【摘要】：故障預(yù)測(cè)技術(shù)能夠有效識(shí)別故障潛在信息,從而避免惡性設(shè)備損壞事故的發(fā)生,以及故障帶來(lái)的高維修成本與高發(fā)電量損失,是實(shí)現(xiàn)故障后檢修到預(yù)測(cè)性檢修智能化轉(zhuǎn)變的關(guān)鍵所在。因此,本文通過(guò)對(duì)風(fēng)電機(jī)組歷史運(yùn)行數(shù)據(jù)、在線監(jiān)測(cè)數(shù)據(jù)和試驗(yàn)數(shù)據(jù)的系統(tǒng)分析與深度挖掘,提煉出隱藏在多源數(shù)據(jù)中的故障特征參數(shù),展開(kāi)對(duì)風(fēng)電機(jī)組與關(guān)鍵部件的故障預(yù)測(cè)技術(shù)研究,為智能化運(yùn)維提供技術(shù)支撐。首先為提高數(shù)據(jù)源質(zhì)量,對(duì)缺失、無(wú)效、失真的“壞數(shù)據(jù)”進(jìn)行辨識(shí)與重構(gòu)。然后依據(jù)SCADA數(shù)據(jù),建立非線性狀態(tài)觀測(cè)(NEST)模型并加以改進(jìn),提出一種考慮樣本優(yōu)化的風(fēng)電機(jī)組齒輪箱故障預(yù)測(cè)方法,并與不同模型進(jìn)行對(duì)比,驗(yàn)證模型的時(shí)效性與優(yōu)越性,通過(guò)對(duì)振動(dòng)信號(hào)加速度值的時(shí)域、頻域分析,提出一種基于Hilbert變換的齒輪箱故障預(yù)測(cè)方法,應(yīng)用D-S證據(jù)理論將以上兩種預(yù)測(cè)結(jié)果進(jìn)行信息融合,獲得更科學(xué)、更符合實(shí)際的故障預(yù)測(cè)結(jié)果。其次依據(jù)現(xiàn)場(chǎng)試驗(yàn),對(duì)偏航系統(tǒng)進(jìn)行功率特性測(cè)試、偏航精度測(cè)試,有效預(yù)測(cè)偏航系統(tǒng)存在偏航制導(dǎo)故障,并對(duì)故障加以修復(fù)、補(bǔ)償,又提出優(yōu)化偏航死區(qū)的新思路。最后從整機(jī)角度考慮,提出一種基于灰色熵AHP與TOPSIS法的風(fēng)電機(jī)組健康狀態(tài)評(píng)估方法,對(duì)同場(chǎng)同期風(fēng)機(jī)進(jìn)行綜合排序,將排名靠后的風(fēng)機(jī)視為性能劣化、存在潛在故障的“嫌疑風(fēng)機(jī)”,從而實(shí)現(xiàn)故障預(yù)測(cè)的目的。通過(guò)與現(xiàn)有方法進(jìn)行對(duì)比,結(jié)果顯示文章所提方法預(yù)測(cè)精度更高,更實(shí)用、有效,可為相關(guān)研究提供參考。
[Abstract]:Fault prediction technology can effectively identify the potential fault information, so as to avoid the occurrence of malignant equipment damage accidents, as well as the high maintenance cost and high power generation loss caused by faults. It is the key to realize the intelligent transformation from post-fault maintenance to predictive maintenance. Therefore, through the systematic analysis and deep mining of the historical operation data, on-line monitoring data and test data of wind turbine, the fault characteristic parameters hidden in multi-source data are extracted in this paper. The fault prediction technology of wind turbine and key components is studied to provide technical support for intelligent operation and maintenance. First of all, in order to improve the quality of data sources, the missing, invalid and distorted "bad data" are identified and reconstructed. Then, according to SCADA data, the nonlinear state observation (NEST) model is established and improved, and a fault prediction method of wind turbine gearbox considering sample optimization is proposed and compared with different models to verify the timeliness and superiority of the model. Through the time domain and frequency domain analysis of the acceleration value of vibration signal, a fault prediction method of gearbox based on Hilbert transform is proposed. The above two prediction results are fusion by using D 鈮，

本文編號(hào)：2495412