基于示功图自动诊断的单井远程无线网络监控系统的设计与实现
|
摘要
能源问题是关系到社会稳定，国民经济持续健康发展的重大问题，而原油又在能源结构中占有举足轻重的地位，应用计算机技术建设数字油田是实现油田管理自动化，保障安全生产、提高生产效率的有效方法与重要手段。
本文是在针对抽油设备长期处于野外工作环境作业，其运行情况难以获得、维护困难这一采油行业实际情况而设计研发的系统基础上形成的。本系统在组成上分三级管理平台，既控制中心管理平台、分控中心管理平台，终端控制管理平台，各级管理平台既是独立的管理系统又是有机的组合体。它通过对抽油机自动化监测、联合站自动化监测，解决了设备运行问题、安全问题以及维护问题，极大地提高了生产效率。
本系统还通过对抽油机各种数据的采集和检测，为用户提供丰富的辅助决策信息，并对示功图、电功图等数据进行深层次分析，从而有效确定抽油机运行状态、故障类别、泵效、产业量等数据，为下一步制定生产决策，提高生产效率与原油产量，提供可靠的参考依据。
单井远程无线网络监控系统高效的整合了采油行业的业务流程，通过抽油井以及联合站的数据采集、分析、整合与监控，极大的提高了油田自动化水平。满足了采油数字化的需求。
Now, our country had treated the energy issues as a more and more important problem. The energy issue has already effected the social stability and the economic development. Especially in recent years, there is a large increment on the dependence on foreign oil. As a result, how to lower production costs, improve labor efficiency and reduce security risks had became to the task we must face to in the following period. Using computer technology to build digital oilfield is an important method to realize the oilfield automatic management and security production. It is also an effective way to improve the production efficiency. The system is the project between KingEagle technology cooperation and a company of DaQing oilfield, and its main objective is to achieve long-range surveillance and unmanned production, and realize the the digital management of oil field.
Mechanical oil pumping is the most important method of crude oil production in the world, and sucker rod oil pumping system is the most popular method of mechanical oil pumping. The equipments of sucker rod oil pumping system are vital important in the whole production of crude oil. But as is known to all, the oil pumps word on the field, and it is hard to maintain them and to know their work status, and the losing of the oil pumps' malfunction is very big. Therefore, this paper researches on the remote monitoring and fault diagnosis system of sucker rod oil well, and show the design schema and there realization information. Main achievements of this paper are showed in the following paragraph.
The paper first show the design schema of the oil pump remote monitoring and management system. Based on the actual oilfield business model, we have designed the system as a three-tier management platform. They are the control center management platform, sub-control center management platform and control terminal management platform. All the Platforms are both independent management system and an organic body composition. The system can automatic monitor the united-station and the pumping equipment, which can solve the safety and maintenance problems, and greatly improve the production efficiency. In this system, through detecting and analyzing the data we had got, we can get the knowledge of the pumps' state, which give the making of the decision a reliable proof a favorable reference.
To make the system more user-friendly, and increase the automation level of the oil field, we also designed rich diagram display method, report forms, curve, alarm methods, GIS and other function modules. In those modules, we had used voice synthesis, geographic information systems and other advanced technologies. During the developing period, we use RUP as our development process, and use many design patterns to improve our design. For the reason of using these advanced object-oriented experience, the oil pump remote monitoring and management system now is powerful, flexible, and user recognized.
There is no doubt that in the oil pump remote monitoring and management system, the fault diagnosis technology of the oil pump is one of the most crucial technologies. This paper complete show the fault diagnosis schema of the oil pump, including the diagram's pretreatment, feature extraction and standardized, classifier design and other issues. During the diagram's pretreatment period, we use mathematical morphology's erosion, dilation, thinning and other algorithm. During the diagram's feature extraction period, this paper introduce and the design four method for the learning and training of the classifier. They are the heart shape round distance, the heart shape rectangle distance, the standard distance and the difference distance.
During the classifier designing period, this paper use simulated annealing algorithm to optimize the training of the BP neural network. This is because the BP neural network has its own disadvantage. As we all know, during the training or the neural network, the parameter of the learning rate is very important and difficult to select. If learning rate is too large, when the neural network is close to the globe optimal solution, the neural network may surge around this solution. If learning rate is too small, the neural network may converge to the globe optimal solution too slowly, and which is more important that the neural network may stay at the local optimal solutions, and be unable to come over them. All the case above will make the training of the training of neural networks unsuccessful.
