基于天气分型的风光出力互补性分析方法

基于天气分型的风光出力互补性定量分析方法能够科学指导风光互补发电系统优化调度。针对现有天气分型方法中主成分分析法无法提取非线性特征,分布领域嵌入(t-SNE)算法未考虑样本实际分布等不足,提出了基于核主成分分析(KPCA)和自组织特征映射(SOFM)神经网络的天气分型及风光出力互补性分析方法。首先,基于数值天气预报数据,利用KPCA进行特征向量提取;然后,以特征向量为输入条件,构建基于SOFM神经网络的天气类型划分模型;最后,基于波动互补率和爬坡互补率评估指标,从波动性和爬坡性2个角度定量分析不同天气类型下风光出力互补程度和最佳并网容量比例。结果表明不同天气类型下风光出力波动互补性及最佳并网容量比例差异明显,验证了所提方法的有效性。

Analysis Method for Complementarity Between Wind and Photovoltaic Power Outputs Based on Weather Classification

The quantitative analysis method of the complementarity between wind power output and photovoltaic power output based on weather classification can scientifically guide the optimal dispatch of wind-photovoltaic complementary power generation systems. In order to overcome the shortcomings of the existing weather classification methods that principal component analysis cannot extract nonlinear features and t-distributed stochastic neighbor embedding (t-SNE) based algorithm does not consider the actual distribution of samples, a weather classification and complementarity analysis method for wind and photovoltaic power output based on the kernel principal component analysis (KPCA) and the self-organizing feature map (SOFM) neural network is proposed. Firstly, the KPCA is employed to extract the feature vectors based on numerical weather prediction data. Then, a weather pattern classification model based on the SOFM neural network is constructed by using the feature vectors as input conditions. Finally, Based on the evaluation indicators for complementary ratio of fluctuation and complementary ratio of ramp, the complementary degree and the optimal grid-connected capacity ratio of wind and photovoltaic power output under different weather patterns are quantitatively analyzed from two perspectives of fluctuation and ramp. The results demonstrate that the fluctuation complementarity and the optimal grid-connected capacity ratio of wind and photovoltaic power output have obvious difference under different weather patterns, which verifies the effectiveness of the proposed method.