基于改进相似日优化HBA-BiLSTM-KELM的光伏发电功率预测
RESEARCH ON PV SYSTEM POWER PREDICTION BASED ON IMPROVED SIMILAR DAY AND HBA-BiLSTM-KELM NEURAL NETWORK
李超然 1潘鹏程 1杨伟荣 2徐恒山 1魏业文1
作者信息
- 1. 新能源微电网湖北省协同创新中心,宜昌 443002;三峡大学电气与新能源学院,宜昌 443002
- 2. 云南电网有限责任公司曲靖供电局,曲靖 655099
- 折叠
摘要
为提高光伏发电系统输出功率的预测精度,提出基于改进相似日和蜜獾算法(HBA)优化改进双向长短期记忆神经网络(BiLSTM)与核极限学习机(KELM)的光伏发电预测方法.首先,使用CRITIC权重法动态计算各气象因素对光伏发电功率的影响权重,通过逐时刻计算历史日和待预测日的加权欧氏距离确定相似日.其次,使用HBA优化BiLSTM和KELM模型参数,然后使用HBA参数优化后的BiLSTM进行功率预测,优化后的KELM进行误差优化预测.最后将初步预测功率和误差预测功率相加得到最终预测功率.仿真结果表明:该模型平均绝对百分比误差为0.91%,具有较高的光伏系统输出功率预测精度.
Abstract
In order to improve the output power prediction accuracy of PV generation system,this paper proposed a prediction model based on improved similar days and honey badger algorithm to improve bidirectional long-short term memory neural network and kernel extreme learning machine.Firstly,The CRITIC weight method is used to dynamically calculate the influence weight of each meteorological factor on photovoltaic power generation.The similar days are determined by calculating the weighted Euclidean distance between the historical days and the predicted days.Secondly,HBA is used to optimize the parameters of BiLSTM and KELM models,and then the BiLSTM optimized by HBA parameters is used for power prediction,and the optimized KELM is used for error optimization prediction.Finally,the preliminary prediction power and the error prediction power are added to obtain the final prediction power.The simulation results show that the average absolute percentage error of the model is 0.91%,which has high prediction accuracy of output power of photovoltaic system.
关键词
光伏发电/功率预测/神经网络/核极限学习机/蜜獾算法Key words
PV power generation/power forecasting/neural network/kernel extreme learning machine/honey badger algorithm引用本文复制引用
基金项目
国家水运安全工程技术研究中心开放基金(B2022002)
宜昌市自然科学基金(A22-3-008)
出版年
2024
NETL
NSTL
万方数据