The Implementation of Hybrid Electric Vehicle Battery Fault and Abnormal Early Warning System Using Keras Neural Network Technology

This paper proposed a method that makes use of Keras artificial neural network (ANN) technology to develop a power battery failure and abnormal warning system for hybrid electric vehicles. The proposed method applies the on-board diagnostic (OBD) interface to collect driving data of hybrid electric vehicles, such as vehicle speed, engine speed, engine cooling water temperature, engine load, accelerator pedal position, control module voltage, main battery voltage as well as current, hybrid vehicle charging status, intake air temperature, airflow, and etc. Then, these data are preprocessed to extract 6 characteristic fields. In the meanwhile, the airflow and vehicle speed data are used to calculate the fuel consumption characteristic field. All these characteristic fields are normalized to between 0 and 1. Finally, a fault warning model for the power battery system of a hybrid electric vehicle is constructed through the Keras ANN. The input layer, hidden layer, and output layer of the Keras ANN used in this paper are 7, 6, and 10 neurons, respectively. This paper uses the 2019 Toyota Prius C as an experimental vehicle. The experimental results show that the proposed power battery failure and abnormal warning system have an accuracy rate of 97.83%. The experimental results also show that the performance of the Keras ANN model is better than that of the decision tree algorithm, random forest tree algorithm, support vector machine, and k-nearest neighbor classification algorithms. The results of this paper have an extremely high application value of the Internet of Vehicles (IoV), especially for the fault detection/warning of hybrid electric vehicles.