【摘要】 针对多数元启发式算法面对质子交换膜燃料电池电压模型参数辨识问题时易发生“早熟”导致参数辨识精度较低的问题，该文在改进蜉蝣算法基础上提出一种混沌映射和自适应Levy飞行的改进蜉蝣算法。首先，引入Logistic方程生成混沌序列，映射到问题空间中提高种群初始化的遍历性；其次，在蜉蝣的速度更新中加入自适应Levy飞行算法，利用Levy飞行大概率小步长，小概率大步长的特性帮助算法跳出局部最优值；此外，加入自适应策略动态调整蜉蝣速度，进一步缩短算法寻优时间；最后，通过4种测试函数在两种不同维度下的寻优结果验证算法的有效性。将所提算法应用于SR-12燃料电池电压模型参数辨识中，结果表明：相较于蜉蝣算法、改进的蜉蝣算法和粒子群算法，所提算法针对加入/未加入白噪声的实验数据均具有更快的收敛速度、更高的辨识精度以及更强的鲁棒性。更多还原

【Abstract】 To address the issue that most heuristic algorithms are prone to converge prematurely when facing the voltage model parameter identification problem of proton exchange membrane fuel cell(PEMFC), resulting in low parameter identification accuracy, this paper proposes an improved mayfly algorithm based on chaos mapping and adaptive Levy flight(LLIMA). Firstly, the logistic equation is introduced to generate chaotic sequences, which are then mapped to the problem space to enhance the exploration capability of population initialization. Secondly, the adaptive Levy flight algorithm is added to the velocity update of the mayfly, which helps the algorithm escape from the local optimum by utilizing the property of Levy flight of large probability with small step size and small probability with large step size. Additionally, the inclusion of an adaptive strategy dynamically adjusts the mayfly velocities, further shortening the optimization time of the algorithm. Finally, the effectiveness of the algorithm is verified by the optimization results of four test functions in two different dimensions. Applying the LLIMA algorithm to parameter identification of the SR-12 fuel cell voltage model demonstrates that compared to the mayfly algorithm, improved mayfly algorithm, and particle swarm algorithm, the proposed LLIMA algorithm achieves higher identification accuracy, faster convergence speed, and stronger robustness for experimental data both with and without added white noise.更多还原