Performing analytics on the load curve(LC)of customers is the foundation for demand response which requires a better understanding of customers'consumption pattern(CP)by analyzing the load curve.However,the perfor...Performing analytics on the load curve(LC)of customers is the foundation for demand response which requires a better understanding of customers'consumption pattern(CP)by analyzing the load curve.However,the performances of previous widely-used LC clustering methods are poor in two folds:larger number of clusters,huge variances within a cluster(a CP is extracted from a cluster),bringing huge difficulty to understand the electricity consumption pattern of customers.In this paper,to improve the performance of LC clustering,a clustering framework incorporated with community detection is proposed.The framework includes three parts:network construction,community detection,and CP extraction.According to the cluster validity index(CVI),the integrated approach outperforms the previous state-of-the-art method with the same amount of clusters.And the approach needs fewer clusters to achieve the same performance measured by CVI.展开更多
Dynamic networks have become popular to enhance the model capacity while maintaining efficient inference by dynamically generating the weight based on over-parameters.They bring much more parameters and increase the d...Dynamic networks have become popular to enhance the model capacity while maintaining efficient inference by dynamically generating the weight based on over-parameters.They bring much more parameters and increase the difficulty of the training.In this paper,a multi-layer dynamic convolution(MDConv) is proposed,which scatters the over-parameters over multi-layers with fewer parameters but stronger model capacity compared with scattering horizontally;it uses the expanding form where the attention is applied to the features to facilitate the training;it uses the compact form where the attention is applied to the weights to maintain efficient inference.Moreover,a multi-layer asymmetric convolution(MAConv) is proposed,which has no extra parameters and computation cost at inference time compared with static convolution.Experimental results show that MDConv achieves better accuracy with fewer parameters and significantly facilitates the training;MAConv enhances the accuracy without any extra cost of storage or computation at inference time compared with static convolution.展开更多
基金Supported by the Major Program of National Natural Science Foundation of China(No.61432006)。
文摘Performing analytics on the load curve(LC)of customers is the foundation for demand response which requires a better understanding of customers'consumption pattern(CP)by analyzing the load curve.However,the performances of previous widely-used LC clustering methods are poor in two folds:larger number of clusters,huge variances within a cluster(a CP is extracted from a cluster),bringing huge difficulty to understand the electricity consumption pattern of customers.In this paper,to improve the performance of LC clustering,a clustering framework incorporated with community detection is proposed.The framework includes three parts:network construction,community detection,and CP extraction.According to the cluster validity index(CVI),the integrated approach outperforms the previous state-of-the-art method with the same amount of clusters.And the approach needs fewer clusters to achieve the same performance measured by CVI.
基金Supported by the National Key Research and Development Program of China(No.2016YFB1000601)the Standardization Pilot Research Project of Chinese Academy of Sciences(No.20194620)。
文摘Dynamic networks have become popular to enhance the model capacity while maintaining efficient inference by dynamically generating the weight based on over-parameters.They bring much more parameters and increase the difficulty of the training.In this paper,a multi-layer dynamic convolution(MDConv) is proposed,which scatters the over-parameters over multi-layers with fewer parameters but stronger model capacity compared with scattering horizontally;it uses the expanding form where the attention is applied to the features to facilitate the training;it uses the compact form where the attention is applied to the weights to maintain efficient inference.Moreover,a multi-layer asymmetric convolution(MAConv) is proposed,which has no extra parameters and computation cost at inference time compared with static convolution.Experimental results show that MDConv achieves better accuracy with fewer parameters and significantly facilitates the training;MAConv enhances the accuracy without any extra cost of storage or computation at inference time compared with static convolution.
