基于改进长短期记忆神经网络的光伏功率预测方法研究
A Modified LSTM-based Method to Predict Photovoltaic Power
卢利军 1王军 1王浩 1张义坤 2何英杰2
作者信息
- 1. 河南许继仪表有限公司,河南许昌 461111
- 2. 西安交通大学电气工程学院,陕西西安 710043
- 折叠
摘要
太阳能光伏输出功率的精确预测对于电网的安全运行非常重要,并且可以降低光伏系统的运营费用.为了通过使用光伏系统历史的性能数据预测未来几天光伏系统的输出功率,提出了一种基于集成长短期记忆网络(Ensemble Long Short-Term Memory Network,Ensemble LSTM)集成模型的方法来预测未来几天的光伏输出功率.为验证该方法的有效性,使用实测太阳能光伏数据进行实验,结果证明与单个LSTM模型对比,所提出的LSTM集成模型非常可靠,并且在输出功率预测准确性方面具有显著优势.
Abstract
Accurate prediction of solar photovoltaic output power is crucial for grids'secure operation and conducive to PV system operational cost reduction.The primary objective of this work was to predict the output power of PV systems for the upcoming days using past performance data.In this regard,a method based on an ensemble Long Short-Term Memory Network(Ensemble LSTM)model was established and validated through an empirical study using practically collected so-lar PV data,whose results demonstrated that the method seems highly reliable compared to individual LSTM models and obviously superior in terms of accuracy for output power prediction.
关键词
长短期记忆网络/光伏功率预测/机器学习Key words
LSTM/PV power prediction/machine learning引用本文复制引用
出版年
2024
NETL
NSTL
万方数据