朱朋成 錢虹 江誠
摘 要:為實現(xiàn)磨煤機狀態(tài)預警,提高磨煤機運行的穩(wěn)定性。以熱力學為基礎,對Hp934型中速磨煤機內(nèi)部煤質(zhì)量、水分質(zhì)量、能量平衡進行分析,確定了表征磨煤機運行狀態(tài)的特征參數(shù)。用相關性分析、正態(tài)分布和置信度算法等方法對大量實際生產(chǎn)數(shù)據(jù)進行挖掘整理,確定了各工況下磨煤機穩(wěn)定運行的各特征參數(shù)邊界,制定了預警規(guī)則,建立了磨煤機的預警模型。測試結(jié)果表明,該模型能判斷磨煤機運行的早期異常,證明了該預警模型的有效性和可行性,能為實現(xiàn)磨煤機的運維提供參考。
關鍵詞:熱力學機理;數(shù)據(jù)挖掘;正態(tài)分布;特征參數(shù);狀態(tài)預警
DOI:10.15938/j.jhust.2020.01.007
中圖分類號: TM621
文獻標志碼: A
文章編號: 1007-2683(2020)01-0043-08
Abstract:In order to realize the early warning of the coal mill and to improve the stability of the coal mill -On the basis of thermodynamics, the internal coal quality, moisture quality, and energy balance of the Hp934 medium-speed coal mill were analyzed, and the characteristic parameters that characterize the operation of the coal mill were determinedUsing correlation analysis, normal distribution and confidence algorithm to excavate a large number of actual production data, determine the boundaries of various characteristic parameters of stable operation of the coal mill under various operating conditions, formulate the early warning rules, and establish the early warning model of coal mill-The test results show that the model can accurately determine the early operation anomalies of the coal mill earlier, prove the effectiveness and feasibility of the early warning model, and provide reference for the operation and maintenance of the coal mill-
Keywords:thermodynamic mechanism, data mining, normal distribution, characteristic parameter,early state warning
0 引 言
目前,在火力發(fā)電廠中,磨煤機是鍋爐燃燒制粉系統(tǒng)的核心設備。磨煤機出現(xiàn)故障,將會直接影響鍋爐機組的穩(wěn)定和經(jīng)濟運行,并導致機組出力異常[1]。傳統(tǒng)的單參數(shù)越限報警,其可靠性低,往往不能在故障發(fā)生前發(fā)生警報,這種形式的報警很難對機組設備的運行起到保護作用。而基于多參數(shù)的設備故障預警可以在故障初期判斷出設備的異常狀態(tài),發(fā)出相應的預警信息,為運行維護人員爭取更多的故障處理時間。隨著大數(shù)據(jù)技術的高速發(fā)展,各行各業(yè)都在對海量的數(shù)據(jù)進行挖掘分析,從而獲取新的知識。發(fā)電企業(yè)也積累了大量設備日常運行的數(shù)據(jù),但這些數(shù)據(jù)沒有得到有效的利用,將數(shù)據(jù)挖掘技術引入到電站設備的預警功能中,對提高設備運行穩(wěn)定性、降低生產(chǎn)成本、提高生產(chǎn)效率十分必要。
針對傳統(tǒng)報警方法的缺點,可以建立多參數(shù)的自適應閾值預警模型。目前,多參數(shù)的選取方法主要有機器學習和機理模型[19,23]。文[2]利用邏輯回歸的方法選取車輛的特征,實現(xiàn)了對車輛的檢驗;文[3,4]提出了基于支持向量機(SVM)的特征選擇算法,優(yōu)化了特征選擇方法的分類性能,選擇較少的特征,提高了分類精度;文[5]通過SVM-RFE算法計算不同特征的權(quán)重,得到特征重要性排序,從而選擇最佳特征子集。文[6]使用隨機森林和決策樹來選擇特征和支持決策。文[7]使用決策樹進行特征選擇,減輕了數(shù)據(jù)集中的冗余?;跈C理模型的特征選擇方法首先對設備的運行特性進行分析,然后利用物理規(guī)律描述設備參數(shù)之間的聯(lián)系,最終確定能夠表征設備運行狀態(tài)的特征參數(shù)[8]。很多研究是利用專家經(jīng)驗或設備廠家參數(shù)去設定模型的閾值、限值,這樣實施起來方便、高效。然而,這些設定值過于主觀,也不能隨著運行情況和環(huán)境的變化做出調(diào)整。文[9]使用數(shù)據(jù)挖掘技術從大型系統(tǒng)參數(shù)集中提取繼電器設置信息。對這些結(jié)果的挖掘確定了繼電器閾值設置的邊界,確保其可以檢測孤島操作。基于數(shù)據(jù)挖掘的自適應閾值的方法[10],可以分析特征參數(shù)的相關性與分布規(guī)律,從而確定設備異常的邊界。
