The analysis of statistic data from real-time driving cycle of hybrid electric vehicle shows relationships that help to optimize control strategy. This paper presents a remote self-learning platform based on general packer radio service (GPRS) for control strategy optimization to hybrid electric vehicle. The platform adopts an in-vehicle device to acquire real-time driving data through CAN bus and communicate wirelessly with central server by GPRS network and INTERNET. A fast clustering method is introduced to classify real-time driving data as different driving cycle. According to driving cycle, online statistical analysis is implemented in central server to obtain energy consumption result, and the optimized control strategy is updated by in-vehicle device. During the process of update to control strategy, the wireless communication quality is a vital factor for completeness of data. A Model-based control algorithm to solve wireless network congestion problem is realized depending on the characteristics of the communication quality. The test result shows that the communication process can be improved greatly and the communication quality is increased 30% at least by this algorithm. The transmission load would be reduced automatically in order to make the communication be normal as soon as possible when the network is under congestion.
The study shows that the remote self-learning platform is effective to define and characterize driving cycle of hybrid electric vehicle. In the sample application, remote self-learning platform plays an importance role in optimizing the control and energy management strategy of hybrid electric vehicle.