|
基于時間序列的趨勢提取方法的研究
劉冰琪1 解 初2 2023/10/31 18:25:44
1，泰山科技學(xué)院 山東泰安 271000
2，國網(wǎng)山東省電力公司泰安供電公司 山東泰安 271000
摘 要：時間序列是指對同一統(tǒng)計指標(biāo)下的數(shù)值根據(jù)不同的時間發(fā)生次序進(jìn)行排列，進(jìn)而得到的一組現(xiàn)實的、真實的數(shù)列。該數(shù)列具有高維度、大規(guī)模，結(jié)構(gòu)復(fù)雜且易受噪聲干擾等特點，這也使得基于原始時間序列數(shù)據(jù)進(jìn)行數(shù)據(jù)趨勢特征信息的挖掘存在較大的難度。時間序列的趨勢提取方法能夠?qū)⒏呔S時間序列數(shù)據(jù)轉(zhuǎn)化為低維的時間序列數(shù)據(jù)，同時能夠保留反映原始時間序列數(shù)據(jù)特征的重要信息，該方法在時間序列數(shù)據(jù)挖掘方面具有廣泛應(yīng)用。主要趨勢提取的方法有自頂向下趨勢提取方法、自底向上趨勢提取方法以及滑動窗口趨勢提取方法等，本文基于工業(yè)系統(tǒng)中閉環(huán)控制系統(tǒng)的時間序列數(shù)據(jù)，對上述三種趨勢提取方法進(jìn)行對比分析，并確定最優(yōu)趨勢提取方法。
關(guān)鍵詞：時間序列；趨勢提�。婚]環(huán)控制系統(tǒng)；性能評價
Research on Trend Extraction Method Based on Time Series
LIU Bing-qi1 Xie Chu2
1,Taishan College of Science and Technology Tai’an Shandong 271000
2,State Grid Shandong Electric Power Company Tai’an Power Supply Company Tai’an Shandong 271000
Abstract：Time series refers to a group of realistic and real data series obtained by arranging the values under the same statistical index according to different time sequence. The sequence has the characteristics of high dimension, large scale, complex structure and easy to be disturbed by noise, which also makes it difficult to mine data trend feature information based on original time series data. The trend extraction method of time series can transform high-dimensional time series data into low-dimensional time series data, and retain important information reflecting the characteristics of original time series data. This method has been widely used in time series data mining. The main trend extraction methods include top-down trend extraction method, bottom-up trend extraction method and sliding window trend extraction method. Based on the time series data of the closed-loop control system in the industrial system, this paper compares and analyzes the above three trend extraction methods and determines the optimal trend extraction method.
Key words：time series；trend extraction；closed loop control system；rating of merit
1 時間序列
時間序列是基于某種統(tǒng)計指標(biāo)按時間先后順序進(jìn)行排序進(jìn)而形成的數(shù)列。隨著工業(yè)自動化水平的發(fā)展，基于時間序列的預(yù)測及異常檢測越來越廣泛。通過對時間序列的趨勢特征[1]進(jìn)行提取，挖掘時間序列所反映的發(fā)展過程、發(fā)展方向，進(jìn)而預(yù)測時間序列未來可能達(dá)到的水平，或基于歷史數(shù)據(jù)，分析檢測當(dāng)（未完，下一頁）
附件下載：基于時間序列的趨勢提取方法的研究
|