| 摘要: | 本研究利用兩種不同的類神經網路來模擬近似鋰電池的放電行為。第一種是利用等前向式類神經網路，並利用均方誤差演算法來減少真實的電池剩餘電能與類神經網路預測值彼此之間的誤差。第二種類神經網路是叫做輻射半徑基底函數網路。這種網路可以將不同的結果分類。利用神經網路預測結果與不同樣式物件的距離，它可以利用與某物件最短的距離來將輸入值分類。這兩種類神經網路都能對剩餘的電池電能得到滿意的預測結果且成功的近似電池放電的行為模式。
In this paper, the approximation of battery discharge behavior by using neural network architecture is studied. The first one uses the feed forward network with least mean square error algorithm to minimize the differences between the true remaining battery power and the predicted value from the neural network. The second network is a radial basis network. The radial basis network is used to classify different battery discharge behaviors. It computes the distance between the battery capacity and the results produced by the network trained. It will decide the shortest distance and then classify the best category it will fit. Both network produce satisfactory predictions for the remaining battery power and can approximated the battery discharge behavior very well. |
