基于GSO-ELM的锂离子电池剩余寿命间接预测方法研究

发布时间：2018-05-01 17:22

本文选题：锂离子电池 + 剩余寿命预测　；参考：《大连海事大学》2017年硕士论文

【摘要】：锂离子电池作为一种能量供应部件,具有体积小、重量轻、温度范围宽和高能量比等诸多优点,是很多复杂系统、关键电子设备的组成部分,对整个系统起着至关重要的作用。在系统运行时,若没有对锂离子电池的剩余寿命(remaining useful life,RUL)进行监测以便及时更换新的电池或维修,一旦发生故障则会直接导致系统性能降低,甚至引发灾难性的故障并造成重大损失,尤其对于航空航天系统,因此,锂离子电池RUL预测具有极其重要的实用价值和研究意义。本文针对锂离子电池的RUL预测问题展开研究,提出了基于萤火虫优化算法的极限学习机(glowworm swarm optimization-extreme learning machine,GSO-ELM)锂离子电池RUL预测模型。针对传统单隐层前馈神经网络学习算法存在的缺陷,在单隐层前馈神经网络中引入了极限学习机算法,并对极限学习机算法的理论进行了详细说明。针对锂电池容量难以在线监测的问题,采用等压降放电时间序列作为间接健康因子,并建立了 RUL间接预测框架。选择NASA锂电池实验数据(B5,B6,B7,B18)为研究对象,首先提取了实际容量;然后选取恒压放电模式,提取了电压、电流以及充放电循环周期,得到了等压降放电时间序列;最后采用偏相关系数法证明了电池实际容量和等压降放电时间序列之间存在着强关联性,从而证明了采用等压降放电时间序列进行间接RUL预测方法的可行性。采用萤火虫算法对极限学习机的不足之处进行优化,建立了萤火虫优化算法的极限学习机(GSO-ELM)模型,并将GSO-ELM算法模型用于锂离子电池间接RUL预测框架中。将GSO-ELM模型各个参数对模型性能的影响进行了详细分析并验证,选取合适的参数进行RUL预测,结果表明,GSO-ELM算法继承了 ELM学习速度快的优点,而且,与ELM相比,它的跟踪效果要更好。其中,由实验数据(B5)可见,用GSO-ELM算法得到的RUL误差为2,平均相对误差为5.36%,表明了该算法不但减小了 RUL预测的误差,而且提高了算法结果的稳定性。
[Abstract]:As a kind of energy supply component, Li-ion battery has many advantages, such as small volume, light weight, wide temperature range and high energy ratio. It is a part of many complex systems and key electronic equipment, which plays an important role in the whole system. When the system is running, if the residual life of the lithium ion battery is not monitored in order to replace the new battery or maintain it in a timely manner, it will directly result in a deterioration in system performance in the event of failure. Even causing catastrophic failure and causing great losses, especially for aerospace systems, lithium ion battery RUL prediction has an extremely important practical value and research significance. In this paper, the RUL prediction model of lithium-ion battery based on firefly optimization algorithm is proposed, and the RUL prediction model of glow worm swarm optimization-extreme learning machine GSO-ELM) is proposed. Aiming at the shortcomings of the traditional single hidden layer feedforward neural network learning algorithm, this paper introduces the ultimate learning machine algorithm into the single hidden layer feedforward neural network, and explains the theory of the ultimate learning machine algorithm in detail. To solve the problem that the capacity of lithium battery is difficult to be monitored on line, the isobaric discharge time series is used as the indirect health factor, and the indirect prediction framework of RUL is established. The experimental data of NASA lithium-ion battery (NASA) were selected as the research object. The actual capacity was extracted firstly, then the voltage, current and cycle of charge and discharge were extracted, and the time series of isobaric discharge were obtained by selecting the constant voltage discharge mode and the cycle period of charge and discharge. Finally, the partial correlation coefficient method is used to prove that there is a strong correlation between the actual capacity of the battery and the discharge time series of isobaric drop, which proves the feasibility of the indirect RUL prediction method using the isobaric drop discharge time series. The limitation of the ultimate learning machine is optimized by using the firefly algorithm. The GSO-ELM model of the firefly optimization algorithm is established, and the GSO-ELM algorithm model is applied to the indirect RUL prediction framework for lithium-ion batteries. The influence of each parameter of GSO-ELM model on the model performance is analyzed and verified in detail, and the appropriate parameters are selected for RUL prediction. The results show that the GSO-ELM algorithm inherits the advantages of fast ELM learning speed, and compared with ELM, the GSO-ELM algorithm has the advantages of fast learning speed. It has better tracking effects. It can be seen from the experimental data (B5) that the RUL error obtained by GSO-ELM algorithm is 2 and the average relative error is 5.36, which indicates that the algorithm not only reduces the error of RUL prediction, but also improves the stability of the result.
【学位授予单位】：大连海事大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM912

【相似文献】