改进型CNN-LSTM深度学习神经网络的台区户变拓扑关系识别

Recognition of Topological Relationship Between Substation Transformers and Users Based on Improved CNN-LSTM Deep Learning Neural Network

针对电力台区内各种数据信息繁多复杂、数据处理能力滞后及用户利用率低下等问题,提出一种新型的台区户变拓扑关系识别方法。通过构建卷积神经网络(convolutional neural network,CNN)和长短期记忆(long short-term memory,LSTM)神经网络模型,将台区内配电变压器的有功功率、无功功率、电压值、电流值和用户侧的多种用电数据信息转换为CNN-LSTM深度学习神经网络模型;并在CNN模型中融入LSTM模块,以将台区户变拓扑宏观数据关系转换为微观数据信息识别,大大提高台区户变拓扑关系识别和应用能力。通过设置CNN-LSTM深度学习神经网络不同的层次,计算台区户变拓扑关系。通过算例分析,大大提高了用户识别能力,为台区户变拓扑关系识别提供了技术思路。

A new method for identifying the topological relationship between substation transformers and users was proposed to address the issues of complex and diverse data information,lagging data processing capabilities,and low user utilization in the power substation area.By constructing convolutional neural network(CNN)and long short-term memory(LSTM)neural network models,the active power,reactive power,voltage value,current value,and various electricity consumption data information of the distribution transformer in the substation area were transformed into a CNN-LSTM deep learning neural network model;and the LSTM module was integrated into the CNN model to convert the topological macro data relationship between substation transformers and users into micro data information for recognition,greatly improving the recognition and application capabilities of substation transformers and users topological relationship.By setting different levels of CNN-LSTM deep learning neural network,the topological relationship of substation area and user variables was calculated.Through example analysis,the user recognition ability has been greatly improved.The technical ideas for topology relationship recognition of substation transformers and users is provided.