基于风速属性优化聚类的时序卷积特征聚合风速预测

Time Series Convolution Feature Aggregation Based Wind Speed Prediction Considering Wind Speed Attribute Optimization Clustering

准确的风速预测对于规模化风电并网及安全运行具有重要意义。利用快速相关滤波筛选风速关联属性因素并结合近邻传播原理优化风速聚类集合,提出基于时序卷积特征聚合的风速预测模型。考虑风速属性与风速序列间的隐含关联性,筛选高关联属性因素构建模型样本集,并通过鲸群算法优化近邻传播聚类分类相似典型集。构建时序卷积层提取多维风速属性特征,并嵌入特征聚合层完成特征降维与信息融合,最终结合记忆层输出风速预测值。以东北地区风场为研究对象进行风速超短期预测并与实测数据对比,验证了预测模型的准确性和泛化能力。

Accurate wind speed prediction is of great significance for large-scale wind power integration and safe operation.The paper proposes a wind speed prediction model based on temporal convolution feature aggregation,using fast correlation filter to screen attribute factors related to the wind speed and optimizing the wind speed clustering set based on affinity propagation.Considering the implicit correlation between the wind speed attributes and wind speed series,the high correlation factors are selected to construct the model sample set,and the whale swarm algorithm is used to optimize the affinity propagation clustering to divide the similar typical set.The temporal convolution layer is constructed to extract multi-dimensional wind speed attribute features,and the feature aggregation layer is embedded to conduct feature dimension reduction and information fusion.Finally,the wind speed prediction value is output with a memory layer.Taking the wind field in northeast China as the research object,the ultra-short-term prediction of the wind speed is carried out and compared with the measured data,verifying the accuracy and generalization ability of the prediction model.