For the reason above, this paper uses simulated annealing algorithm to optimize the training of BP neural network, dynamic deciding the neural network's learning rate. At the beginning of the training of the BP network, we select a larger learning rate to study, and if the network's error is reducing, the learning rate remains the original value, and if the error changes the worse, the paper use simulated annealing algorithm to reduce the BP neural network's learning rate and use the new learning rate to continue the training of the network. The results show that BP neural network optimized by simulated annealing algorithm can better avoid the oscillation of the network and convergence to the local optimal solutions, and also the training of the network using the mixed algorithm is faster, and the result has a better extensive ability.
The practical application of the oil pump remote monitoring and management system based on the automatic fault diagnosis shows a better effect. There are still some following works to do. They are the automatic fault diagnosis to other types of oil pumping equipment, such as screw pump, the electric submersible pump.
本文是在针对抽油设备长期处于野外工作环境作业，其运行情况难以获得、维护困难这一采油行业实际情况而设计研发的系统基础上形成的。本系统在组成上分三级管理平台，既控制中心管理平台、分控中心管理平台，终端控制管理平台，各级管理平台既是独立的管理系统又是有机的组合体。它通过对抽油机自动化监测、联合站自动化监测，解决了设备运行问题、安全问题以及维护问题，极大地提高了生产效率。
本系统还通过对抽油机各种数据的采集和检测，为用户提供丰富的辅助决策信息，并对示功图、电功图等数据进行深层次分析，从而有效确定抽油机运行状态、故障类别、泵效、产业量等数据，为下一步制定生产决策，提高生产效率与原油产量，提供可靠的参考依据。
单井远程无线网络监控系统高效的整合了采油行业的业务流程，通过抽油井以及联合站的数据采集、分析、整合与监控，极大的提高了油田自动化水平。满足了采油数字化的需求。
Now, our country had treated the energy issues as a more and more important problem. The energy issue has already effected the social stability and the economic development. Especially in recent years, there is a large increment on the dependence on foreign oil. As a result, how to lower production costs, improve labor efficiency and reduce security risks had became to the task we must face to in the following period. Using computer technology to build digital oilfield is an important method to realize the oilfield automatic management and security production. It is also an effective way to improve the production efficiency. The system is the project between KingEagle technology cooperation and a company of DaQing oilfield, and its main objective is to achieve long-range surveillance and unmanned production, and realize the the digital management of oil field.
Mechanical oil pumping is the most important method of crude oil production in the world, and sucker rod oil pumping system is the most popular method of mechanical oil pumping. The equipments of sucker rod oil pumping system are vital important in the whole production of crude oil. But as is known to all, the oil pumps word on the field, and it is hard to maintain them and to know their work status, and the losing of the oil pumps' malfunction is very big. Therefore, this paper researches on the remote monitoring and fault diagnosis system of sucker rod oil well, and show the design schema and there realization information. Main achievements of this paper are showed in the following paragraph.
The paper first show the design schema of the oil pump remote monitoring and management system. Based on the actual oilfield business model, we have designed the system as a three-tier management platform. They are the control center management platform, sub-control center management platform and control terminal management platform. All the Platforms are both independent management system and an organic body composition. The system can automatic monitor the united-station and the pumping equipment, which can solve the safety and maintenance problems, and greatly improve the production efficiency. In this system, through detecting and analyzing the data we had got, we can get the knowledge of the pumps' state, which give the making of the decision a reliable proof a favorable reference.
To make the system more user-friendly, and increase the automation level of the oil field, we also designed rich diagram display method, report forms, curve, alarm methods, GIS and other function modules. In those modules, we had used voice synthesis, geographic information systems and other advanced technologies. During the developing period, we use RUP as our development process, and use many design patterns to improve our design. For the reason of using these advanced object-oriented experience, the oil pump remote monitoring and management system now is powerful, flexible, and user recognized.
There is no doubt that in the oil pump remote monitoring and management system, the fault diagnosis technology of the oil pump is one of the most crucial technologies. This paper complete show the fault diagnosis schema of the oil pump, including the diagram's pretreatment, feature extraction and standardized, classifier design and other issues. During the diagram's pretreatment period, we use mathematical morphology's erosion, dilation, thinning and other algorithm. During the diagram's feature extraction period, this paper introduce and the design four method for the learning and training of the classifier. They are the heart shape round distance, the heart shape rectangle distance, the standard distance and the difference distance.
