利用庞特里亚金最小原理和深度强化学习方法优化混合储能系统的功率分配，用

发布时间：2024-12-26 11:46

使用强化学习提高深度强化学习算法的效率 #生活技巧# #学习技巧# #深度学习技巧#

摘要

基于电池-超级电容器混合储能系统（HESS）的电动汽车（EV）需要一个高效的在线能量管理系统（EMS）来延长电池寿命。间接最优控制方法，如庞特里亚金最小原理（PMP），因其固有的瞬时优化和计算简便性而备受关注。然而，在给定的驾驶场景下确定 PMP 的成本变量和非临时变量具有挑战性，会严重影响最大限度减少电池衰减的最优解。在本研究中，我们提出了一种在线混合 EMS，将 PMP 与深度强化学习（RL）相结合，以精确预测最佳成本值，同时使电池超级电容器 HESS 辅助电动汽车的电池衰减最小化。此外，我们还提出了一种估算初始成本值的分析方法，以创建 RL 的成本值行动空间，该方法也可用于估算类似应用的初始成本值猜测。我们通过仿真验证了提议的 EMS 在驾驶周期不确定和未经训练的驾驶周期下的性能。结果表明，深度 RL 可以有效地进行最优成本估算，从而满足超级电容器充电的可持续性。预测的成本值可将超级电容器的充电状态保持在理想范围内，同时最大限度地降低不确定驾驶曲线下的电池电流。此外，我们还比较了建议的 EMS 和离线优化 EMS 的电池能量消耗，以研究建议的智能 EMS 的可行性。建议的 EMS 优于离线 EMS，在 900 次充放电循环中取得了显著改善。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

Optimal power-split of hybrid energy storage system using Pontryagin’s minimum principle and deep reinforcement learning approach for electric vehicle application

The battery supercapacitor hybrid energy storage system (HESS) based electric vehicles (EVs) require an efficient online energy management system (EMS) to enhance the battery life. The indirect optimal control method, like Pontryagin’s minimum principle (PMP), is gaining attention for its inherent instantaneous optimization and computational simplicity. However, determining the costate, non-casual variable of PMP for a given driving scenario is challenging and significantly affects the optimal solution of minimizing the battery degradation. In this study, we propose an online hybrid EMS by combining PMP and deep reinforcement learning (RL) to precisely predict the optimal costate and simultaneously minimize the battery degradation in a battery supercapacitor HESS assisted EV. Also, we propose an analytical approach to estimate the initial costate to create the costate action space of RL, which could also be used to estimate the initial costate guess for similar applications. We validate the performance of the proposed EMS under driving cycle uncertainties and untrained driving cycles via simulation. The results demonstrate the effectiveness of deep RL for optimal costate estimation to satisfy the charge sustainability of the supercapacitor. The predicted costate maintains the supercapacitor state-of-charge in the desired range while minimizing battery current for uncertain driving profiles. Further, we have compared the battery energy depletion in the proposed EMS and an offline optimal EMS to study the feasibility of the proposed intelligent EMS. The proposed EMS outperforms the offline EMS, achieving a significant improvement of 900 charge–discharge cycles.

网址：利用庞特里亚金最小原理和深度强化学习方法优化混合储能系统的功率分配，用 https://www.yuejiaxmz.com/news/view/574186

上一篇: 数学建模最优化方法：动态规划

下一篇: Go 语言中的性能优化方法有哪些

利用庞特里亚金最小原理和深度强化学习方法优化混合储能系统的功率分配，用

摘要

相关内容

随便看看

最新动态分享

热点动态分享

专题

推荐动态分享