1,720,967 research outputs found
Advancing vehicle electrification: control strategies for hybrid electric vehicles and battery management algorithms for state estimation
No full textL'abstract è presente nell'allegato / the abstract is in the attachmen
States Estimation for Parallel-Connected Battery Module: A Moving Horizon Approach
No full textIn this work, a moving horizon estimation (MHE)-based method is developed for estimating battery cells state in parallel-connected modules. Unlike conventional approaches, the proposed method acknowledges the impact of cell-to-cell (CtC) variations and heterogeneity propagation on module performance. A nonlinear observability analysis is performed to assess the feasibility of reconstructing individual cell states from module voltage and current measurements, considering interconnection resistance, state of charge (SOC)-dependent parameters, and different numbers of cells. The results indicate that states are distinguishable when the interconnection resistance is not null, and observability improves as the number of cells in parallel decreases. To the best of our knowledge, this is the first application of MHE in the context of battery modules, validated with real-world battery data. In contrast with conventional estimation methods, this study leverages MHE’s ability to handle equality constraints, allowing for the solution of Kirchhoff’s laws without complicating the module dynamics, maintaining the estimation accuracy. The proposed estimation algorithm demonstrates robustness against measurement noise and model uncertainties, with a maximum SOC error below 2.65%. Furthermore, the MHE results are compared against two widely used observers, the extended Kalman filter (EKF) and unscented Kalman filter (UKF), showing consistently higher estimation accuracy across all experimental conditions
Acceleration control strategy for Battery Electric Vehicle based on Deep Reinforcement Learning in V2V driving
Full text linkThe transportation sector is seeing the flourishing of one of the most interesting technologies, autonomous driving (AD). In particular, Cooperative Adaptive Cruise Control (CACC) systems ensure higher levels both of safety and comfort, enhancing at the same time the reduction of energy consumption. In this framework a real-time velocity planner for a Battery Electric Vehicle, based on a Deep Reinforcement Learning algorithm called Deep Deterministic Policy Gradient (DDPG), has been developed, aiming at maximizing energy savings, and improving comfort, thanks to the exchange of information on distance, speed and acceleration through the exploitation of vehicle-to-vehicle technology (V2V). The aforementioned DDPG algorithm relies on a multi-objective reward function that is adaptive to different driving cycles. The simulation results show how the agent can obtain good results on standard cycles, such as WLTP, UDDS and AUDC, and on real-world driving cycles. Moreover, it displays great adaptability to driving cycles different from the training one
Online Temperature-aware Equivalent Consumption Minimization Strategy for Mild Hybrid Electric Powertrains
Full text linkIn this work, an online energy management strategy for mild hybrid electric vehicles is developed to minimize the fuel consumption while simultaneously preventing battery overheating. Since mild hybrids are typically equipped with a passively cooled battery pack, the energy management strategy design needs to keep the battery temperature below an upper limit, preventing accelerated aging and thermal runaway. To address this issue, the equivalent consumption minimization strategy (ECMS) approach is extended to develop a real-time capable controller, termed thermal ECMS (Th-ECMS), that is sensitive to the thermal dynamics of the battery and that can enforce constraints on its temperature. The rationale for our formulation is based on Pontryagin's minimum principle from optimal control theory. The online Th-ECMS is developed on the basis of the offline version of Th-ECMS, introduced in a previous work. Exploiting the a priori knowledge of the driving mission, the offline Th-ECMS calibrates the equivalence factors and obtains the optimal solution, which is compared with the globally optimal dynamic programming solution. This offline calibration method is run on a large number of driving missions and the collected data is used to train a feed-forward neural network that estimates optimal equivalence factors as functions of the battery state of charge, battery temperature, and distance yet to travel. The trained network is then used to populate two look-up tables mapping the equivalence factors, and implementable on the vehicle electronic control unit. Finally, the online Th-ECMS obtains the equivalence factors through the look-up tables in real-time. The online strategy was tested in four different driving missions, achieving a fuel economy remarkably similar to the optimal solution and successfully avoiding battery overheating
Battery Electric Vehicle Control Strategy for String Stability based on Deep Reinforcement Learning in V2V Driving
