多算法多模型与在线第二次学习结合的短期电力负荷预测方法

发布时间：2018-02-03 17:57

本文关键词： 短期电力负荷预测多样性采样异构模型多算法多模型在线第二次学习　出处：《计算机应用》2017年11期 　论文类型：期刊论文

【摘要】：为了提高短期电力负荷预测精度,首次提出多算法多模型与在线第二次学习结合的预测方法。首先,利用互信息方法和统计方法对输入变量进行选择;然后,通过Bootstrap方法对数据集进行多样性采样,利用多个不同的人工智能算法和机器学习算法训练得到多个差异化较大的异构预测模型;最后,用每个待预测时刻最近一段时间的实际负荷值、第一次学习生成的多异构预测模型的负荷预测值构成新训练数据集,对新训练数据集进行在线第二次学习,得到最终预测结果。对中国广州市负荷进行预测研究,与最优单模型、单算法多模型和多算法单模型相比,在每日总负荷预测中,全年平均绝对百分误差(MAPE)分别下降了21.07%、7.64%和5.00%,在每日峰值负荷预测中,全年MAPE分别下降了16.02%、7.60%和13.14%。实验结果表明,推荐方法有效地提高了负荷预测精度,有利于智能电网实现节能降耗、调度精细化管理和电网安全预警。
[Abstract]:In order to improve the accuracy of short-term power load forecasting, a forecasting method combining multi-algorithm and multi-model with on-line second learning is proposed for the first time. Firstly, mutual information method and statistical method are used to select input variables. Then, the diversity of data sets is sampled by Bootstrap method, and many heterogeneous prediction models are obtained by using different artificial intelligence algorithms and machine learning algorithms. Finally, using the actual load value of the most recent period of time to be predicted, the load forecasting value of the first learning multi-heterogeneous forecasting model is used to form a new training data set. Through the second online learning of the new training data set, the final prediction results are obtained. The load forecasting of Guangzhou City, China, is compared with the optimal single model, single algorithm multi-model and multi-algorithm single model. In the daily total load forecasting, the average absolute percent error (MAPE) decreased by 21.07% and 5.00%, respectively, in the daily peak load forecasting. The annual MAPE decreased by 16.02% 7.60% and 13.14% respectively. The experimental results show that the recommended method can effectively improve the accuracy of load forecasting and is conducive to the realization of energy saving and consumption reduction in smart grid. Meticulous management of dispatching and early warning of power grid security.
【作者单位】：湖南大学信息科学与工程学院;
【基金】：国家自然科学基金资助项目(61374172) 国家科技成果转化项目(201255)~~
【分类号】：TM715;TP18
【正文快照】： 0引言电力负荷预测是智能电网建设的基本环节之一,精准、时效的负荷预测是智能电网实现节能降耗、调度精细化管理和电网安全预警的基础。由于影响电力负荷变化的因素呈现出较高的随机性,使得电力负荷的变化也具有较高的随机性,加大了预测的难度。已有电力负荷预测模型总体分为

【相似文献】