| Electric buses are an important component of low-carbon transportation.However,due to the limitation of battery capacity,the range of limited electric buses has become a major factor affecting the large-scale popularization of electric buses.Studies have shown that reasonable and accurate estimates of energy consumption can also greatly reduce the impact of limited driving range of electric vehicles when battery technology is difficult to break through in a short period of time.At present,most electric vehicle energy consumption estimates are concerned about the deterministic estimation of travel energy consumption.The impact of uncertain factors in the actual driving factors on energy consumption has not been fully valued.To solve this problem,two probability models based on Bayesian regression and quantile regression are proposed respectively to estimate the probability distribution of the travel energy consumption of electric buses,and the estimation performance of the probability distribution of the proposed models is discussed in detail.The main research work includes:Based on Bayesian regression and quantile regression,the probability estimation models of electric bus energy consumption were constructed respectively.Considering the influence of speed,braking,temperature and other factors,the basic structure of electric bus energy consumption model was determined based on four physically explicable characteristics,namely speed intensity,braking intensity,slow running intensity and temperature intensity.In the probability estimation model based on Bayesian regression,the normal distribution is introduced into the energy consumption model to describe the uncertainty of energy consumption,the method of estimating the prior distribution of the model is designed,and the method of solving the posterior probability of the model is studied;In the probability estimation model based on quantile regression,the discrete quantile distribution is introduced into the energy consumption model to describe the uncertainty of energy consumption,and the solution method of the model is studied.Based on the actual driving data of electric buses,the two energy consumption probability estimation methods are applied.On the basis of correcting the information loss and abnormal sampling point in the original data,the stroke power was corrected using estimated battery attenuation combined with voltage,current,and SOC data;The two proposed probability estimation models were trained and validated using the time series validation,and the deterministic estimation performance and probabilistic estimation performance of the two probability estimation models were compared and evaluated respectively.At the same time,the impact of temperature on the uncertainty of electric bus energy consumption was analyzed and discussed.The rationality of the estimation results of model probability distribution is studied.This paper proposes a method to separate the uncertainty distribution of travel energy consumption by directly utilizing the energy consumption difference of adjacent trips,and then applies this method to the training verification data in Chapter3 to obtain the uncertainty distribution of travel energy consumption with clearer physical meaning.This result is compared with the probabilistic model estimation results proposed in this article,further verifying the effectiveness of the model proposed in this article.At the same time,the main components of energy consumption estimation errors are discussed in depth,providing a useful reference for more reasonable application and evaluation of electric vehicle energy consumption models. |
