基于云计算和极限学习机的分布式电力负荷预测算法
发布时间：2018-07-14 10:42
【摘要】：为了提高电力负荷预测精度,应对电力系统智能化所带来的数据海量化高维化带来的单机计算资源不足的挑战,提出了一种在线序列优化的极限学习机短期电力负荷预测模型。针对电力负荷数据特性,对极限学习机预测算法进行在线序列优化;引入分布式和multi-agent思想,提升负荷预测算法预测准确率;采用云计算的MapReduce编程框架对提出的算法模型进行并行化改进,提高其处理海量高维数据的能力。选用EUNITE提供的真实电力负荷数据进行算例分析,在32节点云计算集群上进行实验,结果表明基于该模型的负荷预测精度均优于传统支持向量回归预测算法和泛化神经网络预测算法,且提出的算法具有优异的并行性能。
[Abstract]:In order to improve the precision of power load forecasting and to meet the challenge of the shortage of single computer computing resources brought by the large amount of data brought by the intelligentization of the power system, a short-term power load forecasting model of the online sequence optimized limit learning machine is put forward. The limit learning machine prediction algorithm is online based on the characteristics of the power load data. Sequence optimization; introducing the distributed and multi-agent ideas to improve the prediction accuracy of the load forecasting algorithm; using the MapReduce programming framework of the cloud computing to improve the proposed algorithm model and improve its ability to deal with massive and high dimensional data. Use the real electrical load data provided by EUNITE to carry out an example analysis, in the 32 node cloud meter. The experimental results show that the load forecasting accuracy based on this model is better than the traditional support vector regression prediction algorithm and the generalization neural network prediction algorithm, and the proposed algorithm has excellent parallel performance.
【作者单位】： 华北电力大学控制与计算机工程学院;
【基金】：河北省科学研究项目(Z2012077,Z2010290)
【分类号】：TM715.1
[Abstract]:In order to improve the precision of power load forecasting and to meet the challenge of the shortage of single computer computing resources brought by the large amount of data brought by the intelligentization of the power system, a short-term power load forecasting model of the online sequence optimized limit learning machine is put forward. The limit learning machine prediction algorithm is online based on the characteristics of the power load data. Sequence optimization; introducing the distributed and multi-agent ideas to improve the prediction accuracy of the load forecasting algorithm; using the MapReduce programming framework of the cloud computing to improve the proposed algorithm model and improve its ability to deal with massive and high dimensional data. Use the real electrical load data provided by EUNITE to carry out an example analysis, in the 32 node cloud meter. The experimental results show that the load forecasting accuracy based on this model is better than the traditional support vector regression prediction algorithm and the generalization neural network prediction algorithm, and the proposed algorithm has excellent parallel performance.
【作者单位】： 华北电力大学控制与计算机工程学院;
【基金】：河北省科学研究项目(Z2012077,Z2010290)
【分类号】：TM715.1
【参考文献】
相关期刊论文 前10条
1 王奔;冷北雪;张喜海;单纕，
本文编号：2121390
本文链接：https://www.wllwen.com/kejilunwen/dianlilw/2121390.html
