摘要
针对工业控制过程中高频时间序列是否存在某些关键特征形态子序列以及该子序列位于时间序列中具体位置的问题,提出了基于shapelet的工业时间序列中关键形态特征子序列的识别定位算法。其中shapelet为时间序列中最具辨别性的连续子序列,利用shapelet集可以适用于长度不一的子序列相似性计算,并且序列识别结果具有解释性。为了提高时间序列中关键形态特征子序列的识别定位速度和准确性,首先基于遗传算法从时间序列数据集中提取并筛选出具有形态特征的shapelet集,其次采用数据标准化和滑动欧式距离的方法计算shapelet与时间序列中子序列的相似性度量值,用来评估形态特征的相似性;然后定义了自适应相似性阈值以及滞后时间的概念,实现识别和定位时间序列中存在的特征形态子序列,提高了关键形态子序列的识别精度;最后利用公开标准数据集和连铸过程中拉速时间序列集验证了方法的可行性和准确性。
Aiming at the problem of whether there are certain key characteristic shape subsequences in high-frequency time series in industrial control processes and the specific location of this subsequence in the time series,identification and positioning algorithm for key shape characteristic subsequences in industrial time series is proposed based on shapelet.Shapelet are the most discriminative continuous subsequences in time series.The shapelet set can be applied to the similarity calculation of subsequences of different lengths,and the sequence identification results are interpretable.In order to improve the speed and accuracy of identifying and positioning key shape characteristic subsequences in time series,shapelet sets with specific shape are first extracted and screened from the time series data set based on genetic algorithms.Secondly,the method of data standardization and sliding Euclidean distance is used to calculate the similarity measurement value between the shapelet and the subsequence in the time series,which is used to evaluate the similarity of shape characteristic.Then,the concepts of adaptive similarity threshold and lag time are defined to achieve accurate identification and positioning of characteristic shape subsequences existing in time series and improve the recognition accuracy of key shape subsequences.Finally,the feasibility and accuracy of the method were verified using public standard data sets and time series data of casting speed during continuous casting process.
作者
杜学飞
DU Xuefei(China Coal Research Institute,China Coal Technology and Engineering Group,Beijing 100013,China)
出处
《连铸》
北大核心
2024年第6期75-82,共8页
Continuous Casting
基金
中国煤炭科工集团自立重点资助项目(2023-TD-ZD012-001)。
关键词
shapelet
遗传算法
时间序列
相似性度量
形态特征
连铸过程
shapelet
genetic algorithm
time series
similarity measurement
shape characteristic
continuous casting process
