Laboratory experiments were conducted to investigate the transformation and performance of a granular sequence batch reactor(SBR) under the conventional organic loading rate(OLR) condition.Aerobic granules were succes...Laboratory experiments were conducted to investigate the transformation and performance of a granular sequence batch reactor(SBR) under the conventional organic loading rate(OLR) condition.Aerobic granules were successfully cultivated in a SBR by means of alternative feeding load combined with reducing settling time after 60 d operational period.Subsequently,the black fungal granules were presented in reactor because of the filamentous overgrowth on the surface of aerobic granules.A small amount of fungal granules had no effect on the performance of granular SBR.Aerobic granules completely vanished and fungal granules eventually became the dominant species in subsequent 90 d operation after granulation.The three-dimensional excitation emission matrix(EEM) spectra result shows that the extracellular polymeric substances(EPS) component in both granules has no much difference,whereas the content of EPS in fungal granules is higher than that in bacterial granules.Due to their low bioactivity,the chemical oxidation demand(COD) and NH4-N removal efficiencies gradually decrease from 90.4%–96.5% and 99.5% to 71.8% and 32.9% respectively while the fungal granules become dominant in the SBR.展开更多
随着建筑物能源消耗的不断升高,高精度与高泛化能力的非侵入式负荷监测技术的研究具有重大意义。针对当前负荷分解方法存在的问题,提出了一种基于多尺度特征融合与多任务学习框架的非侵入式负荷监测方法。将实例-批归一化网络与U形网络...随着建筑物能源消耗的不断升高,高精度与高泛化能力的非侵入式负荷监测技术的研究具有重大意义。针对当前负荷分解方法存在的问题,提出了一种基于多尺度特征融合与多任务学习框架的非侵入式负荷监测方法。将实例-批归一化网络与U形网络结合,提取总负荷数据的上下文信息,并利用跨越连接实现对不同尺度的细节特征与全局特征的融合。针对多特征特点,引入高效通道注意力网络,使模型聚焦重要特征。引入多任务学习框架与后处理操作,去除输出的假阳性片段,实现对目标电器的精准识别。将所提模型与几种代表性模型在UK-DALE(UK domestic appliance-level electricity)数据集与REDD(reference energy disaggregation data set)上进行对比实验,结果表明,所提模型的性能优于对比模型,具有出色的负荷分解能力与状态识别能力。展开更多
基金Project(51078036) supported by the National Natural Science Foundation of China
文摘Laboratory experiments were conducted to investigate the transformation and performance of a granular sequence batch reactor(SBR) under the conventional organic loading rate(OLR) condition.Aerobic granules were successfully cultivated in a SBR by means of alternative feeding load combined with reducing settling time after 60 d operational period.Subsequently,the black fungal granules were presented in reactor because of the filamentous overgrowth on the surface of aerobic granules.A small amount of fungal granules had no effect on the performance of granular SBR.Aerobic granules completely vanished and fungal granules eventually became the dominant species in subsequent 90 d operation after granulation.The three-dimensional excitation emission matrix(EEM) spectra result shows that the extracellular polymeric substances(EPS) component in both granules has no much difference,whereas the content of EPS in fungal granules is higher than that in bacterial granules.Due to their low bioactivity,the chemical oxidation demand(COD) and NH4-N removal efficiencies gradually decrease from 90.4%–96.5% and 99.5% to 71.8% and 32.9% respectively while the fungal granules become dominant in the SBR.
文摘随着建筑物能源消耗的不断升高,高精度与高泛化能力的非侵入式负荷监测技术的研究具有重大意义。针对当前负荷分解方法存在的问题,提出了一种基于多尺度特征融合与多任务学习框架的非侵入式负荷监测方法。将实例-批归一化网络与U形网络结合,提取总负荷数据的上下文信息,并利用跨越连接实现对不同尺度的细节特征与全局特征的融合。针对多特征特点,引入高效通道注意力网络,使模型聚焦重要特征。引入多任务学习框架与后处理操作,去除输出的假阳性片段,实现对目标电器的精准识别。将所提模型与几种代表性模型在UK-DALE(UK domestic appliance-level electricity)数据集与REDD(reference energy disaggregation data set)上进行对比实验,结果表明,所提模型的性能优于对比模型,具有出色的负荷分解能力与状态识别能力。
