基于CNN-AE-MAML的低压配电网自适应分类方法

Adaptive Classification Method of Low Voltage Distribution Network Based on CNN-AE-MAML

低压配电网分类有利于提高低压配电网经济运行管理措施及新能源规划运行方案制定的效率。随着各类新能源、充电桩等新型源荷的不断接入,低压配电网原有负荷特征发生变化,一方面导致配电网负荷特征复杂,另一方面导致变化后可用的新负荷特征数据较少,给配电网分类带来挑战。针对以上挑战,提出一种基于卷积自编码器和模型不可知元学习(convolutional neural network-auto encoder-model agnostic meta learning,CNN-AE-MAML)的低压配电网自适应分类方法。首先,利用卷积自编码器(convolutional neural network auto encoder,CNN-AE)提取可表征低压配电网的配变负荷、光伏发电特征,采用谱聚类(spectral clustering,SC)对低压配电网进行分类;然后,构建基于softmax配电网类型识别方法,利用低压配电网实际数据的降维特征识别配电网类型;此外,利用模型不可知元学习(model-agnostic meta-learning,MAML)方法训练CNN-AE特征提取模型,使CNN-AE模型在少量数据下能自适应提取配电网新负荷特征,最终实现低压配电网准确、快速自适应分类。最后,利用低压配电网实际数据验证了所提方法的有效性。

The classification of low-voltage distribution network is conducive to improving the efficiency of formulating economic operation management measures and new energy planning operation schemes of low-voltage distribution network.With the continuous access of various new sources of energy,charging piles and other new sources,the original load characteristics of the low-voltage distribution network have changed,which on the one hand leads to complex load characteristics of the distribution network,and on the other hand leads to less available load characteristic data after the change,which brings challenges to the classification of the distribution network.Aiming at the above challenges,this paper proposes an adaptive classification method of low-voltage distribution network based on CNN-AE-MAML.Firstly,convolutional neural network auto encoder(CNN-AE)is used to extract the dimensionality reduction features of the distribution load of low-voltage distribution network and the photovoltaic power generation.Spectral clustering(SC)was used to classify low-voltage distribution networks.Then,the distribution network type identification method based on softmax is constructed to identify the distribution network type by using the dimensional-reduction features of the actual data of low-voltage distribution network.In addition,the model agnostic meta-learning(MAML)method is used to train the CNN-AE feature extraction model,so that the CNN-AE model can adaptively extract the new load features of the distribution network under a small amount of data,and finally achieve accurate and fast adaptive classification of the low-voltage distribution network.