知识导航
资源分类

基于泄漏积分型回声状态网络的在线学习光伏功率预测

基于泄漏积分型回声状态网络的在线学习光伏功率预测

Online-learning PV Power Forecasting Based on Leaky-integrator Echo State Network

Online-learning PV Power Forecasting Based on Leaky-integrator Echo State Network

doi:
10.3969/j.issn.1003-8930.2018.02.001
摘要:
为提高光伏功率预测结果的准确性,提出了基于泄漏积分型回声状态网络LIESN(leaky-integrator echo state network)的具有在线学习功能的预测方法.预测模型中采用泄漏积分神经元增强储备池的短期记忆能力,通过最小二乘在线学习算法增加临近时间样本对权值的影响;综合考虑预测精度与运行时间,分析了LIESN关键参数对预测性能的影响,并提出了LIESN关键参数的设定方法.实例证明,在线学习LIESN的预测精度优于BP神经网络、经典ESN及离线学习LIESN模型,测试结果的归一化均方根误差达到0.0986,验证了方法的有效性.
Abstract:
To enhance the accuracy of photovoltaic(PV)power forecasting,an online-learning method based on leaky-integrator echo state network(LIESN)is presented.In the forecasting model,leaky-integrator neurons are introduced to improve the short-term memory ability of the reservoir,and least square online-learning method is used to increase the influence of adjacent time samples on weights.With the comprehensive consideration of forecasting accuracy and run?ning time,the influences of key parameters of LIESN on the forecasting performance are analyzed,and a method is pro?posed to set the key parameters of LIESN.Practical examples indicate that the forecasting accuracy of online-learning LIESN was superior to those of BP neural network,plain ESN and offline-learning LIESN models,and the normalized root mean square error of the test result reached 0.0986,which verifies the validity of the proposed method.
作者 徐正阳 路志英 刘洪
Author: XU Zhengyang LU Zhiying LIU Hong
作者单位 天津大学智能电网教育部重点实验室,天津,300072
期 刊: 电力系统及其自动化学报 ISTIC EI SCI PKU CSSCI
Journal: Proceedings of the CSU-EPSA
年,卷(期) 2018, 30(2)
分类号 TM315
关键词: 回声状态网络 知识脉络 泄漏积分 知识脉络 光伏功率预测 知识脉络 在线学习 知识脉络
Keywords: echo state network leaky-integrator PV power forecasting online learning
机标分类号
基金项目 国家重大科学仪器设备开发专项资助项目,国家自然科学基金资助项目
参考文献和引证文献
返回顶部参考文献
返回顶部引证文献
返回顶部本文读者也读过
返回顶部相似文献
返回顶部相关博文

知识产权声明| 服务承诺| 联系我们| 人才招聘| 客户服务| 关于我们

京ICP证:010071  互联网出版许可证:新出网证(京)字042号  京公网安备11010802020237号
万方数据知识服务平台--国家科技支撑计划资助项目(编号:2006BAH03B01)©北京万方数据股份有限公司  万方数据电子出版社