2080 nm long-wavelength,high-power dissipative soliton resonance in a dumbbell-shaped thulium-doped fiber laser

We demonstrate a 2080 nm long-wavelength mode-locked thulium(Tm)-doped fiber laser operating in the dissipative soliton resonance(DSR) regime. The compact all-fiber dumbbell-shaped laser is simply constructed by a 50/50 fiber loop mirror(FLM), a 10/90 FLM, and a piece of large-gain Tm-doped double-clad fiber pumped by a 793 nm laser diode. The 10/90 FLM is not only used as an output mirror, but also acts as a periodical saturable absorber for initiating DSR mode locking. The stable DSR pulses are generated at the center wavelength as long as 2080.4 nm, and the pulse duration can be tunable from 780 to 3240 ps as the pump power is increased. The maximum average output power is 1.27 W, corresponding to a pulse energy of 290 nJ and a nearly constant peak power of 93 W. This is, to the best of our knowledge, the longest wavelength for DSR operation in a mode-locked fiber laser.     

基于并行的F-LSTM模型及其在电力通信设备故障预测中的应用

电力通信网设备时序故障预测的目标是通过过去设备告警数据,预测设备在下一个时间段是否发生故障,这对设备的管理和维护起着重要作用。为了预测电力设备未来的状态,提出一种Forward-LSTM(F-LSTM)学习模型,对设备故障的时序特征和非时序特征(静态信息)进行并行训练,探索出一种新的对静态-时序数据的训练方法,将其应用在电力通信网故障预测中。F-LSTM结合了两个组件,一个学习时序特征的长短期记忆神经网络(LSTM)与一个处理静态数据的前向全连接神经网络(forward full connection neural networks,FC),数据的静态/时序属性被自动判断并传递给FC或LSTM来并行训练。对于具有同时产生动态数据与静态数据的电力通信网络,Forward-LSTM(F-LSTM)模型能以较高速度与精度预测其故障发生的位置。此外,本文采用一种加权的损失函数,可以更好地捕捉设备故障的时序规律。选取某电力通信网络系统中2016—2017年设备故障数据,对本方法进行测试。实验结果显示,与Xgboost模型相比,F-LSTM模型对故障预测的召回率提高5%,同时F-LSTM模型较LSTM模型缩减了计算量,加快了模型的训练速度。