Recognition of driving conditions on urban roads for electric vehicles based on intelligent hybrid search algorithm and LVQNN

This paper proposes an optimized driving condition recognition model based on the learning vector quantized neural network (LVQNN) for electric vehicles (EVs) driving on urban roads. After analyzing the actual driving data of EVs, four different categories of urban road driving conditions were proposed as the expected results of the recognition model. The intelligent hybrid search algorithm combining the particle swarm algorithm and the tabu search algorithm has been developed to select an optimal subset of 11 driving condition feature parameters. The subset describes the short travel segment features that constitute each driving condition, which are used as inputs to the recognition model. After training and testing, the overall recognition accuracy of the model can reach 92.1%. On this basis, the competing layer’s number of neurons of LVQNN is optimized by K-fold cross-validation method to reduce the adverse effects of unreasonable numbers and improve its recognition rate. Through simulation calculations, the recognition accuracy of the optimized model for the four EV urban road driving conditions is higher than that of the unoptimized model, and the total recognition accuracy reaches 93.6%, which indicates that the proposed model can be effectively applied to the recognition of urban road driving conditions of EVs.