摘要
针对传统方法难以实现对电力系统异常电量的高效、准确辨识的问题,提出一种基于自回归滑动平均模型(ARMA)和K-means聚类的电量数据异常识别方法。在分析电量数据中的趋势性、周期性和季节性特征的基础上,首先利用历史数据建立ARMA模型进行用电量预测模型训练,并通过极值点步进线性回归策略逐步提取出线性化的旋转分量,提高旋转分量的生成效率。然后计算预测值和真实值之间的残差,最后对残差项进行DBSCAN聚类,实现电量异常数据的识别。对某电网20个区域的用电量数据进行案例分析,并与常见异常识别方法进行对比,通过检测率和误报率评价指标验证了该研究方法的有效性。
It is difficult to use traditional methods to efficiently and accurately identify abnormal quantity of power system.In view of the problem,this paper proposes a method of identifying abnormal power data based on autoregressive moving average model(ARMA)and K-means clustering.On the basis of analyzing the tendency,periodicity and seasonal character of the electric quantity data,the paper firstly establishes the ARMA model with historical data to conduct training on electricity consumption prediction model,and uses the extreme point stepwise linear regression strategy to gradually extract the rotation component of linearization,so as to improve the generation efficiency of rotating component.Then,the paper calculates the residual between the predicted value and the real value.Finally,the paper performs DBSCAN clustering for residual items to realize the recognition of abnormal data of electric quantity.The paper also makes a case study on the data of electricity consumption in 20 regions of a power grid,comparing with the common anomaly identification method,from which the evaluation indexes of detection rate and false alarm rate verify the effectiveness of the proposed research method.
作者
梁捷
梁广明
LIANG Jie;LIANG Guangming(Measurement Center,Guangxi Power Grid Co.,Ltd.,Guangxi Nanning 530023;Nanning Baihui Pharmaceutical Group Co.,Ltd.,Guangxi Nanning 530003)
出处
《湖北电力》
2019年第4期65-70,共6页
Hubei Electric Power
基金
广西电网有限责任公司重点科技项目(项目编号：GXKJXM20181236)
