Search

ABSTRAK Penurunan performa baterai terjadi akibat siklus pengisian dan pengosongan berulang yang melebihi batas, mempercepat degradasi. Penelitian ini bertujuan untuk meningkatkan akurasi estimasi State of Charge (SOC) baterai menggunakan Artificial Neural Network (ANN) dengan algoritma Elman Backpropagation. Metode digunakan karena menambahkan lapisan context neuron yang mampu menangkap pola dinamis pada data baterai. Pengujian dilakukan dengan membandingkan hasil estimasi SOC dari metode ini dengan metode Coulomb Counting. SOC baterai diestimasi dari 100% hingga 60%, dan hasil menunjukkan bahwa meskipun Coulomb Counting awalnya memberikan SOC lebih tinggi, estimasi dari kedua metode menjadi lebih mirip seiring waktu. Error estimasi berkisar antara 0,1% hingga 14,7%. Algoritma Elman Backpropagation terbukti mampu memberikan estimasi SOC yang lebih akurat, namun masih memerlukan kalibrasi lebih lanjut untuk meningkatkan presisi. Kata kunci: Artificial Neural Network, Baterai, Coulumb Counting, Elman Backpropagation, State of Charge. ABSTRACT The decline in battery performance occurs due to repeated charge and discharge cycles that exceed limits, accelerating degradation. This study aimed to improve the accuracy of State of Charge (SOC) estimation using an Artificial Neural Network (ANN) with the Elman Backpropagation algorithm. The method used was unique in adding a context neuron layer capable of capturing dynamic patterns in battery data. Testing was conducted by comparing SOC estimates from this method with the Coulomb Counting method. The battery's SOC was estimated from 100% to 60%, and the results showed that although Coulomb Counting initially provided higher SOC estimates, the estimates from both methods became more similar over time. Estimation errors ranged from 0.1% to 14.7%. The Elman Backpropagation algorithm proved to provide more accurate SOC estimates, although further calibration is needed to improve precision. Keywords: Artificial Neural Network, Battery, Coulumb Counting, Elman Backpropagation, State of Charge.

This study focuses on conceptualization and development of a battery management system (BMS) with two main functions, battery monitoring and management, in the context of brushless direct current motors (BLDCs). The main challenge in variable estimation is to protect the battery from potential risks during the charge and discharge cycle. The new proposed resolution combines a comprehensive BMS with monitoring capabilities for charge (SoC), health (SoH), voltage, current and battery temperature. In addition, a protective mechanism is incorporated to prevent variables from overshooting safety parameters. This research uses two different methodologies for estimating SOC, coulomb counting and open circuit voltage. In experimental tests, resistance potentiometers of 1,650, 3,300 and 0 were used, with SoC estimates of 37%, 19% and 65%, while coulomb counting method has a marginal error of 1.13%. On the contrary, the open-circuit voltage method generated a SoC estimate of 0% for all potentiometer resistance, with an error rate of 0.64 %. As a result, the open circuit voltage method is chosen because of its superior accuracy compared to the coulomb counting method. The state assessment of the battery showed a value of 100% after seven cycles. In addition, a protective system has been implemented to ensure that battery variables remain within the safe thresholds throughout the charge and discharge process. Consequently, the implementation of this BMS is expected to significantly improve overall performance and extend battery life.