The invention provides a near space aircraft energy system management method based on deep reinforcement learning. The method comprises the steps of collecting real-time data of an energy system; the method comprises the following steps: constructing a model of an Actor-Critic framework; training the model; inputting the real-time data into an Actor network to obtain a corresponding charging and discharging strategy, and outputting an action coefficient to each node by the Actor network; discretizing the action coefficient, and outputting the switching action of each node battery; and the switching action is converted into the switching value of the energy system battery. The method has the beneficial effects that the strategy can be adjusted according to the change of environmental conditions by adopting the deep reinforcement learning model, and the effect of optimizing energy use is achieved; the intelligent charging and discharging strategy can reduce deep discharging and frequent charging and discharging circulation of the battery, so that the service life of the battery is prolonged; the deep reinforcement learning can learn to maintain stable operation of the aircraft under various conditions, and accidental shutdown or performance reduction caused by improper energy management is reduced.
本发明提供一种基于深度强化学习的临近空间飞行器能源系统管理方法,包括采集能源系统的实时数据;构建Actor‑Critic架构的模型;对模型进行训练;将实时数据输入至Actor网络中,得到对应的充放电策略,Actor网络将动作系数输出至各节点;将动作系数经过离散化后,输出各节点电池的开关动作;将开关动作转换为能源系统电池的开关量。本发明的有益效果是通过采用深度强化学习模型能够根据环境条件的变化调整其策略,达到最优化能源使用的效果;智能充放电策略可以减少电池的深度放电和频繁的充放电循环,从而延长电池的使用寿命;深度强化学习可以学习在各种条件下维持飞行器的稳定运行,减少因能源管理不当导致的意外停机或性能下降。
Near space aircraft energy system management method based on deep reinforcement learning
基于深度强化学习的临近空间飞行器能源系统管理方法
2024-10-08
Patent
Electronic Resource
Chinese
IPC: | H02J CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER , Schaltungsanordnungen oder Systeme für die Abgabe oder Verteilung elektrischer Leistung / B64G Raumfahrt , COSMONAUTICS / G06N COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS , Rechnersysteme, basierend auf spezifischen Rechenmodellen |
Dual-motor energy management method and system based on deep reinforcement learning
European Patent Office | 2024
|HEV energy management method based on deep reinforcement learning A3C algorithm
European Patent Office | 2020
|