Research On Energy Management For Vehicle Hybrid Power System Of Lithium Battery And Super-Capacitor

Lithium battery and super-capacitor have their own characteristics and advantages.For a hybrid power system,how to study the dynamic energy management between lithium battery and super-capacitor is the key to the performance of electric vehicle.And accurate estimation of state of health(SOH)can ensure the safety and reliability of lithium battery.The main contributions of this paper are as follows:(1)This paper proposes a method for estimating the SOH of lithium battery,based on convolutional neural network(CNN)and transfer learning.CNN is used to automatically extract features from voltage,current,and capacity data,and accelerated aging data and normal speed aging data are combined by transfer learning for training.Accelerated aging data and the last several cycles of discarded battery are used to pre-train a base model,and the first several cycles of normal speed aging data are used to fine-tune the model.Then the learned model can accurately estimates the SOH of the lithium battery.(2)This paper proposes an energy management strategy based on Deep Q-network algorithm.According to the characteristics of lithium battery and super-capacitor,appropriate variables are selected as the state and action,and a reward function is designed.Simulation experiments are carried out under the three driving conditions of ECE,UDDS,and HWFET,and the proposed energy management strategy is used to distribute the energy between lithium battery and super-capacitor.The proposed method can meet the demand power.The super-capacitor can absorb the energy released by the vehicle during braking and avoids energy waste.The super-capacitor assists to reduce the use of lithium battery.The output current variance rate of the lithium battery is reduced,so the lifespan of the lithium battery can be prolonged.(3)In addition,neural network and reinforcement learning method is utilized for energy management strategy.Based on the predicted value of the next state from neural network,the optimal action sequence of a certain length is selected according to the principle of maximum cumulative reward.Based on the model predictive control method,the first action of the action sequence is selected as the optimal control action under the current state.The proposed energy management strategy is adopted in a hybrid driving mode composed of four driving conditions of MBDC,UDDS,WVUSUB and HWFET,to realize the energy distribution between lithium battery and super-capacitor.