摘要
光伏发电的波动性和随机性对电力系统安全稳定运行具有不良影响,为提高日前短期光伏功率预测精度进而提升光伏电站运营及电网调度效率,提出一种基于时间卷积神经网络(TCN)结合多头注意力机制(MHA)的光伏功率预测方法。首先TCN接收数据,利用膨胀卷积的结构改变感受野范围,利用因果卷积的设置提取光伏数据的时序特征;经过升维后输入MHA中,选择合适的多头个数,得到多个子空间,将输入特征进行不同维度的空间映射,进一步分配注意力权重;最后降维输入全连接层结合得到的特征信息对次日24 h的光伏功率进行预测。实验在实际光伏场站数据上进行,结果表明,所提模型的预测精度优于对比模型。
The volatility and randomness of photovoltaic power generation have adverse effects on the safe and stable operation of power system.In order to improve the accuracy of short-term photovoltaic power prediction and improve the efficiency of photovoltaic power station operation and power grid dispatching,a photovoltaic power prediction method based on temporal convolutional network(TCN)and multi-head attention(MHA)mechanism is proposed.First,TCN receives the data,uses the structure of inflated convolution to change the range of receptive field,and uses the setting of causal convolution to extract the time series features of photovoltaic data;after upgrading,input into MHA,select the appropriate number of multiple heads,get multiple subspaces,map the input features in different dimensions,and further assign attention weights.Finally,the characteristic information obtained from the reduced-dimensional input full connection layer is used to predict the photovoltaic power for 24 hours the next day.The experiment is carried out on the actual photovoltaic field station data,and the results show that the prediction accuracy of the proposed model is better than that of the comparison model.
出处
《科技创新与应用》
2023年第29期8-12,共5页
Technology Innovation and Application
基金
内蒙古自治区重点研发和成果转化项目(2022YFSJ0033)
内蒙古自治区应用技术研究与开发资金项目(2021GG0046)。
关键词
光伏发电
短期功率预测
深度学习
时间卷积神经网络
注意力机制
photovoltaic power generation
short-term power prediction
deep learning
temporal convolutional network(TCN)
attention mechanism
作者简介
第一作者:孙永叡(1998-),男,硕士。研究方向为深度学习在光伏功率预测领域的应用;通信作者:任晓颖(1979-),女,博士,副教授。研究方向为可再生能源与清洁能源、控制工程。
