利用遞歸神經網路於無線充電電池管理

本文利用遞迴歸類神經網路實現無線充電電池管理。隨著電動車科技的發展，使得鉛酸電池與鋰電池廣泛的使用與電動車產業中，顯現出充電電池管理的重要性。本論文將發展鉛酸電池與鋰電池充電管理系統，以電池製造商的資料規格與兩段定電壓-電流充電法建立充電系統模型，目的是為建立類神經網路訓練所需要資料庫，在充電管理系統將利用量測電池充電電壓與充電電流，並透過遞迴歸神經網路來估測電池的開路電壓與瞬時效率。本文除了建立鉛酸電池充電系統模型之外，還透過實驗來驗證遞迴類神經網路對電池充電狀態的估測，在微控制器上處理充電資訊以及演算充電策略找出最佳充電曲線，期望能有效地控制電池充電。

關鍵字

遞迴神經網路；無線充電；推挽式轉換器

並列摘要

In this paper, utilize the recurrent neural network to implement wireless charging management of battery. As the development of electric vehicle technology lead to lead-acid battery and lithium battery are used widely in electric vehicle industry. Obviously it is important for battery charging management. This paper will develop a lead-acid battery charging management system. Establish a charging system model based on the battery manufacturer's data specifications and a two-stage constant voltage-current charging method. The purpose is to establish a database for the neural network training needs in charge management. The system will measure the battery charging voltage and current and estimate the open circuit voltage and instantaneous efficiency of the battery by recurrent neural network. In addition to establishing a lead-acid battery charging system model, this paper observes whether the state of battery charging can be estimated by using the recurrent neural network. It is expected to do the charging information and calculating charging strategy to find the optimized charging curve on the microcontroller. Expect to effectively control battery charging.