An energy management strategy (EMS) is important for hybrid electric vehicles (HEVs) since it plays a decisive role on the performance of the vehicle. However, the variation of future driving conditions deeply influences the effectiveness of the EMS. Most existing EMS methods simply follow predefined rules that are not adaptive to different driving conditions online. Therefore, it is useful that the EMS can learn from the environment or driving cycle. In this paper, a deep reinforcement learning (DRL)-based EMS is designed such that it can learn to select actions directly from the states without any prediction or predefined rules. Furthermore, a DRL-based online learning architecture is presented. It is significant for applying the DRL algorithm in HEV energy management under different driving conditions. Simulation experiments have been conducted using MATLAB and Advanced Vehicle Simulator (ADVISOR) co-simulation. Experimental results validate the effectiveness of the DRL-based EMS compared with the rule-based EMS in terms of fuel economy. The online learning architecture is also proved to be effective. The proposed method ensures the optimality, as well as real-time applicability, in HEVs.

Regenerative braking is an effective approach for electric vehicles (EVs) to extend their driving range. A fuzzy-logic-based regenerative braking strategy (RBS) integrated with series regenerative braking is developed in this paper to advance the level of energy-savings. From the viewpoint of securing car stability in braking operations, the braking force distribution between the front and rear wheels so as to accord with the ideal distribution curve are considered to prevent vehicles from experiencing wheel lock and slip phenomena during braking. Then, a fuzzy RBS using the driver’s braking force command, vehicle speed, battery SOC, battery temperature are designed to determine the distribution between friction braking force and regenerative braking force to improve the energy recuperation efficiency. The experimental results on an “LF620” prototype EV validated the feasibility and effectiveness of regenerative braking and showed that the proposed fuzzy RBS was endowed with good control performance. The maximum driving range of LF620 EV was improved by 25.7% compared with non-RBS conditions.

Autonomous vehicles have received considerable attention, yet high-speed path following control remains a critical and challenging issue. At high speeds, achieving perfect control based on an accurate dynamic model of autonomous vehicles is difficult due to the nonlinear model parameters that change at the driving limit. Moreover, the strong coupling relationship between longitudinal control and lateral control also affects control performance. To address these challenges, a reinforcement learning-based control approach (RL-controller) is developed in this paper and applied to the high-speed path following task. The proposed RL-controller overcomes the challenges of explicit modeling and tight coupling control. Firstly, to obtain efficient learning features, our RL-controller is developed using a deep soft actor-critic method that incorporates multi-layer dense connection networks with skip connections and feature reuse. Secondly, an estimation method of the lane curvature is developed, which is chosen as prior knowledge to augment the input state of reinforcement learning. We model the lane using a transformer-based network structure that utilizes a third-order polynomial and train the model on our self-collected dataset. Thirdly, intuitive reward functions are built from the cross-track error, the angle error and the sideslip angle. At last, we verify our proposed RL-controller utilizing an open autonomous driving simulator. Compared with imitation learning and model-based optimal control methods, our RL-controller outperforms them in terms of the driving speed and compound error.

A reliable and safe energy storage system utilizing lithium-ion batteries relies on the early prediction of remaining useful life (RUL). Despite this, accurate capacity prediction can be challenging if little historical capacity data is available due to the capacity regeneration and the complexity of capacity degradation over multiple time scales. In this study, data decomposition, transformers, and deep neural networks (DNNs) are combined to develop a model of RUL prediction for lithium-ion batteries. Complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) is used for battery capacity sequential data to account for the capacity regeneration effect. The transformer networks are leveraged to predict each component of capacity regeneration thus improving the model's ability to handle long sequences while reducing the amount of data. The global degradation trend is predicted using a deep neural network. We validated the early prediction performance of the model using two publicly available battery datasets. Results show that the prediction model only uses 25%-30% data to achieve high accuracy. In the two public data sets, the RMSE errors were 0.0208 and 0.0337, respectively. A high level of accuracy is achieved with the model proposed in this study, which is based on fewer capacity data.

Vehicle dynamic parameters are of vital importance to establish feasible vehicle models which are used to provide active controls and automated driving control. However, most vehicle dynamics parameters are difficult to obtain directly. In this paper, a new method, which requires only conventional sensors, is proposed to estimate vehicle dynamic parameters. The influence of vehicle dynamic parameters on vehicle dynamics often involves coupling. To solve the problem of coupling, a two-stage estimation method, consisting of multiple-models and the Unscented Kalman Filter, is proposed in this paper. During the first stage, the longitudinal vehicle dynamics model is used. Through vehicle acceleration/deceleration, this model can be used to estimate the distance between the vehicle centroid and vehicle front, the height of vehicle centroid and tire longitudinal stiffness. The estimated parameter can be used in the second stage. During the second stage, a single-track with roll dynamics vehicle model is adopted. By making vehicle continuous steering, this vehicle model can be used to estimate tire cornering stiffness, the vehicle moment of inertia around the yaw axis and the moment of inertia around the longitudinal axis. The simulation results show that the proposed method is effective and vehicle dynamic parameters can be well estimated.

The penetration of electric vehicles (EVs) in vehicle-to-grid (V2G) interaction can effectively assist the grid in achieving frequency regulation and peak load balancing. However, the customer perceives that participating in V2G services would result in the additional cycling of the battery and the accelerated aging of the EVs’ power battery, which has become a major obstacle to the widespread adoption of V2G services. Most existing methods require long-term cycling data and battery parameters to quantify battery aging, which is not suitable for the V2G scenario with large-scale and short-time intervals. Consequently, the real-time and accurate quantification of battery aging for optimization remains a challenge. This study proposes a charging scheduling method for EVs that can accurately and online quantify battery aging. Firstly, V2G scheduling is formulated as an optimization problem by defining an online sliding window to collect real-time vehicle information on the grid, enabling online optimization. Secondly, battery aging is more accurately quantified by proposing a novel amplitude-based rain-flow cycle counting (MRCC) method, which utilizes the charging information of the battery within a shorter time period. Lastly, an intelligent optimization algorithm is employed to optimize the charging and discharging power of EVs, aiming to minimize grid fluctuations and battery aging. The proposed method is validated using a V2G scenario with 50 EVs with randomly generated behaviors, and the results demonstrate that the proposed online scheduling method not only reduces the EFCC of the battery by 8.4%, but also achieves results close to global optimization.