LSTM Battery Aging Prediction from Voltage-Time Curves
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Lithium-sulfur (Li—S) batteries face challenges due to the polysulfide shuttle phenomenon, which depletes active sulfur mass and forms deleterious layers, and existing synthesis methods for TiO2—TiN composites result in inconsistent structures, compromising their stability and effectiveness. Additionally, battery performance evaluation is time-consuming, and predicting aging effects is difficult.
Innovation Solution
A processor-implemented method using a trained long short-term memory (LSTM) network to predict aging effects in batteries, combined with a TiO2—TiN/S composite with a Super P® coated separator, and a controlled oxidation process to synthesize TiO2—TiN composites with a uniform structure for enhanced conductivity and polysulfide adsorption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If high-temperature oxidation process is used to synthesize TiO2—TiN composite, then the composite can be formed, but the structure becomes inconsistent and phase separation occurs
Solution Approach 1:
The patent changes the synthesis parameters from high-temperature oxidation to a two-step process involving nitration at lower temperature followed by controlled oxidation. This parameter change enables consistent structural formation while maintaining manufacturing feasibility, resolving the contradiction between structural consistency and ease of manufacture.
Solution Approach 2:
The synthesis process is segmented into two distinct steps: first nitration of TiN to form intermediate product, then controlled oxidation to form the final TiO2—TiN composite. This segmentation allows precise control over the oxidation degree and structural formation, eliminating phase separation while simplifying the overall manufacturing process.
2Reliability
If TiO2 material is used for polysulfide adsorption, then polysulfide capture is improved, but electrical conductivity decreases
Solution Approach 1:
The patent creates a composite material TiO2—TiN where TiO2 provides polysulfide adsorption capability and TiN provides electrical conductivity. This composite structure resolves the contradiction by combining the advantages of both materials, achieving both reliable polysulfide capture and maintained electrical conductivity.
3Measurement precision
If cyclic charge-discharge tests are performed for battery evaluation, then performance data is obtained, but testing time becomes excessively long
Solution Approach 1:
The patent performs preliminary structural optimization of the TiO2—TiN composite before battery assembly, ensuring the material has consistent structure and high conductivity. This preliminary action reduces the need for extensive cyclic testing, as the material performance is already optimized, thereby reducing testing time while maintaining evaluation accuracy.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The method provides accurate predictions of battery aging up to 100 cycles in advance, improving efficiency and safety, and the synthesized TiO2—TiN composite enhances Li—S battery performance and longevity, making them more practical for commercial use.
Implementation Method 1
inputting the data of voltage and time curve into a trained long short-term memory (LSTM) network; outputting the predicted data of voltage and time curve
Implementation Method 2
engage in Lewis acid interactions to adsorb polysulfides
Implementation Method 3
The synthesis of titanium nitride (TiN), although noted for its catalytic influence on polysulfide conversion and exceptional polarization effect, often involves complex, high-temperature processes or the use of hazardous chemicals
Data Source
AI summary
A processor implemented method of predicting the aging effects in a battery, the method comprises capturing data of voltage and time of the battery; inputting the data of voltage and time curve into a trained LSTM network; and outputting the predicted data of voltage and time curve; wherein the trained LSTM network is configured to be trained by the following steps: selecting the input data sample; defining the initial hidden state, the initial cell state, bias, weight, current weight; setting an epoch, an initial learning rate, a gradient threshold, and a drop factor; and training the LSTM network as per the set parameters.