在眾多二次電池中鋰離子電池是作為電動載具電池較為合適的，而鋰離子電池中又為以磷酸鋰鐵電池更為合適，因其特性工作電壓高、循環壽命高、自放電率低與能量效命高等優點，因此本論文選用磷酸鋰鐵電池作為實驗對象。而鋰鐵電池會因溫度、充放電電流等外在因素影響電池容量，由此得知能夠準確的預估電池殘電量是相當困難的，然而類神經網路的非線性、可變性、多輸入輸出與可容錯等特性使得類神經網路能夠準確地預測電池殘電量，在實驗中利用充、放電測試主機在不同外在條件下取得電池充、放電資料，將電池溫度、放電電流與電池端電壓作為類神經網路之輸入，電容量作為目標，在本研究中使用MATLAB 程式內之類神經網路建立預估電池殘電量之類神經網路。最後利用LabVIEW 圖控軟體設計一套電池電容量運算與電池特性監控程式並結合實驗設備使其能夠調控對電池放電之電流大小，再將放電資料儲存，利用此放電資料輸入至類神經網路電池殘電量估測模組中預估電池殘電量後比較實際電容量與預估電容量誤差，而驗證顯示使用倒傳遞網路中的SCG 演算法擁有較高的精準度其實際殘電量與預估殘電量平均誤差為7%。

Due to the Lithium iron phosphate with many advantages, such as high worked voltage, high cycle life, low discharge and high energy desity, so it appropriately use in electric vehicles. Many variables (ex. temperature, discharge current…ect.) will affect capacitance of battery, so accuracy of the estimated state of charge for batteries is very difficult. The characteristics of neural network are nonlinear, variability and muti-input/output. Hence the neural network could accurately estimate the state of charge for battery. Developing the battery monitor system for obatining battery’s performance date and calculating the capacitance of battery by using LabVIEW software. By a series experiments the estimated SOC of battery is less than 7%.