Study on the Driving Cycle Recognition Method for Hybrid Electric Vehicle Based on Driving Characteristics

Focusing on the adaptability of energy management methods for hybrid electric vehicle (HEV) to the driving cycle, the paper analyses and compares two driving cycle identification methods using two different algorithms based on driving characteristics. This paper selects 10 typical driving cycles and divided them into 104 kinematic segments and calculates 12 driving characteristic parameters of each kinematic segment to form a data matrix. Based on K-means cluster principle, the data matrix is clustered into three categories corresponding to three common driving cycles, namely, freeway driving cycle, suburban driving cycle, and urban driving cycle. What's more, randomly generate 3000 sets of data within the threshold of the characteristic parameters of each category to be used as learning data, which will be inputted into the classification learners. Then two classification learners including the K-nearest neighbor classification (KNN) algorithm and the feedforward neural network (FNN) algorithm are used to learn and verify with the original data as the test data. Regarding the overall recognition results of the three types of driving cycles, the FNN algorithm outperforms the KNN algorithm. This research realizes the effective recognition of driving cycles and provides new ideas and methods for the application of different algorithms in driving cycle recognition, aiming to enhance energy management in hybrid electric vehicles.