An Analysis of Electric Vehicle Driving Pattern Based on a Hybrid Model with the Growing Hierarchical Self-Organizing Maps and Deep Recurrent Neural Networks

• Main Authors: CHOU, CHIEN-CHENG, 周建成
• Other Authors: LEE, CHUNG-HONG
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/68714222415939225769

碩士 === 國立高雄應用科技大學 === 電機工程系博碩士班 === 105 === In recent years, the countries of the world actively develop green energy technology because of global climate change and energy shortages. The electric vehicle (EV) is part of green energy technology. Electric vehicles have the merits of high energy efficiency and non-carbon emissions, and these merits of electric vehicles can facilitate the urban afforestation. Therefore, the countries of the world are committed to promoting electric vehicles. In order to speed up the popularity of electric vehicles, this study is committed to analyzing the driving behaviors of the electric vehicle drivers and solve the problems of drivers by this study. The author collected the electric vehicle data of the electric vehicle fleet in this study. Then, the author collated the driver's energy consumption habits by data preprocessing as the sample datasets of the experiment. The author used the hybrid model to analyze the data. The model combined with the Growing Hierarchical Self-Organizing Map (GHSOM) and the Deep Recurrent Neural Network (DRNN). In the first half of the experiment, the author made the cluster experiment first and then conducted the classification experiment. The author defined the clustered results as classification labels by the GHSOM cluster experiment, and the sample datasets are divided into training samples and testing samples. In the second half of the experiment, the author used the training samples to train the DRNN model and input the testing samples when the learning was completed. Finally, the average recognition rate of RNN model is more than 90%, and the author find out an ideal neural network parameter settings.