基于萤火虫支持向量机的抽油机工况故障诊断研究

发布时间：2018-10-25 14:29

【摘要】：在国内外油田生产过程中,机械采油仍然占有很大一部分。在整个机械采油中抽油机等采油设备对于油田的生产非常重要。由于抽油机大多是在野外工作,长期处在高温度、高载荷等恶劣的环境中,而且抽油机的井下工况也是极其复杂,导致抽油机经常出现故障。抽油机的故障所造成的经济损失是极其巨大的,因此需要对抽油机的工况进行故障诊断。基于人工智能的机器学习方法能够构造多类分类学习模型并进行故障诊断,已成为当前研究的热点之一。因此本文提出一种基于萤火虫优化支持向量机的抽油机工况故障诊断方法,主要内容如下:首先,针对当前油田生产所遇到的问题设计了抽油机工况诊断系统整体结构框架。研究了理论示功图形成原理以及部分典型的故障示功图特征,并对示功图进行预处理,示功图的处理方式包括图像的灰度化、滤波、图像二值化以及边缘检测等,将示功图转化为边界最大区域填充的图像。其次,针对Hu矩特征值包含的信息不全面的问题,采用基于小波变换和矩特征相结合的小波不变矩方法来提取抽油机的示功图特征值。由于小波变换抗干扰性强以及反映局部信息的能力,因此采用小波不变矩对抽油机示功图进行局部和全局的特征提取,进而建立典型示功图样本库。再次,深入分析了支持向量机算法,支持向量机参数的选择及参数组合(惩罚因子c与核函数参数σ)会影响其分类精度,本文将引入萤火虫算法对其优化。为避免传统的萤火虫算法易出现局部最优解,特对其进行改进,将改进后的萤火虫算法应用到支持向量机的参数选择上,通过与改进粒子算法优化支持向量机以及传统萤火虫优化支持向量机相比较,基于萤火虫优化的支持向量机的分类效果有了很大的提高。最后,将改进后萤火虫算法优化的支持向量机应用到抽油机工况故障诊断中,建立支持向量机故障诊断模型。将测试数据输入到诊断模型中进行试验,实验结果表明基于改进萤火虫优化的支持向量机故障诊断方法具有较高的分类准确度。通过上述理论研究后,本文采用C#编程语言与ORACLE数据库相结合设计了抽油机工况故障诊断系统,并对其进行现场测试。测试结果表明,系统可以运行稳定,诊断结果准确。
[Abstract]:In the production process of oil fields at home and abroad, mechanical oil production still occupies a large part. Oil extraction equipment such as pumping unit is very important to the production of oil field. Because the pumping unit mostly works in the field for a long time in the bad environment such as high temperature high load and so on and the downhole working condition of the pumping unit is extremely complex resulting in the pumping unit often appear malfunction. The economic losses caused by the faults of pumping units are extremely huge, so it is necessary to diagnose the working conditions of pumping units. The machine learning method based on artificial intelligence can construct multi-class classification learning model and make fault diagnosis, which has become one of the hot research topics. Therefore, this paper presents a fault diagnosis method of pumping unit based on firefly optimized support vector machine. The main contents are as follows: firstly, the whole structure framework of pumping unit condition diagnosis system is designed to solve the problems encountered in oil field production. This paper studies the formation principle of theoretical indicator diagram and some typical fault indicator diagram features, and preprocesses the indicator diagram, which includes image grayscale, filtering, image binarization and edge detection, etc. Convert the indicator diagram into an image filled with the maximum boundary area. Secondly, in order to solve the problem of incomplete information contained in Hu moment eigenvalue, wavelet invariant moment method based on wavelet transform and moment feature is used to extract the eigenvalue of indicator diagram of pumping unit. Because wavelet transform has strong anti-interference and the ability to reflect local information, wavelet invariant moment is used to extract the local and global features of the indicator diagram of pumping unit, and then the sample library of typical indicator graph is established. Thirdly, the support vector machine (SVM) algorithm is deeply analyzed. The selection and combination of SVM parameters (penalty factor c and kernel function parameter 蟽) will affect the classification accuracy. In this paper, the firefly algorithm is introduced to optimize the SVM. In order to avoid the local optimal solution of the traditional firefly algorithm, the improved firefly algorithm is applied to the parameter selection of support vector machine. Compared with the improved particle algorithm optimization support vector machine and the traditional firefly optimization support vector machine, the classification effect of the support vector machine based on the firefly optimization has been greatly improved. Finally, the support vector machine (SVM) optimized by the improved firefly algorithm is applied to the fault diagnosis of pumping unit, and the fault diagnosis model of SVM is established. The experimental results show that the SVM fault diagnosis method based on improved firefly optimization has high classification accuracy. After the above theoretical research, this paper uses C # programming language and ORACLE database to design the pumping unit working condition fault diagnosis system, and carries on the field test to it. The test results show that the system can run stably and the diagnosis result is accurate.
【学位授予单位】：东北石油大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TE933.1

【参考文献】