Full text linkThis works presents a Reinforcement Learning (RL) agent to implement a Cooperative Adaptive Cruise Control (CACC) system that simultaneously enhances energy efficiency and comfort, while also ensuring string stability. CACC systems are a new generation of ACC which systems rely on the communication of the so-called egovehicle with other vehicles and infrastructure using V2V and/ or V2X connectivity. This enables the availability of robust information about the environment thanks to the exchange of information, rather than their estimation or enabling some redundancy of data. CACC systems have the potential to overcome one typical issue that arises with regular ACC, that is the lack of string stability. String stability is the ability of the ACC of a vehicle to avoid unnecessary fluctuations in speed that can cause traffic jams, dampening these oscillations along the vehicle string rather than amplifying them. In this work, a real-time ACC for a Battery Electric Vehicle, based on a Deep Reinforcement Learning algorithm called Deep Deterministic Policy Gradient (DDPG), has been developed, aiming at maximizing energy savings, and improving comfort, thanks to the exchange of information on distance, speed and acceleration through the exploitation of vehicle-to-vehicle technology (V2V). The aforementioned DDPG algorithm is also designed in order to achieve the string stability. It relies on a multi-objective reward function that is adaptive to different driving cycles. The simulation results show how the agent can obtain energy savings up to 11% comparing the first following vehicle and the Lead on standard cycles and good adaptability to driving cycles different from the training one
Battery temperature aware equivalent consumption minimization strategy for mild hybrid electric vehicle powertrains
Full text linkAn energy management strategy for mild hybrids that prevents battery overheating is introduced in this digest. Energy management strategy design for mild hybrids requires particular care to prevent overheating of the battery pack as they typically do not have an active cooling system. To tackle this issue, we extend the well-known equivalent consumption minimization strategy approach to develop a real-time capable fuel-optimal controller that is sensitive to the battery’s thermal dynamics and that can enforce constraints on its temperature. The rationale for our formulation is developed using Pontryagin’s minimum principle from optimal control theory. The same principle is also used to design an off-line numerical procedure for the energy management strategy’s calibration. The effectiveness of the procedure is corroborated by numerical experiments on two different drive cycles, whose results are also compared with the solution obtained with a dynamic programming algorithm. Several peculiar aspects of our solution procedure, such as the method used to incorporate state constraints and the approximate boundary value problem solution method using a particle swarm optimization algorithm, are also detailed and discussed. The proposed controller is computationally light-weight and can be readily extended to on-line control provided that a suitable co-state selection procedure is employed, based on the data collected by using our calibration method on a large number of driving missions
Safe Reinforcement Learning for Energy Management of Electrified Vehicle with Novel Physics-Informed Exploration Strategy
Full text linkThis paper introduces a novel physics-informed exploration strategy for a deep reinforcement learning (DRL)-based energy management system (EMS), specifically targeting the challenge of dealing with constrained action sets. RL-based controllers for electrified vehicle energy management systems have faced obstacles stemming from the selection of infeasible actions, obstructing their practical deployment. The absence of a mechanism for assessing control action feasibility prior to application has compounded this issue, primarily due to the model-free nature of RL-based controllers. Adding a safety layer to the RL-based controller addresses the abovementioned issue, but this often results in suboptimal policies and necessitates an in-depth understanding of the powertrain. Alternatively, theoretical remedies incorporate penalty terms into the immediate reward function to manage infeasible conditions. However, this approach can slow down the training process as the agent learns to avoid infeasible actions. To surmount these challenges, this paper introduces a novel physics-informed exploration strategy, coupled with prioritized experience replay, enabling the agent to swiftly learn to avoid selecting infeasible control actions without the need for a separate safety layer. Real-time simulation results highlight the superior performance of the proposed DRL-based controller over the baseline DRL-based controller with a safety layer, particularly in terms of overall fuel consumption
Influence of the reward function on the selection of reinforcement learning agents for hybrid electric vehicles real-time control
No full textRegression based battery state of health estimation for multiple electric vehicle fast charging protocols
Full text linkIn this work, a data-driven estimation method is developed to estimate the battery state of health (SOH), exploiting SOH features that can be obtained during fast-charging events. A newly expanded experimental dataset with six cells, cycled 1200 to 1800 times until 70% SOH is reached, is used and made available. Our investigation focuses on the variability that can be encountered in charging events due to different charging protocols (particularly for fast charging) and partial charging events. In particular, we investigated nine different SOH features, introducing novel formulations to increase their flexibility with respect to different charging events. Then, we assessed the practical implementability of these features and employed correlation and feature importance analyses to identify the most effective. Finally, we developed a linear regression model for SOH estimation using the selected features as inputs. The model shows an RMS prediction error as low as 1.09% over the battery lifetime and a maximum error no greater than 3.5% until SOH falls below 80%, corresponding to the end-of-life for automotive applications. The estimator is also shown to be robust against significant errors of the state of charge (SOC) input value (as high as 5%), ensuring it will perform well even when SOC is not accurately known
- …