During the classifier designing period, this paper use simulated annealing algorithm to optimize the training of the BP neural network. This is because the BP neural network has its own disadvantage. As we all know, during the training or the neural network, the parameter of the learning rate is very important and difficult to select. If learning rate is too large, when the neural network is close to the globe optimal solution, the neural network may surge around this solution. If learning rate is too small, the neural network may converge to the globe optimal solution too slowly, and which is more important that the neural network may stay at the local optimal solutions, and be unable to come over them. All the case above will make the training of the training of neural networks unsuccessful.
For the reason above, this paper uses simulated annealing algorithm to optimize the training of BP neural network, dynamic deciding the neural network's learning rate. At the beginning of the training of the BP network, we select a larger learning rate to study, and if the network's error is reducing, the learning rate remains the original value, and if the error changes the worse, the paper use simulated annealing algorithm to reduce the BP neural network's learning rate and use the new learning rate to continue the training of the network. The results show that BP neural network optimized by simulated annealing algorithm can better avoid the oscillation of the network and convergence to the local optimal solutions, and also the training of the network using the mixed algorithm is faster, and the result has a better extensive ability.
The practical application of the oil pump remote monitoring and management system based on the automatic fault diagnosis shows a better effect. There are still some following works to do. They are the automatic fault diagnosis to other types of oil pumping equipment, such as screw pump, the electric submersible pump.
引文
[1]．俞凯，浅析2020年我国石油能源需求及对策，http://www.jsmlr.gov.cn/gb/jsmlr/gtzy/zygl/ywzj/zywz/index.html,2005.12
[2]．2006年原油市场分析报告，http://chinaccm.com/4814811/481101/news/20060626/143054.asp,2006.03
[3]．毛宝瑚，油气田自动化，石油大学出版社，2005．01
[4]．抽油井示功图的测试和分析编写小组，抽油井示功图的测试和分析，石油工业出版社．1983．05
[5]. Erich Gamma, Richard Helm, Ralph Johnson, John Vlissides, Design patterns: elements of reusable object-oriented software, Addison-Wesley Longman Publishing Co., Inc., Boston, MA, 1995. 01
[6]. Elisabeth Freeman, Eric Freeman, Bert Bates, Kathy Sierra, Head First Design Patterns, O' Reilly & Associates, Inc., 2004. 10
[7]. Ivar Jacobson, Grady Booeh, James Rumbaugh, The unified software development process, Addison-Wesley Longman Publishing Co., Inc., Boston, MA, 1999. 02
[8]. G. Booch, R. A. Maksimchuk, M. W. Engel, B. J. Young, J. Conallen, K. A. Houston, R. Martin, J. W. Newkirk, Object-Oriented Analysis and Design with Applications (3rd Edition), Addison-Wesley Professional, 2004. 04
[9]. G. Booch, J. Rumbaugh, I. Jacobson, Unified Modeling Language User Guide, The (2nd Edition), Addison-Wesley Professional, 2005.05
[10]．杨淑莹，VC++图像处理程序设计(第2版)，清华大学出版社，2005．01
[11]．杨淑莹，图像模式识别：VC++技术实现，清华大学出版社，北京交通大学出版社，2005．07
