Perancangan Sistem Deteksi Kesalahan Pada Proton Exchange Membrane (PEM) Electrolyzer Menggunakan Extended Kalman Filter

Alifah, Rizki Nur (2026) Perancangan Sistem Deteksi Kesalahan Pada Proton Exchange Membrane (PEM) Electrolyzer Menggunakan Extended Kalman Filter. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5009221051-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (2MB) | Request a copy

Abstract

Penelitian ini membahas mengenai perancangan sistem deteksi kesalahan pada PEM Electrolyzer menggunakan pendekatan Extended Kalman Filter (EKF) dan K-Nearest Neighbor (KNN). Permasalahan utama dalam sistem PEM Electrolyzer adalah adanya potensi gangguan pada aspek elektrokimia dan transport massa yang dapat mempengaruhi kinerja sistem, khususnya pada kondisi operasi dinamis. Oleh karena itu, diperlukan metode deteksi kesalahan yang mampu mengidentifikasi perubahan kondisi sistem secara cepat dan akurat. Pada penelitian ini digunakan dua jenis fault, yaitu membrane conductivity degradation yang dimodelkan sebagai peningkatan resistansi ohmik, serta anode gas transport limitation yang dimodelkan melalui penurunan limiting current. EKF digunakan untuk mengestimasi state sistem berupa alpha (α) dan beta (β) serta tegangan estimasi sebagai representasi kondisi sistem aktual. Selisih antara tegangan hasil pengukuran dan hasil estimasi digunakan sebagai residual untuk mendeteksi adanya penyimpangan sistem. Selanjutnya, dilakukan ekstraksi fitur statistik berupa peak, RMS, dan mean dari alpha, beta, dan residual sehingga diperoleh sembilan fitur yang digunakan sebagai input KNN classifier. Hasil simulasi menunjukkan bahwa EKF mampu mengestimasi kondisi sistem dengan baik dengan rata-rata NRMSE sebesar 0,047795371%. Sistem deteksi kesalahan yang dikembangkan mampu mengklasifikasi kondisi normal, fault 1 (membrane conductivity degradation), dan fault 2 (anode gas transport limitation) dengan error total sebesar 19,58%. Namun, performa pada fault anode gas transport limitation masih relatif rendah akibat karakteristik sistem yang lebih nonlinear. Dengan hasil ini, integrasi EKF dan KNN terbukti mampu digunakan sebagai pendekatan hybrid untuk deteksi dan klasifikasi kesalahan pada PEM Electrolyzer berbasis model simulasi.
===================================================================================================================================
This study discusses the design of a fault detection system for a PEM Electrolyzer using the Extended Kalman Filter (EKF) and K-Nearest Neighbor (KNN) approaches. The main issue in PEM Electrolyzer systems is the potential occurrence of faults in electrochemical processes and mass transport, which can affect system performance, particularly under dynamic operating conditions. Therefore, a fault detection method is required to identify changes in system conditions accurately and promptly. In this study, two types of faults are considered, namely membrane conductivity degradation, which is modeled as an increase in ohmic resistance, and anode gas transport limitation, which is modeled through a decrease in limiting current. The EKF is used to estimate the system states, namely alpha (α) and beta (β), as well as the estimated voltage as a representation of the actual system condition. The difference between the measured voltage and the estimated voltage is used as the residual to detect system deviations. Furthermore, statistical feature extraction is performed using peak, RMS, and mean values from alpha, beta, and residual signals, resulting in nine features used as inputs for the KNN classifier. The simulation results show that EKF can estimate the system condition effectively, with an average NRMSE of 0,047795371%. The developed fault detection system can classify normal conditions, fault 1 (membrane conductivity degradation), and fault 2 (anode gas transport limitation) with a total error of 19,58%. However, the performance for anode gas transport limitation remains relatively low due to the more nonlinear characteristics of the system. Based on these results, the integration of EKF and KNN is shown to be applicable as a hybrid approach for fault detection and classification in a simulation-based PEM Electrolyzer model.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	EKF, Fault detection, PEM Electrolyzer
Subjects:	Q Science > QD Chemistry > QD553 Electrochemistry. Electrolysis T Technology > TA Engineering (General). Civil engineering (General) > TA169.6 Fault location (Engineering)
Divisions:	Faculty of Industrial Technology > Physics Engineering > 30201-(S1) Undergraduate Thesis
Depositing User:	Rizki Nur Alifah
Date Deposited:	30 Jul 2026 06:49
Last Modified:	30 Jul 2026 06:50
URI:	http://repository.its.ac.id/id/eprint/139627

Actions (login required)

View Item