Scribd
Upload
13 views1 page

Desc

Uploaded by

sruthi.g2020
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as TXT, PDF, TXT or read online on Scribd
13 views1 page

Desc

Uploaded by

sruthi.g2020
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as TXT, PDF, TXT or read online on Scribd
You are on page 1/ 1

Today, lithium-ion batteries are widely used in various fields such as Electric

Vehicles and Energy Storage Systems5. Therefore, the need for an effective
prediction algorithm for battery charge status and lifespan is increasing. SOH,
which stands for State of Health, refers to the battery’s aging state, meaning its
capacity to hold a charge. If we define the cycle as the number of battery
discharges, ( Q_n ) as the discharge charge amount in the nth discharge cycle, and
( Q_{std} ) as the standard charge amount provided at the time of battery
manufacture, then SOH ( = \frac{Q_n}{Q_{std}} ). SOH is calculated using the
Capacity Method with data from NASA PCoE6. A value is calculated for each Discharge
Cycle, ranging between 0 and 17. As the cycle increases, SOH generally decreases.
The standard capacity of SOH at the time of battery manufacture is 100%, and it is
commonly used until SOH decreases to 60% ~ 70%. We aim to conduct a SOH trend
prediction simulation according to the number of Discharge Cycles. Given time-
series data from the first to the Nth Discharge Cycle, we will predict the SOH and
Capacity of the N+1 Discharge Cycle, as well as the number of Discharge Cycles that
can be used thereafter8.

First, we plan to proceed with the prediction simulation through a simple linear
regression model. Linear regression is a regression analysis technique that models
the linear correlation between the dependent variable and one or more independent
variables.

Figure 1 Figure 1 shows a graph representing the capacity and failure threshold of
a lithium-ion battery. Based on this, we judge that the actual data is roughly
linear and intend to predict SOH and Capacity using linear regression.

The data we have is a record of lithium-ion battery charge and discharge time-
series data measurements. Considering the characteristics of the data, we plan to
use RNN (Recurrent Neural Network), a deep learning model known for utilizing the
temporal attributes of time-series data9. Basic neural network structures only
moved values past the activation function to the output layer from the hidden
layer. However, RNN has the characteristic of sending the results from the
activation function in the hidden nodes not only towards the output layer but also
back into the input for the next calculation in the hidden nodes. Simply put, at
each point in time, it uses the value from the immediate previous point as its
input in a recursive activity. We aim to understand the time-series characteristics
of the data through a model structure that allows past information to continuously
influence future predictions. RNN has the disadvantage of struggling to carry on
learning information when there is a long distance between the related information
and the point of its use. Since predicting the last point of data may require
information from the distant past, we also plan to utilize LSTM (Long Short-Term
Memory), which has the ability to perform learning requiring a long dependency
period10. LSTM learns what information to discard and carry on through various gate
structures and stores this in the cell state. Moreover, it determines the output
based on this cell state.

Using linear regression, RNN, and LSTM mentioned above, we will predict SOH and
Capacity and analyze the results to see what factors can be considered important
characteristics for predicting battery lifespan.

You might also like