Perancangan Estrimator State Of Charge (SOC) Untuk Baterai NMC Berbasis Adaptive Extended Kalman Filter Pada Daya Lepas Tinggi

Susanto, Rokan Alwan (2026) Perancangan Estrimator State Of Charge (SOC) Untuk Baterai NMC Berbasis Adaptive Extended Kalman Filter Pada Daya Lepas Tinggi. Diploma thesis, Institut Teknologi Sepuluh Nopember.

Text 5009221004_Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Baterai Nickel Manganese Cobalt (NMC) merupakan salah satu jenis baterai lithium-ion yang banyak digunakan pada kendaraan listrik karena memiliki kapasitas tinggi dan kemampuan menangani beban tinggi. Pada kondisi pelepasan daya tinggi (high discharge) dengan C-rate hingga 7C, dinamika internal baterai berubah secara cepat sehingga estimasi State of Charge (SOC) yang akurat menjadi sangat penting untuk menjaga keamanan dan efisiensi operasi baterai. Metode estimasi berbasis filter seperti Extended Kalman Filter (EKF) menghadapi keterbatasan pada kondisi ini karena matriks kovarians noise proses (Q) dan pengukuran (R) yang bersifat tetap tidak lagi representatif terhadap kondisi operasi yang berubah secara dinamis. Penelitian ini merancang dan mengevaluasi estimator SOC berbasis Adaptive Extended Kalman Filter (AEKF) yang mampu memperbarui Q dan R secara online menggunakan mekanisme adaptasi berbasis moving window, dikombinasikan dengan Combined Model elektrokimia hasil gabungan model Shepherd, Unnewehr Universal, dan Nernst sebagai model baterai. Parameter Combined Model diidentifikasi secara offline menggunakan metode Ordinary Least Squares (OLS) dari data eksperimen baterai Molicel INR-21700-P50B (5Ah, NMC) pada kondisi 1C hingga 7C dan profil dinamis WLTP dengan periode sampling 1 detik. Pengujian variasi parameter AEKF menunjukkan bahwa ukuran moving window M=1000, nilai awal kovarians noise proses Q₀=1×10⁻³, dan nilai awal kovarians noise pengukuran R₀=1×10⁻⁵ menghasilkan performa optimal. Hasil validasi model baterai pada kondisi 1C, 3C, 5C, dan 7C menunjukkan MAPE tegangan antara 0,36% hingga 0,85% ketika parameter diuji pada sel baterai yang sama (in-sample). MAPE meningkat menjadi 1,14% hingga 2,33% ketika parameter yang sama diuji pada unit sel baterai berbeda (out-of-sample), namun tetap berada di bawah ambang toleransi 5%, yang menunjukkan kemampuan generalisasi Combined Model antar unit sel dengan spesifikasi serupa. Performa estimator AEKF pada parameter optimal menghasilkan RMSE SOC antara 3,21% hingga 3,80% pada seluruh kondisi 1C hingga 7C dan WLTP, seluruhnya di bawah batas toleransi 5% yang ditetapkan, sehingga estimator dinyatakan berhasil dan layak digunakan untuk aplikasi manajemen baterai pada kendaraan listrik yang mendukung SDG 7 (Affordable and Clean Energy), SDG 9 (Industry, Innovation and Infrastructure), dan SDG 12 (Responsible Consumption and Production)
=================================================================================================================================
Nickel Manganese Cobalt (NMC) battery is a lithium-ion battery widely used in electric vehicles due to its high capacity and high-load handling capability. Under high discharge conditions with C-rates up to 7C, the internal dynamics of the battery change rapidly, making accurate State of Charge (SOC) estimation critical for maintaining battery safety and operational efficiency. Filter-based estimation methods such as the Extended Kalman Filter (EKF) face limitations under these conditions because the fixed process noise (Q) and measurement noise (R) covariance matrices are no longer representative of dynamically changing operating conditions. This research designs and evaluates an SOC estimator based on the Adaptive Extended Kalman Filter (AEKF), capable of updating Q and R online using a moving window-based adaptation mechanism, combined with an electrochemical Combined Model formed from the integration of the Shepherd, Unnewehr Universal, and Nernst models as the battery model. The Combined Model parameters were identified offline using the Ordinary Least Squares (OLS) method from experimental data of a Molicel INR-21700-P50B (5Ah, NMC) battery under 1C to 7C conditions and a dynamic WLTP profile with a sampling period of 1 second. AEKF parameter variation testing showed that a moving window size of M=1000, an initial process noise covariance of Q₀=1×10⁻³, and an initial measurement noise covariance of R₀=1×10⁻⁵ yielded optimal performance. Battery model validation under 1C, 3C, 5C, and 7C conditions showed a voltage MAPE ranging from 0.36% to 0.85% when the parameters were tested on the same battery cell (in-sample). The MAPE increased to 1.14%-2.33% when the same parameters were tested on a different battery cell unit (out-of-sample), yet remained below the 5% tolerance threshold, indicating the Combined Model's generalization capability across battery cell units with similar specifications. The AEKF estimator's performance with the optimal parameters yielded an SOC RMSE ranging from 3.21% to 3.80% across all 1C to 7C conditions and the WLTP profile, all below the established 5% tolerance limit, demonstrating that the estimator is successful and feasible for battery management applications in electric vehicles, supporting SDG 7 (Affordable and Clean Energy), SDG 9 (Industry, Innovation and Infrastructure), and SDG 12 (Responsible Consumption and Production).

Item Type:	Thesis (Diploma)
Uncontrolled Keywords:	Adaptive Extended Kalman Filter, Affordable and clean energy, Baterai NMC, NMC battery, Combined Model, State of Charge.
Subjects:	Q Science Q Science > Q Science (General) > Q180.55.M38 Mathematical models Q Science > QA Mathematics Q Science > QA Mathematics > QA402.3 Kalman filtering. T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK2921 Lithium cells.
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Physics Engineering > 30201-(S1) Undergraduate Thesis
Depositing User:	Rokan Alwan Susanto
Date Deposited:	05 Aug 2026 01:42
Last Modified:	05 Aug 2026 01:42
URI:	http://repository.its.ac.id/id/eprint/142652

Actions (login required)

View Item