摘要
探讨了一种基于多输入的深度神经网络模型用于电力设备异常检测的方法,模型采用卷积神经网络和循环神经网络来处理不同传感器采集的图像和时间序列数据,并通过注意力机制动态调整不同数据源的权重。结合案例在电力数据集上进行实验,该模型的表现优于传统的异常检测方法,并且能够检测到多种类型的异常事件。该方法可以应用于电力设备的故障预测、智能维护等领域,但在处理大规模数据时,存在计算复杂度高和训练时间长等问题,可以在提高计算效率的同时进一步优化模型的性能和健壮性。
The article discusses a multi-input based deep neural network model for power equipment anomaly detection,the model uses convolutional neural network and recurrent neural network to process image and time series data collected by different sensors,and dynamically adjusts the weights of different data sources through the attention mechanism.Experiments are conducted on power datasets with case studies,and the model outperforms traditional anomaly detection methods and is able to detect multiple types of anomalous events.The method can be applied in the fields of fault prediction and intelligent maintenance of electric power equipment,but when dealing with large-scale data,there are problems such as high computational complexity and long training time,which can further optimize the performance and robustness of the model while improving computational efficiency.
作者
彭超锋
PENG Chaofeng(State Grid Shaanxi Electric Power Co.,Ltd.,Ultra High Voltage Company,Xi’an 710026,China)
出处
《通信电源技术》
2023年第21期64-66,共3页
Telecom Power Technology
关键词
多源数据融合
变电站
数据采集
卷积神经网络
循环神经网络
multi-source data fusion
substation
data acquisition
convolutional neural network
recurrent neural network
