Samudra, Bima Augusta Sultan (2026) Klasifikasi Kondisi Operasi Baterai Go-Kart Listrik Berdasarkan Parameter Suhu Dan Arus Menggunakan Algoritma K-Nearest Neighbor. Other thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 2040221048-Undergraduate_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
2040221048-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (5MB) | Request a copy |
Abstract
Perkembangan kendaraan listrik meningkatkan kebutuhan akan pemantauan kondisi baterai yang mampu mendukung keamanan, keandalan, dan kinerja kendaraan selama beroperasi. Pada gokart listrik, perubahan beban operasi dapat menyebabkan fluktuasi arus dan peningkatan suhu baterai yang perlu dipantau untuk mengidentifikasi kondisi operasinya. Proyek akhir ini bertujuan mengklasifikasikan kondisi operasi baterai LiFePO4 pada gokart listrik berdasarkan parameter arus dan suhu menggunakan algoritma K-Nearest Neighbor (KNN), serta mengevaluasi kinerja model dalam mengenali setiap kategori kondisi operasi. Penelitian dilakukan dengan pendekatan eksperimental melalui akuisisi data menggunakan ANT Battery Management System (BMS) dan kontroler Votol EM-150 pada beberapa variasi kecepatan operasi gokart. Data hasil akuisisi diproses melalui tahapan preprocessing, meliputi data cleaning, pemeriksaan missing value, penghapusan data duplikat, analisis outlier, penentuan threshold, pelabelan, dan normalisasi menggunakan Min-Max Scaling. Kondisi operasi baterai dikategorikan menjadi Normal, Sedang, Tinggi, dan Kritis, kemudian parameter arus dan suhu baterai digunakan sebagai fitur masukan model KNN dengan pembagian data training dan testing sebesar 80:20. Hasil pengujian menunjukkan bahwa K = 3 merupakan nilai terbaik dengan akurasi sebesar 99,60%. Model menghasilkan F1-score sebesar 1,00 pada kelas Normal, 1,00 pada kelas Sedang, 0,99 pada kelas Tinggi, dan 0,91 pada kelas Kritis. Hasil tersebut menunjukkan bahwa KNN mampu mengenali pola kondisi operasi baterai berdasarkan kombinasi parameter arus dan suhu dengan tingkat ketepatan yang tinggi. Penelitian ini memberikan dasar penerapan klasifikasi berbasis machine learning untuk mendukung pemantauan kondisi operasi baterai pada gokart listrik. Penelitian selanjutnya dapat memperluas jumlah dan variasi dataset serta menambahkan parameter baterai lainnya untuk meningkatkan kemampuan generalisasi model.
=========================================================================================================================================
The development of electric vehicles has increased the need for battery condition monitoring to support vehicle safety, reliability, and performance during operation. In electric go-karts, variations in operating load can cause fluctuations in battery current and temperature, which need to be monitored to identify battery operating conditions. This final project aims to classify the operating conditions of a LiFePO4 battery in an electric go-kart based on current and temperature parameters using the K-Nearest Neighbor (KNN) algorithm and to evaluate the model performance in identifying each operating condition category. An experimental approach was conducted by acquiring data using an ANT Battery Management System (BMS) and a Votol EM-150 controller under various go-kart operating speeds. The acquired data were processed through preprocessing stages, including data cleaning, missing value checking, duplicate removal, outlier analysis, threshold determination, labeling, and Min-Max Scaling normalization. Battery operating conditions were classified into four categories: Normal, Moderate, High, and Critical. Battery current and temperature were then selected as input features for the KNN model using an 80:20 training-testing data split. The results showed that K = 3 achieved the best performance with an accuracy of 99.60%. The model achieved F1-scores of 1.00 for Normal, 1.00 for Moderate, 0.99 for High, and 0.91 for Critical. These results indicate that KNN can effectively identify battery operating conditions based on the combination of current and temperature parameters with a high level of accuracy. This study provides a basis for applying machine learning-based classification to support battery operating condition monitoring in electric go-karts. Future research should expand the dataset and operating variations and incorporate additional battery parameters to improve model generalization.
Actions (login required)
 |
View Item |