| Summary: | 碩士 === 國立臺灣科技大學 === 電機工程系 === 102 === Lithium ion (Li-ion) batteries play an important role in applications such as portable electronic, electrical vehicles and renewable energy systems. To maximize the performance of the Li-ion batteries, an accurate and fast battery state of charge (SOC) estimation technique is essential. In this thesis, an AC impedance based technique is proposed to estimate the SOC of Li-ion batteries. The proposed technique includes an artificial neural network (ANN) which can be used to calculate the SOC according to the AC impedance values, battery current and battery temperature data. In this thesis, these input data is obtained using the Bio-Logic VMP3 potentiostats/galvanostats device, and the ANN is trained using the Neural Network Toolbox in MATLAB.
According to the trained results, the performance of the utilized ANN is related to the number of neurons in the hidden layer. The averages SOC estimation error is lower than 6 %, 5% and 2% when the number of neurons in the hidden layer is set as 15, 25 and 35, respectively. Therefore, the proposed technique can be utilized to estimate the SOC of Li-ion battery precisely. Finally, the trained ANN is implemented using a low cost microcontroller dsPIC33FJ16GS502 from Microchip to validate the correctness of the proposed method.
|
|---|
