The energy management system (EMS) of hybridization and electrification plays a pivotal role in improving the stability and cost-effectiveness of future vehicles. Existing efforts mainly concentrate on specific optimization targets, like fuel consumption, without sufficiently taking into account the degradation of on-board power sources. In this context, a novel multi-objective energy management strategy based on deep reinforcement learning is proposed for a hybrid electric vehicle (HEV), explicitly conscious of lithium-ion battery (LIB) wear. To be specific, this paper mainly contributes to three points. Firstly, a non-parametric reward function is introduced, for the first time, into the twin-delayed deep deterministic policy gradient (TD3) strategy, to facilitate the optimality and adaptability of the proposed energy management strategy and to mitigate the effort of parameter tuning. Then, to cope with the problem of state redundancy, state space refinement techniques are included in the proposed strategy. Finally, battery health is incorporated into this multi-objective energy management strategy. The efficacy of this framework is validated, in terms of training efficiency, optimality and adaptability, under various standard driving tests.

Electrified vehicles are important for reducing emissions and will dominate road transportation in the future. An energy management strategy is critical for hybrid electric vehicles to manage the use of on-board energy resources. This paper proposes a novel double-layer energy management strategy based on hierarchical rewarding DDPG (HR-DDPG) for a heavy-duty hybrid electric vehicle (HEV) to improve energy efficiency adaptively to various scenarios. The proposed hierarchical rewarding structure with two reward functions can guide the DDPG agent to explore the optimal policy more efficiently and deeply and is well-suited to achieve targeted adjustment according to the vehicle operating modes to cope with rapidly changing scenarios. The optimality and adaptability of the new strategy are evaluated by standard and real-world driving data. The results show that the proposed strategy leads to an improvement of 8.11% in energy efficiency compared with the conventional DDPG strategy, and its energy-saving performance can reach up to 93.87% of the capacity of a dynamic programming (DP) based energy management strategy on the comprehensive driving condition II.

The emerging deep reinforcement learning (DRL) algorithms are promising for energy management of plug-in hybrid electric vehicles (PHEVs), but most existing DRL methods cannot meet the requirement for learning robustness and data independence, which are essential to mission-critical systems like PHEVs. By incorporating Bayesian Optimization (BO) with the Soft Actor-Critic (SAC) algorithm, this paper proposes a new energy management strategy (EMS) for HEVs, namely, BO-SAC EMS. The present work contributes solutions to address the above-mentioned two challenges in DRL-based EMS: 1) through BO-based hyper-parameter tuning, the brittle convergence characteristics and robustness of the SAC algorithm have been significantly improved; and 2) by introducing a state-action-reward (SAR) co-design scheme for the SAC algorithm, the dependence on real-world data has been considerably reduced, thus improving training efficiency. Using the original SAC (Origin-SAC) method and dynamic programming (DP) results as the baseline, comparison studies are conducted under ten driving cycles. By estimating the maximum energy consumption per 100 kilometres based on the Six Sigma theory, BO-SAC is shown more robust by saving more than 3% energy in the worst case.

Cooperative adaptive cruise control (CACC) is a critical function that faces significant challenges in maintaining platoon stability and achieving energy efficiency, especially in real-world operations. The CACC of connected and autonomous vehicles (CAVs) based on the multi-agent reinforcement learning (MARL) algorithm is studied to optimize platoon stability and energy efficiency simultaneously. Then the effectiveness of communication information is the key to guaranteeing learning performance in real-world driving, and thus this paper has proposed a communication-efficient MARL by incorporating the quantified stochastic gradient descent (QSGD) and a binary differential consensus (BDC) method into a fully-decentralized MARL framework. We evaluate this BDC-MARL algorithm against several typical non-communicative and communicative MARL algorithms, including IA2C, FPrint, and DIAL, focusing on metrics such as platoon stability, fuel economy, and driving comfort. Our results demonstrate that BDC-MARL achieves superior energy savings, with improvements of up to 5.8%, an average velocity of 15.26m/s, and an inter-vehicle spacing of 20.76m. Additionally, we perform comprehensive analyses of communicative information-sharing efficiency and scalability across varying platoon sizes, further validating the practical effectiveness through real-world scenarios using data from the open-source OpenACC.

The Proton Exchange Membrane Fuel Cell (PEMFC) is a fast-developing battery technology, and the key to its reliability and lifespan improvement lies in the accurate assessment of durability. However, the degradation mechanism of the PEMFC is hard to determine and its internal parameters are highly coupled. Thus, the development of a more accurate life prediction model that meets the actual scenarios is to be investigated urgently. To solve this problem, a multi-feature fusion life prediction method based on the Temporal Convolutional Network-Gated Recurrent Unit (TCN-GRU) is proposed. A TCN algorithm is used as the prediction base model, and two GRU modules are included with the model to strengthen the model’s expression ability and improve its predictive accuracy. Two widely recognized datasets and two operating conditions are utilized for model training and prediction, respectively. Comparisons are made with single-feature parameter models in terms of Root Mean Square Error (RMSE) and the Determination Coefficient (R²). The results show that the prediction accuracy of the TCN-GRU multi-feature fusion model is higher than that of the single-feature models in terms of stability and anti-interference under both operating conditions. The accuracy of the TCN-GRU (three-feature) model is the most optimal in a steady-state condition at 80% of the training set ratio (RMSE = 3.27 × 10⁻³, R² = 0.965). Furthermore, with the increase in the input feature parameter, the TCN-GRU model is closer to the real value, which proves once again that the proposed model can meet the accuracy requirements of the life prediction of the PEMFC.