ESTIMASI REMAINING USEFUL LIFE PADA BATERAI ION LITIUM MENGGUNAKAN DEEP DETERMINISTIC POLICY GRADIENT

Wibowo, Chelsa Rachel (2024) ESTIMASI REMAINING USEFUL LIFE PADA BATERAI ION LITIUM MENGGUNAKAN DEEP DETERMINISTIC POLICY GRADIENT. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5002201121-undergraduate_thesis.pdf - Accepted Version Restricted to Repository staff only Download (7MB) | Request a copy

Abstract

Di era green technology, baterai lithium ion (Li-Ion) menjadi komponen penting sebagai sumber energi listrik yang ramah lingkangan karena tidak menghasilkan emisi gas karbon. Permasalahan utama dalam maintenance baterai Li-Ion adalah estimasi remaining useful life (RUL) yang akurat untuk meningkatkan keandalan sistem dan meminimalisir biaya maintenance. Dalam beberapa tahun terakhir, banyak penelitian telah dilakukan untuk estimasi RUL dengan menggunakan metode berbasis machine learning. Namun masih sangat sedikit penelitian estimasi RUL yang menggunakan metode berbasis Deep Reinforcement Learning (DRL). Selain itu, penelitian yang menggunakan DRL sebelumya masih memiliki permasalahan dalam menangani continuous action space. Oleh karena itu, pada penelitian ini digunakan metode berbasis DRL yaitu Deep Deterministic Policy Gradient (DDPG) yang secara khusus dirancang untuk menangani continuous action guna menemukan policy estimasi RUL yang optimal. Penelitian ini menggunakan dataset baterai lithium ion NASA yang dijalankan melalui operasional discharge, dengan kapasitas sebagai data yang diprediksi dan cycle sebagai satuan waktunya. Hasil penelitian menunjukan bahwa DDPG layak digunakan untuk kasus estimasi RUL baterai berdasarkan nilai RMSE, MAE, dan RULerror yang didapat. Sehingga, hasil penelitian dapat membantu perencanaan maintenance agar lebih efektif.

In the era of green technology, lithium ion (Li-Ion) batteries are an important component as an environmentally friendly source of electrical energy because they do not produce carbon gas emissions. The main problem in Li-Ion battery maintenance is accurate remaining useful life (RUL) estimation to improve system reliability and minimize maintenance costs. In recent years, many studies have been conducted for RUL estimation using machine learning-based methods. However, there are still very few RUL estimation studies that use Deep Reinforcement Learning (DRL)-based methods. In addition, previous research using DRL still has problems in handling continuous action space. Therefore, this research uses a DRL-based method, Deep Deterministic Policy Gradient (DDPG), which is specifically designed to handle continuous action to find the optimal RUL estimation policy. This research uses a NASA lithium ion battery dataset that runs through operational discharge, with capacity as the predicted data and cycle as the unit of time. The results show that DDPG is feasible to use for the case of battery RUL estimation based on the RMSE, MAE, and RULerror values obtained. Thus, the research results can help maintenance planning to be more effective.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Baterai Lithium Ion, Deep Deterministic Policy Pradients, Deep Reinforcement Learning, Sisa Masa Pakai Deep Deterministic Policy Pradients, Deep Reinforcement Learning, Lithium Ion Batteries, Remaining Useful Life
Subjects:	Q Science Q Science > QA Mathematics Q Science > QA Mathematics > QA274.7 Markov processes--Mathematical models. Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Mathematics > 44201-(S1) Undergraduate Thesis
Depositing User:	Chelsa Rachel Wibowo
Date Deposited:	08 Aug 2024 01:19
Last Modified:	08 Aug 2024 01:19
URI:	http://repository.its.ac.id/id/eprint/113772

Actions (login required)

View Item