«上一篇/Previous Article|本期目录/Table of Contents|下一篇/Next Article»

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:

期数:

2025年07期

页码:

125-132

栏目:

人工智能

出版日期:

2025-07-10

文章信息/Info

Title:

Short-term PV Power Prediction Based on IDBO-TCN-LSTM

文章编号:

1673-629X(2025)07-0125-08

作者:

周新宇; 姜志航; 白峻铭; 梁宏涛

青岛科技大学 信息科学技术学院,山东 青岛 266011

Author(s):

ZHOU Xin-yu; JIANG Zhi-hang; BAI Jun-ming; LIANG Hong-tao

School of Information Science and Technology,Qingdao University of Science and Technology,Qingdao 266011,China

关键词:

蜣螂优化算法; Tent映射; Levy飞行; 变异策略; 光伏功率预测; 深度学习

Keywords:

dung beetle optimization algorithm; Tent mapping; Levy flight; variational strategy; photovoltaic power prediction; deep learn-ing

分类号:

TP305

DOI:

10.20165/j.cnki.ISSN1673-629X.2025.0044

摘要:

为了解决光伏功率预测中存在的不稳定性和波动性问题,以及长短期记忆网络(LSTM)在参数确定方面的困难,该文提出了一种基于改进蜣螂优化算法(IDBO)、时间卷积网络(TCN)和长短期记忆网络相结合的短期光伏功率组合预测模型。 首先,利用 Tent 混沌反向学习策略对蜣螂算法的种群进行初始化,以提高算法的探索能力和多样性;引入莱维飞行和动态权重因子更新解决原始算法收敛慢和易陷入局部最优的问题;引入高斯-柯西变异对最优个体进行扰动,增强全局搜索能力。其次,选择获得感受野更大的 TCN 作为特征提取层,以捕捉更加复杂的时间序列模式。 最后,采用改进蜣螂优化算法优化 LSTM 网络模型的参数设置,建立 IDBO-TCN-LSTM 组合预测模型并在澳大利亚光伏数据集上进行仿真实验。对比结果显示,该模型在处理短期光伏预测任务的预测精度上优于其他模型。

Abstract:

In order to solve the instability and volatility problems in PV power prediction, as well as the difficulties in parameter determination of long-short-term memory networks (LSTM),we propose a short-term PV power portfolio prediction model based on the combination of the improved dung beetle optimization algorithm (IDBO),temporal convolutional network (TCN) and long-short-term memory networks. Firstly,the population of the dung beetle algorithm is initialized using the Tent chaotic inverse learning strategy to improve the exploratory ability and diversity of the algorithm; Levy flights and dynamic weight factor updating are introduced to solve the shortcomings of the original algorithm,which is slow to converge and prone to fall into local optimums; Gaussian-Cauchy variation is introduced to perturb the optimal individuals and enhance the global search ability. Secondly,the TCN,which obtains a larger sensory field,is selected as the feature extraction layer to capture more complex time series patterns. Finally, the improved dung beetle optimization algorithm is used to optimize the parameter settings of the LSTM network model,and the combined IDBO-TCN-LSTM pre-diction model is established and simulation experiments are carried out on an Australian photovoltaic ( PV) dataset. The comparative results show that the prediction accuracy of the proposed model is better than that of other models in dealing with short - term PV prediction tasks.

常用功能

更新日期/Last Update: 2025-07-10