Research On Battery Life Prediction Method Of Electric Vehicle Based On Real-time Data

In recent years,in the face of environmental pollution and energy overuse caused by fuel gas emissions,the Chinese government has issued a series of policies to encourage people to reduce fuel emissions.Electric bicycles have become an important part of green travel.Electric bicycles have developed rapidly due to their small size,low price,and suitable for short-distance travel.With the number of electric bicycles has increased,they have become the main methods of transportation for people’s daily travel,takeaway services,and bike-sharing.Lithium batteries are the only source of power for lithium electric bicycles,and their quality affects the performance and safety of the vehicle directly.The battery’s endurance is one of the crucial parameters that people concerned about when choosing vehicles.However,in the actual driving process,because of the influence of various factors,its endurance is greatly reduced.There is a problem that electric bicycles have no real-time remaining driving range or large deviations compared with prediction.This subject takes Mavericks UQis as the research object,and its power battery is 9 parallel 13 series lithium batteries with a rated voltage of 48 V.Based on the hardware platform to collect information such as voltage,internal resistance,temperature,speed and SOC during the driving process of the electric vehicle,and analyzed the relationship between each factor and the remaining driving range.Establishing the corresponding regression prediction model to realize the evaluation of the electric vehicle during the driving process in order to forecast of remaining driving range.First,analyze the force situation and power consumption of the electric bicycle during the driving process.By establishing the energy conservation equation,obtain the driving range and the remaining driving range under the constant speed condition.Simultaneously,analyze the factors that affect the driving range during the driving process.Secondly,based on the actual testing requirements,a hardware testing platform that can detect real-time data of electric vehicles is built based on the existing experimental equipment.According to the principle of DC discharge method,parallel discharge loads,observe the changes of terminal voltage and current,and realize the collection of resistance.For the collected data,the abnormal data is handled by analyzing the changes during the operation process.Simultaneously,in order to obtain the remaining driving mileage that meets the accuracy requirements,the speed points are used to obtain the mileage instead of the actual mileage to obtain the remaining driving mileage through calculation.Combining the collected real-time data,perform the correlation analysis among the characteristics of each influencing factor and the remaining driving range to build the foundation for subsequent predictive modeling.Finally,based on real-time data,GRNN and GBDT were used to establish prediction models among battery internal resistance,temperature,total voltage,driving speed,SOC and remaining driving range.At the same time,the PSO optimization algorithm is used to optimize the parameters in the model.The results show that both models can be used to predict the remaining driving range.For the single result,the results from PSO-GRNN model is better,but the stability of PSO-GBDT model is better.