[12]．章毓晋，图像工程(上册)图像处理(第2版)，清华大学出版社，2006．03
[13]．章毓晋，图像工程(中册)图像分析(第2版)，清华大学出版社，2006．03
[14]．章毓晋，图像工程(下册)图像理解(第2版)，清华大学出版社，2006．03
[15]．李玉冰等，多边形重心的计算方法，计算机应用2005年12月第25卷391-393
[16]．王志伟，薛改珍，金正谦，王进恒，应用人工神经网络识别示功图，石油钻探技术，2001，Vol．29 No．2 P．51-57
[17]．T．M．阿里耶夫等，抽油井自动控制和诊断，石油工业出版社，1993．11
[18]．韩家炜，Micheline Kamber，数据挖掘概念与技术，机械工业出版社，2001．08
[19]．Ian H．Witten，Eibe Frank，数据挖掘：实用机器学习技术(第2版)，机械工业出版社，2006．02
[20]．周春光，梁艳春，计算智能，吉林大学出版社，2001．11
[21]．Simon Haykin，神经网络原理，机械工业出版社，2006．05
[22]. Yin JX, Chen SY, Qiu J. Neural network prediction model and its application based on GA and BP. Journal of Dalian University of Technology, 2002, 42(5): 594-598 (in Chinese with English abstract)
[23]. Yan PF, Zhang CS. Artificia Neural Network and Simulating-Evolution Computation 2ed. Beijing: Tsinghua University Press, 2005. 09
[24]. Ananth Ranganathan, The Levenberg-Marquardt Algorithm, http://www.Cc.Gatech.Edu/people/home/ananth/lmtut.pdf, 2004.06
[25]．卢开澄，计算机算法导引-设计与分析，清华大学出版社，2006．01
[26]．王凌著，智能优化算法及其应用，清华大学出版社，2001．10
[2]．2006年原油市场分析报告，http://chinaccm.com/4814811/481101/news/20060626/143054.asp,2006.03
[3]．毛宝瑚，油气田自动化，石油大学出版社，2005．01
[4]．抽油井示功图的测试和分析编写小组，抽油井示功图的测试和分析，石油工业出版社．1983．05
[5]. Erich Gamma, Richard Helm, Ralph Johnson, John Vlissides, Design patterns: elements of reusable object-oriented software, Addison-Wesley Longman Publishing Co., Inc., Boston, MA, 1995. 01
[6]. Elisabeth Freeman, Eric Freeman, Bert Bates, Kathy Sierra, Head First Design Patterns, O' Reilly & Associates, Inc., 2004. 10
[7]. Ivar Jacobson, Grady Booeh, James Rumbaugh, The unified software development process, Addison-Wesley Longman Publishing Co., Inc., Boston, MA, 1999. 02
[8]. G. Booch, R. A. Maksimchuk, M. W. Engel, B. J. Young, J. Conallen, K. A. Houston, R. Martin, J. W. Newkirk, Object-Oriented Analysis and Design with Applications (3rd Edition), Addison-Wesley Professional, 2004. 04
[9]. G. Booch, J. Rumbaugh, I. Jacobson, Unified Modeling Language User Guide, The (2nd Edition), Addison-Wesley Professional, 2005.05
[10]．杨淑莹，VC++图像处理程序设计(第2版)，清华大学出版社，2005．01
[11]．杨淑莹，图像模式识别：VC++技术实现，清华大学出版社，北京交通大学出版社，2005．07
[12]．章毓晋，图像工程(上册)图像处理(第2版)，清华大学出版社，2006．03
[13]．章毓晋，图像工程(中册)图像分析(第2版)，清华大学出版社，2006．03
[14]．章毓晋，图像工程(下册)图像理解(第2版)，清华大学出版社，2006．03
[15]．李玉冰等，多边形重心的计算方法，计算机应用2005年12月第25卷391-393
[16]．王志伟，薛改珍，金正谦，王进恒，应用人工神经网络识别示功图，石油钻探技术，2001，Vol．29 No．2 P．51-57
[17]．T．M．阿里耶夫等，抽油井自动控制和诊断，石油工业出版社，1993．11
[18]．韩家炜，Micheline Kamber，数据挖掘概念与技术，机械工业出版社，2001．08
[19]．Ian H．Witten，Eibe Frank，数据挖掘：实用机器学习技术(第2版)，机械工业出版社，2006．02
[20]．周春光，梁艳春，计算智能，吉林大学出版社，2001．11
[21]．Simon Haykin，神经网络原理，机械工业出版社，2006．05
[22]. Yin JX, Chen SY, Qiu J. Neural network prediction model and its application based on GA and BP. Journal of Dalian University of Technology, 2002, 42(5): 594-598 (in Chinese with English abstract)
[23]. Yan PF, Zhang CS. Artificia Neural Network and Simulating-Evolution Computation 2ed. Beijing: Tsinghua University Press, 2005. 09
[24]. Ananth Ranganathan, The Levenberg-Marquardt Algorithm, http://www.Cc.Gatech.Edu/people/home/ananth/lmtut.pdf, 2004.06
[25]．卢开澄，计算机算法导引-设计与分析，清华大学出版社，2006．01
[26]．王凌著，智能优化算法及其应用，清华大学出版社，2001．10