classification accuracy may not be high

3. The above two mathematical models are only suitable for predicting specific material properties and do not have universal applicability

Images