針對磨煤機運行環(huán)境復雜和工況多變的情況,本文提出一種以磨煤機運行機理和大數(shù)據(jù)分析相結(jié)合的預警方法。對磨煤機內(nèi)部質(zhì)量、能量平衡進行分析,結(jié)合故障的特征征兆確定預警的特征參數(shù)。在此基礎上采用相關性分析和正態(tài)分布等方法對包含特征參數(shù)在內(nèi)的大量實際生產(chǎn)數(shù)據(jù)進行挖掘,得出了不同工況下預警的各特征參數(shù)自適應邊界,建立了磨煤機狀態(tài)預警的數(shù)學模型,并采用概率密度與置信規(guī)則[11-12]相結(jié)合實現(xiàn)對磨煤機的預警。
1 基于熱力學的磨煤機預警特征參數(shù)的確定
中速磨直吹式制粉系統(tǒng)中的主要設備有給煤機、磨煤機、煤粉分離器及燃燒器等[20-21]。原煤從給煤機進入落煤管,再進入碾磨區(qū),通過碾磨部件之間的相互擠壓變?yōu)槊悍?。一次風進入磨煤機,干燥煤粉的同時也將其吹入粗粉分離器,在分離器中細的煤粉被送到鍋爐中燃燒,較粗的煤粉落到落煤管中繼續(xù)碾磨[13]。
4 結(jié) 語
本文以磨煤機運行熱力學機理為指導,確定表征磨煤機穩(wěn)定運行的特征參數(shù),以給煤量作為磨煤機運行工況,并結(jié)合大量磨煤機生產(chǎn)大數(shù)據(jù),確定了各工況下磨煤機穩(wěn)定運行的特征參數(shù)邊界,同時使用數(shù)據(jù)挖掘的方法,制定了磨煤機預警的判斷規(guī)則。初步結(jié)果表明,磨煤機預警的判斷規(guī)則能為機組運
行人員提供可信度更高磨煤機運行異常提示,有助于實現(xiàn)磨煤機故障的早期預警。
但穩(wěn)定工況下,機組參數(shù)本身有一定的波動性,可考慮用數(shù)據(jù)濾波處理方法做進一步處理;隨樣本數(shù)量不斷積累,特征參數(shù)的上、下邊界也可不斷優(yōu)化;但目前的預警,還沒有能實現(xiàn)磨煤機運行異常的原因分析和故障定位,這還有待于進一步深入探討。
參 考 文 獻:
[1] AGRAWAL V, PANIGRAHI B K, SUBBARAO P M V. Intelligent Decision Support System for Detection and Root Cause Analysis of Faults in Coal Mills[J]. IEEE Transactions on Fuzzy Systems, 2017, 25(4):934.
[2] LEE C, KIM J, PARK E, et al. Multi-feature Vehicle Detection Using Feature Selection[C]. Manchester: IEEE, 2013:234.
[3] DAI K, YU H Y, LI Q. A Semisupervised Feature Selection with Support Vector Machine[J]. Journal of Applied Mathematics, 2013(1):1.
[4] TAN J, ZHANG Z, ZHEN L, et al. Adaptive Feature Selection Via a New Version of Support Vector Machine[J]. Neural Computing & Applications, 2013, 23(3-4):937.
[5] CHEN G S, ZHENG Q Z. Online Chatter Detection of the End Milling Based on Wavelet Packet Transform and Support Vector Machine Recursive Feature Elimination[J]. International Journal of Advanced Manufacturing Technology, 2017, 95(5):1.
[6] Al-MAATHIDI M M, LI F F. Audio Content Feature Selection and Classification a Random Forests and Decision Tree Approach[C]. Nanjing: IEEE, 2015:108.
[7] FERREIRA B, SILVA R G, PEREIRA V. Feature Selection Using Non-binary Decision Trees Applied to Condition Monitoring[C]. Limassol: IEEE, 2017:1.
[8] 錢虹, 宋亮, 陳琪琪,等. 基于生產(chǎn)數(shù)據(jù)挖掘的吹灰需求度置信規(guī)則庫研究[J]. 熱力發(fā)電, 2017, 46(6):113.
QIAN H, SONG L, CHEN QQ, et al. Research on Confidence Rule Base of Soot-blown Demand Based on Production Data Mining[J]. Thermal Power Generation, 2017, 46(6): 113.
[9] El-ARROUDI K, JOOS G. Data Mining Approach to Threshold Settings of Islanding Relays in Distributed Generation[J]. IEEE Transactions on Power Systems, 2007, 22(3):1112.
