검색 상세

Asset 및 Data 상관관계를 활용한 예측정보시스템에 관한 연구

Study on Development of Early Information System Using Both Asset Characteristics and Correlational Data

  • 文漢柱 (Moon, Han Joo, 의용과학대학원 의료영상공학과)

초록/요약

ABSTRACT Study on Development of Early Information System Using Both Asset Characteristics and Correlational Data Han Joo Moon Department of Biomedical Image Engineering Graduate School of Bio-Medical Science Korea University This paper shows that implementing Early Information System can deliver many benefits for the power generation facility and its management. The major advantage of using EIS(Early Information System)is to increase the efficiency of operation and maintenance of the power plant because the EIS can optimize the power plant and monitor entire power plant with reliability index in real time. This can lead to operators enable to diagnosis the entire system and monitor their facility more precisely via intelligent predictive information before the facility will be tipped. To do this, the study conducted two ways of tag grouping so that the EIS can be more optimized and predict the abnormal information before the operation of power plant fails. Firstly the test of tag grouping was performed with an equipment characteristic data only. The results of prediction and analysis after tag grouping according to the system prediction were confirmed that some abnormal signal is recognized even though there is no abnormality in the actual related signal of the equipment system(Feed Water Pump). As the result, it is difficult to analyze the accurate prediction information by only past data and equipment characteristic data. This preliminary research founds out that other parameters or factors should be included to improve accuracy of the predictive system. Therefore, this paper introduces a better tag grouping method that can improve the unstable prediction of EIS which resulted from the previous method. To minimize the research, the final test focuses on Big Data of Feed Water Pump equipment only. This paper shows that with the use of the new method, EIS dramatically overcomes the inaccuracy of the data modeling as well as the EIS can produce more precise predictive information collected from entire system of EIS. This method conducted a process of analyzing data correlation among asset and combined the analyzed correlation data into the equipment characteristics. Finally, the paper demonstrates that a tag grouping and modeling of Feed Water Pump were built based on three major factors such as data correlation, equipment characteristics and historical data. This results in the EIS that can predict the abnormal data far more precisely and deliver real predictive information to users /operators at power plants. Furthermore, this accurate prediction of IES enables users to evaluate and manage their assets before they will be faced with failure of assets or trip. Key Words : Predictive Information, Intelligent Prediction, Early Information, Power Plant, Data Modeling, Big-data Managing, Historical Data, Early Warning

more

목차

I. INTRODUCTION ８
1. BACKGROUND ８
2. RESEARCH TRENDS ９
3. ARTICLE COMPOSITION １０
II. THEORETICAL BACKGROUND １１
1. PRINCIPLES OF POWER PLANT SYSTEM １１
2. SIGNAL SYSTEM OF POWER PLANT FACILITY AND RANGE OF
PLANT MONITORING １２
1) EQUIPMENT SYSTEM OF POWER PLANT １２
2) TYPES OF TAG ALIAS FOR EACH EQUIPMENT SYSTEM １９
3. CHARACTERISTICS AND COMPOSITIONS OF PREDICTIVE
INFORMATION SYSTEM ２１
1) DATA WORK FLOW OF PREDICTIVE INFORMATION SYSTEM
２２
2) DATA MODEL OF PREDICTIVE INFORMATION SOFTWARE ２３
4. INCORRECT RESULT OF PREDICTION DATA FROM POWER PLANT
SYSTEM ２４
1) RECOGNITION ERROR OF BAD DATA FROM HISTORICAL DATA
２４
2) ABNORMAL DATA FROM SPECIFIC DURATION OF PLANT
OPERATION ２５
3)RECOGNITION ERROR FROM MEASUREMENT OF SENSOR DETECTOR
２６
4) MISMATCH AND DETECTION ERROR OF TAG MAPPING
INFORMATION ２７
5) INCORRECT TAG GROUPING FOR PREDICTIVE MODELS
２８
5. PROCESS OF FAILURE PREDICTION USING EIS SOFTWARE
２９
1) TO ANALYZE SIGNALS ２９
2) TO GENERATE DATABASE ２９
3) TO CONVERT HISTORICAL DATA ３０
4) TO BUILD MODELS ３０
5) TO CLASSIFY MODEL GROUPS AND REMOVE ABNORMAL DATA
FROM MODEL GROUPS ３１
6) TO CONSTRUCT EIS-SUCCESS TREE DATABASE BASED ON
PLANT SYSTEM CODE ３１
7) TO APPLY C-FACTOR INTO REAL TIME DATA AND ITS
MATERIAL ３２
8) TO OPTIMIZE THE PREDICTIVE VALUE WITH VARIOUS
METHODS ３３
9) THE FINAL RESULT ON TREND CHART ３５
III. OPTIMIZING MODEL GROUP OF PREDICTIVE INFORMATION
SYSTEM FOR EACH ASSET ３５
1. TO SELECT MODEL IN ORDER FOR PREDICTING ABNORMAL DATA
FROM EQUIPMENT ３５
1) DEAERATOR ３６
2) FEED WATER PUMP/BFP ３７
2. ASSET MODEL GROUPING ３７
3. SEVERITY FOR EACH EQUIPMENT ３８
4. CLASSIFYING SIGNAL OF EQUIPMENT AND ITS GROUPING 3 8
5. RESULT OF SIGNAL GROUPING FOR EQUIPMENT AND ANALYSIS
OF OUTCOME ３９
1) OUTCOMES OF SIGNAL GROUPING FOR EQUIPMENT ３９
2) ANALYSIS ON OUTCOMES OF SIGNAL GROUPING FOR
EQUIPMENT ４１
IV. OPTIMIZING MODEL GROUPING USING COMBINATION OF ASSET
CHARACTERISTICS AND DATA CORRELATION ４２
1. SYSTEM ANALYSIS RELATED TO FEED WATER PUMP ４２
2. DEFINING CORRELATIONAL TAGS FROM TURBINE SYSTEM ４４
3. OPTIMIZING CORRELATIONAL FACTORS BETWEEN FEED WATER
PUMP AND TURBINE SYSTEM ４６
1) PROCESS OF OPTIMIZATION APPLIED BY CORRELATION ４６
2) CORRELATION ANALYSIS BETWEEN FEED WATER PUMP AND
TURBINE SYSTEM ４７
3) TO GENERATE DATABASE FOR FAILURE FACTORS ５０
4) TO CONVERT DATA VIA C-FACTOR IN ORDER TO UTILIZE THE
CORRELATION OF TWO ASSETS ５０
5) MODEL GROUPING VIA CORRELATION ANALYSIS ５１
6) TO REBUILD MODEL AND REMOVE ABNORMAL DATA USING EIS
５２
4. OPTIMIZATION AND RESULT OF ANALYSIS ON UNIFIED MODEL
GROUPING ５４
V. CONCLUSION ５５
VI. REFERENCE ５６
ABSTRACT ５８

more
반출 Meta View 목록