基于数据挖掘和LSSVM的电量大数据多维感知方法

Multidimensional Perception Method for Power Big Data Based on Data Mining and LSSVM

为了解决现有电量大数据多维感知方法存在感知性能差的问题,利用数据挖掘技术和LSSVM算法实现对感知方法的优化设计。利用数据挖掘技术构建多维电量数据模型,通过填补缺失数据、修正错误数据和数据标准化等3个步骤,从趋势性、变动性及负荷等3个方面提取电量大数据特征;利用LSSVM算法预测多维电负荷变化量,进行电量大数据多维感知。实验结果表明,所提方法的拟合误差降低了8.4 kW,敏感度指数提高了0.388,电量大数据多维感知性能较优。

In order to solve the problem of poor perception performance of the existing multidimensional perception methods of power big data,the optimal design of perception methods is realized by data mining technology and LSSVM algorithm.The multidimensional electricity data model is constructed by data mining technology.Through three steps of filling in missing data,correcting wrong data and data standardization,the characteristics of electricity big data are extracted from three aspects of trend,variability and load.The multidimensional electricity load variation is predicted by LSSVM algorithm,and the multidimensional perception of electricity big data is carried out.The experimental results show that the fitting error of this method is reduced by 8.4 kW,the sensitivity index is increased by 0.388,and the multidimensional perception performance of electricity big data is better.