失配条件下的基于变步长法的全局最大功率点跟踪算法仿真研究

针对光伏系统在实际工况条件下的失配问题,提出一种基于变步长法的全局最大功率跟踪算法,该方案可以在多个局部最大功率点条件下快速找到全局最大功率点,实现光伏系统能效的有效提升。相比较现有方案,该方案具备快速、简单和稳定的特点。

基于正则化的深度神经网络语音增强自适应方法

在有监督的语音增强系统中,训练模型的数据与测试数据的环境差异导致系统应用场景受限。针对训练与测试数据之间的失配问题,提出一种基于正则化的自适应方法,提高语音增强系统在不匹配噪声环境下的鲁棒性。采用保守训练的方法,尝试3种正则项,将其加入优化准则中,用少量自适应数据调整噪声无关的DNN。该方法在失配环境下进行测试,实验结果证明,改进系统的性能相比基线系统可以提升最多可达14.37%的短时客观可懂度(short-time objective intelligibility, STOI),并在感知评估语音质量(perceptual evaluation of speech quality, PESQ)上也取得一致提升。

Evolution of misfit dislocation network and tensile properties in Ni-based superalloys:a molecular dynamics simulation

The structural evolution of dislocation network is closely related to y' rafting and tensile properties. In this work, the effects of strain rate and temperature on the structural evolution of interface dislocation network in Ni-based superalloys are studied by molecular dynamics simulations. The correlation between the evolution of dislocation network and tensile properties is also explored. The results indicate that the dislocation network shows different degrees of deformation and damage at various strain rates and temperatures. The ),' rafting depends on the damage structure of dislocation network at various strain rates and tem- peratures. Moreover, the tensile properties of interface in Ni-based superalloys are closely related to the evolution of disloca- tion network and dislocation motion mechanisms.