A Fast Diagnosis for Classification of re-used Li-ion Batteries for PV and EV Systems by the ANN Model

Abstract

Proper usage of the batteries can impact how long the battery in PV/EV systems will last. But the correct estimation of State of Health (SoH) can affect the total cost of the system and its efficiency. As a matter of fact, the battery cost in EV applications is (35–50) % of the total cost of the cars. Their classification of deterioration and which application to send them next is the main concern. In this paper the proposed method was based on ANN algorithm, expressed by two NN structures in cascade. Where the first NN structure use V and I waveform and number of cycles as an optional input, and the output is the internal impedance parameters which is used as main input for the second NN in order to estimate finally the SoH of the battery pack system. A structure with 1 and 2 hidden layers is proposed. The estimation is finished within 42 seconds and with error of 1.8% in the worst case. By correctly estimating the SoH of the battery we can extend its usage for a little longer or preparing it to be used in PV systems, where the need for high current and dynamic characteristics during discharging it's not the same as in EV.