基于自組織神經(jīng)網(wǎng)絡(luò)的火電廠健康狀態(tài)數(shù)據(jù)提取算法
吳勝聰，陳雨軒，沈可心，程浩軒
（三峽大學(xué) 電氣與新能源學(xué)院，湖北 宜昌 443002）
摘 要：火電廠設(shè)備健康數(shù)據(jù)提取是火電廠設(shè)備狀態(tài)評(píng)估數(shù)據(jù)處理的一個(gè)關(guān)鍵步驟，有利于提高設(shè)備狀態(tài)評(píng)估的準(zhǔn)確性與效率。將設(shè)備狀態(tài)數(shù)據(jù)首先利用R 型層次聚類進(jìn)行特征參數(shù)選取與冗余數(shù)據(jù)清除，再采用自組織神經(jīng)網(wǎng)絡(luò)篩選異常值。利用所訴方法對(duì)某發(fā)電廠的汽泵前置泵設(shè)備的監(jiān)測(cè)數(shù)據(jù)進(jìn)行健康狀態(tài)數(shù)據(jù)提取，發(fā)現(xiàn)清除的異常數(shù)據(jù)遠(yuǎn)遠(yuǎn)大于提取出的健康數(shù)據(jù)，表明該方法清除的數(shù)據(jù)滿足預(yù)期，為后續(xù)健康狀態(tài)評(píng)估提供了準(zhǔn)確的參照數(shù)據(jù)，并且降低監(jiān)測(cè)數(shù)據(jù)維度提高評(píng)估效率。
關(guān)鍵詞：大數(shù)據(jù)；自組織神經(jīng)網(wǎng)絡(luò)；R 型聚類；電力設(shè)備狀態(tài)數(shù)據(jù)
中圖分類號(hào)：TM621 文獻(xiàn)標(biāo)識(shí)碼：A 文章編號(hào)：1007-3175(2019)09-0027-06
Health State Data Extraction Algorithm for Thermal Power Plant Based on Self-Organizing Neural Network
WU Sheng-cong, CHEN Yu-xuan, SHEN Ke-xin, CHENG Hao-xuan
（College of Electrical Engineering & New Energy, China Three Gorges University, Yichang 443002, China）
Abstract: The health data extraction of thermal power plant equipment is a key step in the processing of equipment state assessment of thermal power plants, which is conducive to improving the accuracy and efficiency of equipment state assessment. The power equipment status data were carried out characteristic parameters selection and redundant data eliminating by R-type hierarchical clustering, then the outliers of device status data were filtered by self-organizing neural network. The proposed algorithm was used to extract the health status data from the monitoring data on turbine pump booster pump device in certain power plant. It is found that The clearing abnormal data is far greater than the extracted health data, which indicates that the algorithm meets the expectation. This algorithm provides the accurate reference data for subsequent health assessment, reducing the monitoring data dimension and improving evaluation efficiency.
Key words: big data; self-organizing neural network; R-type clustering; power equipment status data
參考文獻(xiàn)
[1] 高起棟. 基于數(shù)據(jù)挖掘技術(shù)的火電廠設(shè)備狀態(tài)監(jiān)測(cè)系統(tǒng)[J]. 工業(yè)技術(shù)創(chuàng)新，2017，4(6)：24-27.
[2] 謝小鵬，林玥廷，林英明. 火電廠設(shè)備狀態(tài)監(jiān)測(cè)與故障預(yù)警的研究[J]. 華電技術(shù)，2018，40(6)：7-9.
[3] 嚴(yán)英杰，盛戈皞，陳玉峰，等. 基于時(shí)間序列分析的輸變電設(shè)備狀態(tài)大數(shù)據(jù)清洗方法[J]. 電力系統(tǒng)自動(dòng)化，2015，39(7)：138-144.
[4] 徐步云，倪禾. 自組織神經(jīng)網(wǎng)絡(luò)和K-means聚類算法的比較分析[J]. 新型工業(yè)化，2014，4(7)：63-69.
[5] 劉煥海，葉劍鋒，阿斯耶姆. 基于自組織特征映射網(wǎng)絡(luò)(SOM) 的聚類分析方法[J]. 軟件導(dǎo)刊，2016，15(12)：133-135.
[6] 馮震. 無(wú)線傳感器網(wǎng)絡(luò)數(shù)據(jù)離群點(diǎn)檢測(cè)若干方法研究[D]. 上海：上海大學(xué)，2017.
[7] 陸可，鄒啟鳴，李鳴，等. 基于R型聚類- 因子分析的指標(biāo)體系簡(jiǎn)化方法[J]. 計(jì)算機(jī)系統(tǒng)應(yīng)用，2016，25(5)：118-123.
[8] 張好勇，張東亮，高樹軍，等. 基于自組織特征映射神經(jīng)網(wǎng)絡(luò)的高壓斷路器故障診斷[J]. 電氣應(yīng)用，2016，35(23)：21-24.
[9] 焦毛. 基于多重分形和SOM神經(jīng)網(wǎng)絡(luò)的水電機(jī)組振動(dòng)故障診斷研究[D]. 西安：西安理工大學(xué)，2018.
