This is a deep learning approach to estimate State-of-Charge of 18650 Li-Ion batteries in real-time with high accuracy.
The dataset, used can be found here, Dataset.
Clone this repo into your working directory and execute the training_code.m file to train an artifical neural network.
You can change the network hyper-parameters to improve training results.
Once the training is complete, you can export the model into various formats as per your use case through builtin matlab commands. In this case, it has been exported as a Tensorflow model and later converted to TFLite format to be deployed on Hardware.
Matlab Code files have been written in Matlab 2020b and all python files have been verified to work in Python 3.9
The Schematic and board files for the PCB HAT designed for Raspberry Pi can be found in the PCB folder. These were designed in Eagle 9.6.2
1) Raspberry Pi 4
2) 3Ah 18650 Li-Ion cell
3) 0-25V Generic Voltage sensor
4) 0-30A ACS712 Current Sensor
5) DHT11 Temperature Humidity Sensor
6) ADS1115 / MCP3208 (ADC)