发电技术 ›› 2024, Vol. 45 ›› Issue (4): 684-695.DOI: 10.12096/j.2096-4528.pgt.23045

• 新能源 • 上一篇    下一篇

一种新型分布式光伏出力区间预测方法

杨康1, 李蓝青1, 李艺丰1, 宋东阔2, 王博仑1, 陈金2, 周霞3, 单宇3   

  1. 1.国网江苏省电力有限公司,江苏省 南京市 210008
    2.国电南瑞科技股份有限公司,江苏省 南京市 211000
    3.南京邮电大学自动化学院/人工智能学院,江苏省 南京市 210023
  • 收稿日期:2023-04-12 修回日期:2023-08-25 出版日期:2024-08-31 发布日期:2024-08-27
  • 通讯作者: 周霞
  • 作者简介:杨康(1990),男,硕士,工程师,主要研究方向为电力电子,yangkangtc@163.com
    周霞(1978),女,博士,副教授,从事电力通信、电力系统分析与控制研究,本文通信作者,zhouxia@njupt.edu.cn
  • 基金资助:
    国家自然科学基金项目(61933005)

A Novel Distributed Photovoltaic Output Interval Prediction Method

Kang YANG1, Lanqing LI1, Yifeng LI1, Dongkuo SONG2, Bolun WANG1, Jin CHEN2, Xia ZHOU3, Yu SHAN3   

  1. 1.State Grid Jiangsu Electric Power Co. , Ltd. , Nanjing 210008, Jiangsu Province, China
    2.Guodian NARI Technology Co. , Ltd. , Nanjing 211000, Jiangsu Province, China
    3.College of Automation/College of Artificial Intelligence, Nanjing University of Posts and Telecommunications, Nanjing 210023, Jiangsu Province, China
  • Received:2023-04-12 Revised:2023-08-25 Published:2024-08-31 Online:2024-08-27
  • Contact: Xia ZHOU
  • Supported by:
    National Natural Science Foundation of China(61933005)

摘要:

目的 分布式光伏功率预测对光伏电站运行和调度具有重要意义,针对点预测方法难以全面描绘分布式光伏功率不确定性的问题,提出了一种基于自适应噪声完备集合经验模态分解(complete ensemble empirical mode decomposition with adaptive noise,CEEMDAN)和麻雀搜索算法优化的最小二乘支持向量机(sparrow search algorithm optimized least square support vector machine,SSA-LSSVM)分布式光伏功率区间预测模型。 方法 首先,通过CEEMDAN将光伏功率序列分解为多个模态分量,再对一次分解得到的高频非平稳分量进行二次分解;其次,采用样本熵(sample entropy,SE)将所有分量重构为趋势分量和振荡分量;然后,通过SSA-LSSVM得到2个分量的点预测值;最后,对振荡分量的点预测误差进行概率密度估计,叠加点预测值得到总体的预测区间结果。 结果 所提区间预测模型具有更高的区间覆盖率且区间平均宽度更窄。 结论 在分布式光伏功率数据处理中加入二次模态分解,再结合样本熵对其子序列进行重构,可有效降低原始预测分量的复杂程度,同时提升模型预测准确性。

关键词: 分布式光伏, 功率预测, 核密度估计, 区间预测

Abstract:

Objectives Distributed photovoltaic power prediction is of great significance for the operation and scheduling of photovoltaic power plants. Point prediction methods are difficult to comprehensively describe the uncertainty of distributed photovoltaic power. This article proposed a distributed photovoltaic power interval prediction model based on complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) and sparrow search algorithm optimized least squares support vector machine (SSA-LSSVM). Methods Firstly, the photovoltaic sequence was broken down into multimodal components through CEEMDAN, and then the high-frequency non-stationary components obtained from the first decomposition were decomposed twice. Secondly, sample entropy (SE) was used to reconstruct all components into trend and oscillation components. Then, the point prediction values of the two components were obtained through SSA-LSSVM. Finally, the probability density estimation was performed on the point prediction error of the oscillation component, and the stacked point prediction value was used to obtain the overall prediction interval result. Results The interval prediction model proposed in this paper has higher interval coverage and narrower average interval width. Conclusions Adding secondary modal decomposition to distributed photovoltaic power data processing and combining sample entropy to reconstruct its sub-sequences can effectively reduce the complexity of the original prediction components and improve the accuracy of model prediction.

Key words: distributed photovoltaic, power prediction, kernel density estimation, intervel prediction

中图分类号: