数据与计算发展前沿 ›› 2021, Vol. 3 ›› Issue (5): 109-117.

doi: 10.11871/jfdc.issn.2096-742X.2021.05.008

• 技术与应用 • 上一篇    下一篇

基于分时LSTM的光伏发电功率预测

许可1,2(),王珏1,*(),姚铁锤1,2(),李凯1()   

  1. 1.中国科学院计算机网络信息中心,北京 100190
    2.中国科学院大学,北京 100049
  • 收稿日期:2021-03-16 出版日期:2021-10-20 发布日期:2021-11-24
  • 通讯作者: 王珏
  • 作者简介:许可,中国科学院计算机网络信息中心,硕士研究生,主要研究方向为人工智能算法,光伏发电功率预测。
    本文承担工作为：时间相关性分析和模型算法实现。
    XU Ke is a master student at Computer Network Information Center of the Chinese Academy of Scien-ces. Her main research interests are artificial intelligence algorithms and time series data prediction.
    In this paper she undertakes the following tasks: time corre-lation analysis and model algorithm realization.
    E-mail: xuke@cnic.cn|王珏,中国科学院计算机网络信息中心,副研究员,博士,主要研究方向为人工智能算法与应用软件。
    本文承担工作为：光伏功率预测模型设计。
    WANG Jue, Ph.D., is an associate rese-arch fellow of Computer Network Information Center of the Chinese Academy of Sciences. His main research interests are artificial intelligence algorithms and application software.
    In this paper he undertakes the following tasks: design of the photovoltaic power forecasting model.
    E-mail: wangjue@sccas.cn|姚铁锤,中国科学院计算机网络信息中心,博士研究生,主要研究方向为人工智能。
    本文承担工作为：系统实现。
    YAO Tiechui is a Ph.D. student at Com-puter Network Information Center of the Chinese Academy of Sciences. His main research interest is artificial intelligence.
    In this paper he undertakes the following tasks: model algor-ithm realization.
    E-mail: yaotiechui@cnic.cn|李凯, 中国科学院计算机网络信息中心,工程师,硕士,主要研究方向为人工智能平台与应用软件。
    本文承担工作为：光伏功率预测模型调优。
    LI Kai, master, is an engineer at Com-puter Network Information Center of the Chinese Academy of Sciences. His main research interests are artificial intelligence platforms and application software.
    In this paper he undertakes the following tasks: optimization of the photovoltaic power forecasting model.
    E-mail: kai.li@cnic.cn
  • 基金资助:
    科技部重大科技项目(2020AAA0105202);中国科学院战略性先导科技专项(A类)(XDA27000000)

Photovoltaic Power Forecast Based on Time-division LSTM

XU Ke1,2(),WANG Jue1,*(),YAO Tiechui1,2(),LI Kai1()   

  1. 1. Computer Network Information Center, Chinese Academy of Sciences, Beijing 100190, China
    2. University of Chinese Academy of Sciences, Beijing 100049, China
  • Received:2021-03-16 Online:2021-10-20 Published:2021-11-24
  • Contact: WANG Jue

RichHTML

45

PDF

285

摘要：

【目的】准确的预测光伏发电功率对电网调度具有十分重要的意义。【方法】光伏发电是一个连续不断的过程,光伏发电功率每时刻的变化取决于当前时刻和历史时刻的气象特征。本文考虑光伏发电的时间相关性,提出了基于分时长短期记忆神经网络（Long Short-Term Memory,LSTM）的光伏发电功率预测模型,利用Pearson相关系数对光伏发电功率的时序特征以及气象特征进行研究,选择与光伏发电功率高度相关的历史气象特征作为光伏发电功率预测模型的输入。【结果】对真实光伏电站进行案例分析,基于分时LSTM的光伏发电功率预测可以有效提高光伏发电功率预测精度,并适应光伏时序数据时间不对齐的特点。【结论】本文所提出的模型与传统LSTM模型相比具有更低的预测误差。

关键词: 分时LSTM, Pearson相关系数, 光伏功率预测

Abstract:

[Objective] An accurate forecast of photovoltaic (PV) power has great significance for power dispatch. [Methods] PV generation is a continuous process. The variation of PV power at each moment is affected by meteorological factors at both current and historical moments. For the time correlation of PV power, we propose a PV power forecasting model based on time-division long short-term memory (LSTM) in this paper. The Pearson correlation coefficient is used to study the series characteristics and meteorological characteristics of PV power. And the meteorological characteristics highly related to the PV power at historical moments are selected as the input of the PV power forecasting model. [Results] Based on the real-world datasets, the PV power forecasting model based on time-division LSTM can adapt to the PV series time misalignment and therefore achieve a high precision rate. [Conclusions] The proposed model has lower forecasting error than the traditional LSTM model.

Key words: time-division LSTM, Pearson correlation coefficient, photovoltaic power forecast