无人机燃料电池混合动力系统人工神经网络控制策略

目前，全球有200多项在研电推进飞机项目，包含传统固定翼飞机改造、新构型设计、全电推进系统和混合电推进等。近年来，电推进飞机的发展方向可分为纯电动固定翼、电动垂直起降和混合电推进3类。本文针对某小型无人机燃料电池和锂电池组成的混合动力系统，建立了由质子交换膜燃料电池混合动力系统、6自由度飞行器模型和基于神经网络控制器组成的整个系统仿真模型。为了满足无人机不同飞行阶段的电力需求，控制器控制电池电流和充放电速率。在燃料电池作为无人机主要动力源的情况下，对混合动力系统的性能进行验证和评估。对人工神经网络控制器和模糊逻辑控制器性能进行了比较，结果表明，人工神经网络控制器的性能明显提高，增强了无人机混合动力系统的整体稳定性。人工神经网络控制器在效率和油耗方面比模糊逻辑控制器略有提高，约为1%。这说明在这类系统中，所设计的神经网络控制器比经典控制器具有更强的鲁棒性，更好的响应能力、可靠性，可满足无人机燃料电池混合动力系统运行在最优水平和最优效率上。

Research on Artificial Neural Network Control Strategy of Fuel Cell Hybrid Power System of Unmanned Aerial Vehicles

At present, there are more than 200 electric propulsion aircraft projects under development around the world, including the transformation of traditional fixed-wing aircraft, new configuration design, all-electric propulsion system and hybrid electric propulsion. In recent years, the development direction of electric propulsion aircraft can be divided into three categories: pure electric fixed wing , electric vertical take-off and landing and hybrid electric propulsion .In this paper, aiming at the hybrid power system composed of fuel cell and lithium battery of a small Unmanned Aerial Vehicles (UAV), the whole system simulation model is established, which is composed of proton exchange membrane fuel cell hybrid power system, 6-degree-of-freedom aircraft model and neural network controller. In order to meet the power demand of UAV in different flight stages, the controller controls the battery current and charge and discharge rate. In the case of fuel cell as the main power source of UAV, the performance of hybrid power system is verified and evaluated. The performance of the artificial neural network controller and the fuzzy logic controller are compared. the results show that the performance of the artificial neural network controller is obviously improved and the overall stability of the UAV hybrid power system is enhanced. The artificial neural network controller is slightly higher than the fuzzy logic controller in terms of efficiency and fuel consumption, which is about 1%. This shows that in this kind of system, the designed neural network controller has stronger robustness, better response ability and reliability than the classical controller, and can meet the UAV fuel cell hybrid power system running at the optimal level and optimal efficiency.