Prediction of Time-Series Discharge Characteristics of Primary Batteries for IoT Device Using Machine Learning

EasyChair Preprint no. 13477

Abstract

Efforts are made to construct a prediction model for the discharge characteristics of IoT device batteries using time-series prediction models based on Transformers. With the aim of calculating SOC (State of Charge) through recursive prediction, conditions and processes conducive to more accurate recursive prediction were investigated. Currently, recursive prediction with iTransformer tend to be good score. It was also suggested that adding noise during recursive prediction may enable more stable long-term predictions.

Keyphrases: IoT device, iTransformer, Non-stationary Transformers, PatchTST, primary battery, SoC, transformer

@Booklet{EasyChair:13477,
  author = {Shuhei Matsushita and Masahiro Fukui and Kiyotsugu Takaba},
  title = {Prediction of Time-Series Discharge Characteristics of Primary Batteries for IoT Device Using Machine Learning},
  howpublished = {EasyChair Preprint no. 13477},

  year = {EasyChair, 2024}}