基于注意力机制的CNN-BIGRU短期电价预测  

CNN-BIGRU Short-term Electricity Price Prediction Based on Attention Mechanism

作　　者：杨超 冉启武[1] 罗德虎 豆旺 YANG Chao;RAN Qiwu;LUO Dehu;DOU Wang(School of Electrical Engineering,Shaanxi University of Technology,Hanzhong 723001,China)

机构地区：[1]陕西理工大学电气工程学院,汉中723001

出　　处：《电力系统及其自动化学报》2024年第3期22-29,共8页Proceedings of the CSU-EPSA

基　　金：陕西省自然科学基金资助项目(17JK0139);陕西省自然科学基础研究计划资助项目(2023-JC-YB-442)。

摘　　要：针对短期电价预测的复杂性和精确度较差的问题,本文提出一种基于注意力机制的卷积神经网络和双向门控循环单元网络的短期电价预测模型。该模型将历史电价数据经过数据预处理后作为输入,首先利用卷积神经网络提取历史电价序列中的特征;其次,将提取的特征向量构造成时间序列输入到双向门控循环单元网络,充分挖掘特征内部的变化规律进行训练;然后,引入注意力机制来突出重要信息的影响并赋予权重,利用注意力机制对双向门控循环单元网络每个时间步的输出进行加权求和;最后,在全连接层通过激活函数计算输出最终预测值。通过实例验证了本文所提模型的准确性。Aimed at the problems of complexity and low accuracy with short-term electricity price prediction,a shortterm electricity price prediction model is proposed in this paper,which is based on the Attention mechanism of convolutional neural network(CNN)and bidirectional gated recurrent unit(BIGRU)network.This model takes the historical data of electricity price after data preprocessing as input.First,the features in the historical electricity price series are extracted by using CNN.Second,the extracted feature vectors are constructed into a time series,which is input into the BIGRU network to fully explore the internal variation rules of features for training.Third,the Attention mechanism is introduced to highlight the influence of important information and give the corresponding weight,and it is used to perform weighted summation on the output at each time step of the BIGRU network.Fourth,the final predicted value is calculated by an activation function in the fully-connected layer.The accuracy of the proposed model is verified by examples.

关 键 词：电价预测 注意力机制 卷积神经网络 双向门控循环单元网络 

分 类 号：TM73[电气工程—电力系统及自动化]