传统光伏发电功率预测存在因气象因素特征提取不综合不精确而导致预测精度不高的问题.为了充分挖掘气象因素对光伏出力的影响,并有效利用深度学习技术在非线性拟合方面的优势,本文提出了一种基于气象因素充分挖掘的双向长短期记忆(Bi-di...传统光伏发电功率预测存在因气象因素特征提取不综合不精确而导致预测精度不高的问题.为了充分挖掘气象因素对光伏出力的影响,并有效利用深度学习技术在非线性拟合方面的优势,本文提出了一种基于气象因素充分挖掘的双向长短期记忆(Bi-directional Long Short Term Memory,BiLSTM)网络光伏发电短期功率预测方法.在对原始数据进行异常值及标准化处理的基础上,采用K近邻算法(K-Nearest Neighbor,KNN)在外界温度、湿度、压强等诸多气象因素中充分挖掘影响光伏出力的关键因素,重构多元数据序列,并在探索输入层时间步长、模型层数及每层维数等超参数的合理设置方案的基础上,构建BiLSTM网络模型,实现光伏发电短期功率的高精度预测.仿真结果表明,与KNN、深度信念网络(DBN)、BiLSTM、PCA-LSTM等经典方法比较,所提KNN-BiLSTM方法具有更高的预测精度.展开更多
An approach to improve the image quality by aperture synthesis in generalized phase-shifting interferometry is proposed.The aperture of the holographic system is enlarged by increasing the recording area to receive a ...An approach to improve the image quality by aperture synthesis in generalized phase-shifting interferometry is proposed.The aperture of the holographic system is enlarged by increasing the recording area to receive a more diffusive spectrum by moving a singular CCD.The sub-object wave fronts are combined with the index of their intensities.This method can avoid the negative effect of environmental disturbance and improve the synthesis efficiency by hundreds of times.The optical experiments have shown that it is simple and robust to stitch the complex object wave front accurately and to improve the imaging quality greatly,especially for the recording of large scale objects.展开更多
文摘传统光伏发电功率预测存在因气象因素特征提取不综合不精确而导致预测精度不高的问题.为了充分挖掘气象因素对光伏出力的影响,并有效利用深度学习技术在非线性拟合方面的优势,本文提出了一种基于气象因素充分挖掘的双向长短期记忆(Bi-directional Long Short Term Memory,BiLSTM)网络光伏发电短期功率预测方法.在对原始数据进行异常值及标准化处理的基础上,采用K近邻算法(K-Nearest Neighbor,KNN)在外界温度、湿度、压强等诸多气象因素中充分挖掘影响光伏出力的关键因素,重构多元数据序列,并在探索输入层时间步长、模型层数及每层维数等超参数的合理设置方案的基础上,构建BiLSTM网络模型,实现光伏发电短期功率的高精度预测.仿真结果表明,与KNN、深度信念网络(DBN)、BiLSTM、PCA-LSTM等经典方法比较,所提KNN-BiLSTM方法具有更高的预测精度.
基金Supported by the National Natural Science Foundation of China under Grant Nos 60977006,11004252 and 60777008the Natural Science Foundation of Shandong Province under Grant Nos ZR2009GM016 and ZR2010AQ022.
文摘An approach to improve the image quality by aperture synthesis in generalized phase-shifting interferometry is proposed.The aperture of the holographic system is enlarged by increasing the recording area to receive a more diffusive spectrum by moving a singular CCD.The sub-object wave fronts are combined with the index of their intensities.This method can avoid the negative effect of environmental disturbance and improve the synthesis efficiency by hundreds of times.The optical experiments have shown that it is simple and robust to stitch the complex object wave front accurately and to improve the imaging quality greatly,especially for the recording of large scale objects.
