基于轻量化Transformer模型的多变量风电功率预测
Multivariate Wind Power Prediction with Lightweight Transformer Model
宋倩 1蓝俊欢2
作者信息
- 1. 河池学院 大数据与计算机学院,广西 河池 546300
- 2. 广西电网有限责任公司河池供电局,广西 河池 546300
- 折叠
摘要
风电功率预测对电力调度和新能源管理极为重要.为准确高效预测多变量风电功率,提出一种基于轻量化Transformer(Light Transformer)的风电功率预测方法.首先,采用滚动序列建模方法,确定输入数据,然后采用Transformer预测模型,改进和精简原始结构,在前馈网络模块中使用GeLU激活函数来代替传统的ReLU激活函数,提升模型的质量,轻量化网络结构,利用多头注意力机制,加快模型训练速度,提升预测模型精度.
Abstract
Wind power prediction is extremely important for power dispatching and new energy management.In order to accurately and efficiently predict the multivariate wind power,a wind power prediction method based on Light Transformer is proposed.Firstly,the rolling sequence modeling method is used to determine the input data,and then the Transformer prediction model is used to improve and streamline the original structure and use the GeLU Activation Function instead of the traditional GeLU Activation Function in the feed-forward network module to improve the quality of the model.It lightens the network structure,utilizes the multi-attention mechanism,accelerates the training speed of the model and improves the accuracy of the prediction model.
关键词
轻量化Transformer模型/风电功率/激活函数Key words
Lightweight Transformer model/wind power/Activation Function引用本文复制引用
基金项目
2022年广西高校中青年教师科研基础能力提升项目(2022KY0606)
2023年河池学院校级科研平台(2023XJPT011)
出版年
2024
国家科技期刊平台
NSTL
万方数据