[10]朱曉東,袁坤杰,王艷玲.基于隱馬爾科夫模型的石油鉆井事故預警方法[J].鄭州大學學報(工學版), 2018, 39(4):51.
ZHU XD, YUAN KJ, WANG YL. Early Warning Method of Oil Drilling Accidents Based on Hidden Markov Model[J]. Journal of Zhengzhou University(Engineering Science Edition), 2018, 39(4): 51.
[11]CHANG L, ZHOU Z J, YOU Y, et al. Belief Rule Based Expert System for Classification Problems with New Rule Activation and Weight Calculation Procedures[J]. Information Sciences, 2016, 336(C):75.
[12]LI Bin, WANG Hongwei, YANG Jianbo, et al. A Belief-rule-based Inference Method for Aggregate Production Planning Under Uncertainty[J]. International Journal of Production Research, 2013, 51(1):83.
[13]魏樂, 苑召雄, 閆媛媛,等. 基于系統(tǒng)動力學的MPS中速磨煤機建模與仿真[J]. 熱力發(fā)電, 2016, 45(8):26.
WEI L, YUAN ZX, YAN YY, et al. Modeling and Simulation of MPS Medium-speed Coal Mill Based on System Dynamics[J]. Thermal Power Generation, 2016, 45(8): 26.
[14]NIEMCZYK P, BENDTSEN J D, RAVN A P, et al. Derivation and Validation of a Coal Mill Model for Control[J]. Ifac Proceedings Volumes, 2012, 20(5):519.
[15]孟令虎, 劉鑫屏. 雙進雙出鋼球磨煤機制粉系統(tǒng)煤粉流量軟測量[J]. 熱力發(fā)電, 2018,47(2):97.
MENG LH, LIU XP. Soft Measurement of Pulverized Coal Flow in a Double-inlet and Double-outlet Ball Mill Coal Pulverizing System[J]. Thermal Power Generation, 2018,47(2): 97.
[16]TENG W, CHENG L P, ZHAO K J. Application of Kernel Principal Component and Pearson Correlation Coefficient in Prediction of Mine Pressure Failure[C]// IEEE Chinese Autonation Congress, Jinan, China, 2017:5704.
[17]AGTERBERG F. Pareto-Lognormal Modeling of Known and Unknown Metal Resources[J].Natural Resources Research, 2017, 26(3):265.
[18]錢虹, 馬萃萃. 基于事件觸發(fā)機制的核電站智能診斷專家系統(tǒng)置信規(guī)則庫的研究[J]. 原子能科學技術, 2017, 51(3):485.
QIAN H, MA CC. Research on Confidence Rule Base of Intelligent Diagnosis Expert System for Nuclear Power Plant Based on Event Triggering Mechanism[J]. Atomic Energy Science and Technology, 2017, 51(3): 485.
[19]宋智超, 康健. 特征選擇方法中三種度量的比較研究[J]. 哈爾濱理工大學學報, 2018(1):111.
SONG ZC, KANG J. A Comparative Study of Three Metrics in Feature Selection Methods[J]. Journal of Harbin University of Science and Technology, 2018(1): 111.
[20]榮盤祥, 張亮, 孫國兵,等. 磨煤機組合方式優(yōu)選在鍋爐系統(tǒng)中的應用[J]. 電機與控制學報, 2017,21(12):100.
RONG PX, ZHANG L, SUN GB, et al. Application of Coal Mill Combination Method Optimization in Boiler System[J]. Journal of Electrical Engineering and Control, 2017,21(12): 100.
[21]溫玄, 王培東, 張海英. 中儲式球磨機制粉系統(tǒng)控制器設計[J]. 哈爾濱理工大學學報, 2010, 15(2):47.
WEN X, WANG PD, ZHANG HY. Design of the Controller for the Powder System of the Ball Mill Mechanism in the Middle Storage[J]. Journal of Harbin University of Science and Technology, 2010, 15(2): 47.
[22]李輝, 李洋, 楊東,等. 基于EMD相關去噪的風電機組振動噪聲抑制及特征頻率提取[J]. 電機與控制學報, 2016, 20(1):73.
LI H, LI Y, YANG D, et al. Vibration Noise Suppression and Characteristic Frequency Extraction of Wind Turbine Based on EMD Correlation Denoising[J]. Journal of Electrical Engineering and Control, 2016, 20(1): 73.
[23]榮盤祥, 曾凡永, 黃金杰. 數(shù)據(jù)挖掘中特征選擇算法研究[J]. 哈爾濱理工大學學報, 2016, 21(1):106.
RONG PX, ZENG FY, JIN JJ. Research on Feature Selection Algorithm in Data Mining[J]. Journal of Harbin University of Science and Technology, 2016, 21(1): 106.
(編輯:溫澤宇)