一种基于机器学习的电力能耗异常检测与预测的方法

A Method of Abnormal Detection and Prediction of Power Consumption Based on Machine Learning

在预处理电力能耗数据时,现有的电力数据异常检测与预测方法没有补全由删除或修正数据导致的缺失项,致使时间序列中出现明显的错漏。为了提高电力能耗异常检测与预测的准确性,基于机器学习设计一种电力能耗异常检测与预测方法。在数据的预处理过程中,通过清洗、转换与提取电力能耗数据,补全时间序列中的缺失值,保证被删除与修正的数据不会影响到整体的数据处理,在此基础上设计基于机器学习的电力能耗异常检测算法以及电力能耗预测算法。选择春、夏、秋、冬4个季节的电力数据,比较机器学习方法与其他3种方法在电力能耗异常数据检测与预测的结果。由实验结果可知,所提方法的ROC值在不同的时间中均大于其他算法,且其预测结果的RMSE与MAE误差指标均小于其他算法,可见所提方法的预测结果准确性高于其他算法。

The existing power data anomaly detection and prediction methods do not complete the missing items caused by deleting or correcting the data in preprocessing the power energy consumption data,resulting in obvious errors and omissions in the time series.In order to improve the accuracy of power energy consumption anomaly detection and prediction,the power energy consumption anomaly detection and prediction method is designed based on machine learning.In the process of data preprocessing,we clean,convert and extract the power energy consumption data,complete the missing values in the time series,and ensure that the deleted and corrected data will not affect the overall data processing.On this basis,the power energy consumption anomaly detection algorithm and power energy consumption prediction algorithm based on machine learning are designed.The power data in spring,summer,autumn and winter are selected,and the detection and prediction results of abnormal power consumption data by machine learning method and other three methods are compared.The experimental results show that the ROC value of the proposed method is greater than those of other algorithms in different times,and the RMSE and MAE error indexes of the prediction results are less than those of other algorithms.It can be seen that the accuracy of the prediction results of the proposed method is greater than other algorithms.