基于双分支特征提取的太阳辐射预测方法

Dual branch feature extraction based solar radiation forecasting

针对太阳辐射预测过程中气象特征复杂、时序特征难以充分利用而导致光伏功率出力扰动的问题,提出一种基于双分支特征提取的太阳辐射逐日预测方法。气象分支采用多尺度卷积神经网络提取动态变化的多维气象特征;时序分支使用双向门控循环网络初步提取时序特征,将学习到的双向时序特征输入门控循环网络进一步学习其潜在规律;基于注意力机制自适应地赋予各分支合适的权值,优化多尺度卷积的提取操作和气象、时序特征的融合过程。经过实验验证了该预测方法的准确性和有效性。

Aiming at the complicated meteorological characteristics and the difficulty of full utilization of time sequence characte-ristics,leading to the disturbance of photovoltaic power output in the process of solar radiation forecasting,a daily solar radiation forecasting method based on double-branch feature extraction was proposed.The branch of meteorological took a multi-scale convolutional neural network to extract the multi-dimensional characteristics of the dynamic changes of the meteorological data set.The timing branch used a bidirectional gated recurrent unit neural network(BiGRU)to extract the time series features,the output of the BiGRU was used as the input of the gated recurrent unit neural network(GRU)to further learn the potential features.Based on the attention mechanism,the appropriate weights were assigned to each branch adaptively,and the extraction operation of multi-scale convolution and the fusion process of meteorological and time series features were optimized.The effectiveness and accuracy of the proposed method were validated by experiments.