基于原子稀疏分解理论的短期风电功率滑动预测  

Short-term wind power forecasting based on atomic sparse decomposition theory

作　　者：崔明建[1] 孙元章[1] 柯德平[1] 王树鹏[2] 

机构地区：[1]武汉大学电气工程学院,湖北武汉430072 [2]中国地质大学数理学院,湖北武汉430074

出　　处：《电力自动化设备》2014年第1期120-127,共8页Electric Power Automation Equipment

基　　金：国家重点基础研究发展计划(973计划)资助项目(2012-CB215101)~~

摘　　要：采用一种具有很强的非平稳信号跟踪、预测能力的原子稀疏分解(ASD)法,作为人工神经网络(ANN)的前置分解方法,将风电功率序列分解为原子分量和残差分量,对原子分量进行自预测,残差分量进行ANN预测,再通过追加最新的风电功率实时数据来更新ASD的结果,进而滑动预测下一个时刻的风电功率。以实际风电场数据进行验证,结果证明了该模型可以有效地处理风电功率非平稳性,产生更为稀疏的分解效果,显著地降低了绝对平均误差、均方根误差计算值的统计区间。ASD（Atomic Sparse Decomposition）,which has excellent ability to track and forecast unstable signal,is applied as the pre-decomposition of ANN（Artificial Neural Network） to decompose the wind power series into atomic component and residual component. The former is self-forecasted while the latter is forecasted by ANN. The latest real-time data of wind power are added to update the result of ASD for forecasting the wind power of next instant. The model is verified by the practical data of a wind farm, which shows that,the instability of wind power is effectively dealt with to produce more sparse decomposition effect,significantly reducing the statistical intervals of absolute mean error and root mean square error.

关 键 词：风电 预测 原子稀疏分解 人工神经网络 模型 

分 类 号：TM614[电气工程—电力系统及自动化]