基于时间卷积网络和门控循环单元的短期用电量预测方法

Short-term Electricity Consumption Forecasting Method Based on Temporal Convolutional Network and Gated Recurrent Unit

针对智能电网建设环境下用电数据所呈现出的采集频率低、时变性显著等特点,提出了一种基于时间卷积网络和门控循环单元的短期用电量预测方法。考虑电类特征、环境特征和时间特征,从常见用户用电量的影响因素中筛选出模型的输入数据,分别训练时间卷积网络和门控循环单元两种深度学习模型并建立所提方法的整体架构。对某地区低采集频率用电数据进行仿真分析,与传统的长短期记忆网络、一维卷积及多层感知机等方法相比,所提方法具有更高的预测精度,有效可行。

Aiming at the low acquisition frequency and significant time-varying characteristics of electricity consumption data in the smart grid construction environment,a short-term electricity consumption prediction method is proposed based on time convolutional network and gated recurrent unit network.Considering electrical characteristics,environmental characteristics and time characteristics,the input data of the model is selected from the influencing factors of common user power consumption,and two deep learning models of time convolutional network and gated recurrent unit network are trained separately.Thus,the proposed method is established.The simulation analysis of low-frequency electricity consumption data in a certain area shows that the proposed method has higher prediction accuracy and is effective and feasible compared with traditional long-short-term memory networks,one-dimensional convolution and multilayer perceptrons.