融合框架下的电力工程数据特征提取与评估方法

Feature extraction and evaluation method of power engineering data under the framework of fusion

针对现有电力工程评估校核方法数据处理效率低、智能化与信息化程度不足的问题,提出了一种基于多层感知机(MLP)、门控循环单元(GRU)和图卷积神经网络(GCN)的多任务融合数据评估模型。该模型在对工程数据进行预处理的基础上,利用MLP、GRU和GCN从多元数据中提取深层特征。在自适应权重的多任务学习模型中引入张量融合方法,完成数据信息的特征级融合,再经共享层与输出层的全连接处理得到评估结果。实验结果表明,所提模型评估结果的均方根误差为0.035,平均绝对值误差为0.014,决定系数为0.993,均优于现有特征融合数据处理方法。

A multi-task fusion data evaluation model based on Multi-layer Perceptron(MLP),Gated Recurrent Unit(GRU),and Graph Convolutional Networks(GCN)is proposed to address the issues of low data processing efficiency and insufficient intelligence and informatization of existing power engineering evaluation and verification methods.On the basis of pre-processing engineering data,this model utilizes MLP,GRU,and GCN to extract deep features from multivariate data.The tensor fusion method is introduced into the multi-task learning model with adaptive weights to achieve feature level fusion of data information,and then the evaluation results are obtained through full connection processing between the shared layer and the output layer.The experiment results show that the root mean square error of the evaluation results of the proposed model is 0.035,the average absolute error is 0.014,and the determination coefficient is 0.993,all of which are superior to existing feature fusion data processing methods.