Raftonado Situmorang, Muhammad Ridho Dewanto, Barokatun Hasanah, Kholiq Deliasgarin, Bagus Gilang Oktafian

State Of Charge Estimation on Lithium ION Batteries Using Quantum Neural Network

Introduction

State of charge estimation on lithium ion batteries using quantum neural network. Estimate Lithium-ion battery State of Charge (SoC) using Quantum Neural Networks (QNN) for electric vehicles & renewable energy. Real-time Python prototype with performance analysis.

Abstract

Battery applications can be found in electric vehicles, renewable energy power plants and various other portable devices. In this final project research, the author uses the Quantum Neural Network (QNN) method to estimate the State of Charge (SoC) on a lithium-ion battery designed using PYTHON. This research includes the design of a prototype SoC estimation system on lithium-ion batteries using the QNN method, real-time SoC data collection, and comparison of SoC estimation performance using QNN with real-time data. The results of real-time testing of lithium-ion batteries using ACS712 voltage and current sensors for five cycles show the following voltage results: first cycle 10.70 V to 12.68 V, second cycle 10.56 V to 12.66 V, third cycle 10.60 V to 12.69 V, fourth cycle 10.60 V to 12.00 V, and the fifth cycle 10.41 V to 12.07 V. Meanwhile, the current sensor results for five cycles showed a range of 0.1 A to 0.5 A. Each test result per cycle showed a higher increase, although there were small fluctuations, and the overall trend line showed the consistency of the voltage sensor's performance without significant degradation during repeated tests, indicating good stability of the voltage sensor. Then, methods with qubit rotation, linear entanglement, and Neural Network were tested. SoC prediction results using QNN with qubit rotation showed MAPE and RMSE values of 0.14 and 61%, respectively. Furthermore, testing the SoC prediction results on QNN with linear entanglement shows MAPE and RMSE values of 0.08 and 29%, respectively. While the SoC prediction results.

Review

This research explores the application of Quantum Neural Networks (QNN) for estimating the State of Charge (SoC) in lithium-ion batteries, a highly pertinent topic given the widespread use of such batteries in electric vehicles and renewable energy systems. The authors propose a prototype SoC estimation system developed in Python, focusing on real-time data collection and performance comparison of the QNN method against actual battery behavior. The methodology involves real-time testing with an ACS712 sensor over five charge/discharge cycles, followed by the application of different QNN configurations for SoC prediction. The study details the experimental setup, providing voltage and current ranges observed over the five cycles, and notes the stability of the voltage sensor during these repeated tests. A key contribution is the investigation into various QNN approaches, specifically QNN with qubit rotation and QNN with linear entanglement. The results indicate that the linear entanglement QNN configuration performs better, achieving Mean Absolute Percentage Error (MAPE) and Root Mean Square Error (RMSE) values of 0.08 and 29%, respectively, which is a notable improvement over the qubit rotation method (MAPE 0.14, RMSE 61%). The attempt to leverage quantum computing principles for a critical battery management task is an interesting and timely endeavor. While the abstract presents an intriguing application of QNN, several critical aspects require clarification and deeper analysis. The reported RMSE values, particularly 61% and even 29%, are considerably high for a practical SoC estimation system, raising concerns about the model's accuracy and reliability for real-world deployment. The abstract abruptly ends before fully presenting the results for the linear entanglement method, which is a significant oversight. For a comprehensive evaluation, the research would benefit from a direct comparison of the QNN's performance against established classical SoC estimation methods (e.g., Extended Kalman Filter, Neural Networks, Support Vector Machines) to benchmark its effectiveness and computational advantages. Further details on the specific battery chemistry, temperature conditions, and a more in-depth discussion of the QNN architecture and its training process would enhance the paper's scientific rigor and practical implications.

Full Text

Comments

You need to be logged in to post a comment.