为提高风电出力的预测精度,提出一种基于Bayes优化的长短期记忆人工神经网络(long-short term memory,LSTM)的预测模型。首先,利用经验模态分解对风电历史出力序列进行分解,并对各分量及原始数据分别提取8个统计特征量,与预测前6个时刻...为提高风电出力的预测精度,提出一种基于Bayes优化的长短期记忆人工神经网络(long-short term memory,LSTM)的预测模型。首先,利用经验模态分解对风电历史出力序列进行分解,并对各分量及原始数据分别提取8个统计特征量,与预测前6个时刻出力值共同组成预测特征集。然后,采用绳索算法(least absolute shrinkage and selection operator,LASSO)从预测特征集中提取具有统计意义的特征子集,作为预测模型的输入。最后,提出基于Bayes超参数寻优的LSTM网络优化方法,以提高预测精度。选取湖北某市风电出力历史数据进行预测实验,结果表明:相较于BP神经网络、SVM、RBF网络、GRNN网络等预测模型,所提模型预测精度较高,特征提取方法较为合理。展开更多
Flexible asymmetric supercapacitor is fabricated with three dimensional(3D)Fe2O3/Ni(OH)2 composite brush anode and Ni(OH)2/MoO2 honeycomb cathode.Particularly for 3D composite brush anode,a layer of thin Fe2O3 film is...Flexible asymmetric supercapacitor is fabricated with three dimensional(3D)Fe2O3/Ni(OH)2 composite brush anode and Ni(OH)2/MoO2 honeycomb cathode.Particularly for 3D composite brush anode,a layer of thin Fe2O3 film is firmly adhered on a 3D Ni brush current collector with the assist of Ni(OH)2,functioning as both adherence layer and pseudocapacitive active material.The unique 3D Ni brush current collector possesses large surface area and stretching architecture,which facilitate to achieve the composite anode with high gravimetric capacitance of 2158 F/g.In terms of cathode,Ni(OH)2 and MoO2 have a synergistic effect to improve the specific capacitance,and the resulting Ni(OH)2/MoO2 honeycomb cathode shows a very high gravimetric capacitance up to 3264 F/g.The asymmetric supercapacitor(ASC)has balanced cathode and anode,and exhibits an ultrahigh gravimetric capacitance of 1427 F/g and an energy density of 476 W·h/kg.The energy density of ASC is 3-4 times higher than those of other reported aqueous electrolyte-based supercapacitors and even comparable to that of commercial lithium ion batteries.The device also shows marginal capacitance degradation after 1000 cycles'bending test,demonstrating its potency in the application of flexible energy storage devices.展开更多
文摘为提高风电出力的预测精度,提出一种基于Bayes优化的长短期记忆人工神经网络(long-short term memory,LSTM)的预测模型。首先,利用经验模态分解对风电历史出力序列进行分解,并对各分量及原始数据分别提取8个统计特征量,与预测前6个时刻出力值共同组成预测特征集。然后,采用绳索算法(least absolute shrinkage and selection operator,LASSO)从预测特征集中提取具有统计意义的特征子集,作为预测模型的输入。最后,提出基于Bayes超参数寻优的LSTM网络优化方法,以提高预测精度。选取湖北某市风电出力历史数据进行预测实验,结果表明:相较于BP神经网络、SVM、RBF网络、GRNN网络等预测模型,所提模型预测精度较高,特征提取方法较为合理。
文摘Flexible asymmetric supercapacitor is fabricated with three dimensional(3D)Fe2O3/Ni(OH)2 composite brush anode and Ni(OH)2/MoO2 honeycomb cathode.Particularly for 3D composite brush anode,a layer of thin Fe2O3 film is firmly adhered on a 3D Ni brush current collector with the assist of Ni(OH)2,functioning as both adherence layer and pseudocapacitive active material.The unique 3D Ni brush current collector possesses large surface area and stretching architecture,which facilitate to achieve the composite anode with high gravimetric capacitance of 2158 F/g.In terms of cathode,Ni(OH)2 and MoO2 have a synergistic effect to improve the specific capacitance,and the resulting Ni(OH)2/MoO2 honeycomb cathode shows a very high gravimetric capacitance up to 3264 F/g.The asymmetric supercapacitor(ASC)has balanced cathode and anode,and exhibits an ultrahigh gravimetric capacitance of 1427 F/g and an energy density of 476 W·h/kg.The energy density of ASC is 3-4 times higher than those of other reported aqueous electrolyte-based supercapacitors and even comparable to that of commercial lithium ion batteries.The device also shows marginal capacitance degradation after 1000 cycles'bending test,demonstrating its potency in the application of flexible energy storage devices.
