引用本文: | 范杏蕊,李元诚.基于改进Autoformer模型的短期电力负荷预测[J].电力自动化设备,2024,44(4):171-177 |
| FAN Xingrui,LI Yuancheng.Short-term power load forecasting based on improved Autoformer model[J].Electric Power Automation Equipment,2024,44(4):171-177 |
|
摘要: |
针对短期电力负荷预测因受天气、温度、节假日等多重不确定性因素影响而造成精度低的问题,提出一种基于改进Autoformer模型的短期电力负荷预测模型。改变序列分解预处理的惯例,设计深度模型的内部分解模块,该模块提取模型中隐藏状态的内在复杂时序趋势,使得模型具有复杂时间序列的渐进分解能力;提出Nystrom自注意力机制,该机制利用Nystrom方法来逼近标准的自注意力机制。某地电力负荷预测实验结果表明,所提模型比基于标准Autoformer模型的短期电力负荷预测模型的时间复杂度更低,准确率更高。 |
关键词: 短期电力负荷预测 时序分解模块 Nystrom自注意力机制 Sdformer模型 |
DOI:10.16081/j.epae.202305011 |
分类号:TM715 |
基金项目: |
|
Short-term power load forecasting based on improved Autoformer model |
FAN Xingrui, LI Yuancheng
|
College of Control and Computer Engineering, North China Electric Power University, Beijing 102206, China
|
Abstract: |
Aiming at the low accuracy problem of short-term power load forecasting caused by the influence of multiple uncertain factors such as weather, temperature and holiday, a short-term power load forecasting model based on an improved Autoformer model is proposed. By changing the pre-processing convention of sequence decomposition, an internal decomposition module of depth model is designed, which extracts the intrinsically complex time series trend of hidden state in the model, and makes the model have the ability to decompose complex time series asymptotically. The Nystrom self-Attention mechanism is proposed, which uses the Nystrom method to approximate the standard self-Attention mechanism. The experimental results of power load forecasting in a region show that the proposed model has lower time complexity and higher accuracy than the standard Autoformer model. |
Key words: short-term power load forecasting timing decomposition module Nystrom self-Attention mechanism Sdformer model |