抽油机运动周期精确测量方法研究

发布时间：2018-03-25 05:02

本文选题：示功图　切入点：加速度　出处：《河南科技大学》2017年硕士论文

【摘要】：示功仪是油田生产维护用的一种常用仪表,对示功图的分析是掌握油井产量、能耗、泵况等多种状态的重要手段。抽油机运动周期作为示功图的重要参数,其准确性直接影响示功图的精度。常用的加速度二次积分算位移的方法中,仅利用加速度信号求取的周期存在误差,而在油井抽油机正常工作过程中,悬点加速度、载荷、电动机功率及扭矩等物理量都是随着抽油机负载的变化而发生周期性变化的。基于此,本文提出一种多传感器融合的抽油机运动周期的计算方法。本文主要具体研究内容和结论概括如下:1、通过对抽油机井系统、悬点运动方程、理论示功图进行分析并结合现场实际环境,发现实测信号中不仅混有许多噪声,而且由于抽油机负载变化,上下冲程中加速度信号波峰有偏心,并不是简单的正弦波;此外加速度、载荷、电功率、扭矩等随负载变化的周期信号频率非常低;进一步论证了单一加速度信号计算位移的局限性。基于此提出了总体方案,选用抽油机的加速度、载荷、电功率信号作为输入,借助多传感器融合技术,设计抽油机运动周期的计算方法。2、为了提高周期的测量精度,对比分析多种滤波算法对加速度、载荷、功率信号预处理效果,通过理论分析和仿真实例证明Meyer小波更适合对低频非平稳加速度进行处理,卡尔曼滤波对功率信号有良好的处理效果,复合滤波算法对相对变化平稳的载荷信号也有良好的处理效果。3、通过对比时域的极值点归并法,频域M-Rife算法及函数拟合等周期算法,结果表明,频域M-Rife算法对功率信号周期计算更加准确;极值点归并法对载荷信号进行周期计算效果更好;对于拟合法,原始信号直接进行函数拟合,发现噪声干扰过大拟合效果不佳。对预处理后的信号选用不同阶次正弦函数类拟合叠加,通过比较残差平方和、确定系数、均方根来评价拟合效果优劣,对比后发现拟合法对加速度信号进行周期计算更加合适。4、通过多次理论仿真,并且在长时间、大量实验数据分析积累的基础上,分析了实际生产中某油井连续50次预测的加速度,载荷,功率数据,借助于MATLAB平台进行计算。在最终数据融合中,对比两种确定加权因子的方法优劣,确定选用基于模糊型指数信任度函数确定加权因子的方法计算出三种信号的权值,并最终计算出较准确的抽油机运动周期。对比其他方法本文提出的算法能得出更高精度的周期值,达到理想的测试效果。
[Abstract]:The dynamometer is a kind of commonly used instrument for oil field production and maintenance. The analysis of indicator diagram is an important means of mastering oil well production, energy consumption, pump condition and so on. The motion period of pumping unit is an important parameter of indicator diagram. The accuracy of the method directly affects the accuracy of the indicator diagram. In the commonly used methods for calculating displacement by quadratic integration of acceleration, there are errors only in the period obtained by using the acceleration signal, but in the normal working process of the oil well pumping unit, the acceleration and load at the suspension point exist. The power and torque of the motor vary periodically with the load of the pumping unit. In this paper, a multi-sensor fusion method for calculating the motion period of pumping unit is proposed. The main contents and conclusions of this paper are summarized as follows: 1. By analyzing the theoretical indicator diagram and combining with the actual environment, it is found that the measured signal is not only mixed with a lot of noise, but also the peak of acceleration signal is eccentric in the upper and lower stroke because of the variation of the load of the pumping unit, so it is not a simple sine wave. In addition, the frequency of periodic signals such as acceleration, load, electric power, torque and so on are very low. The limitation of calculating displacement by single acceleration signal is further demonstrated. Based on this, the overall scheme is proposed, and the acceleration of pumping unit is selected. Load, electric power signal as input, with the aid of multi-sensor fusion technology, the calculation method of pumping unit motion cycle is designed. In order to improve the measuring accuracy of the period, the comparison and analysis of various filtering algorithms for acceleration, load, Through theoretical analysis and simulation examples, it is proved that Meyer wavelet is more suitable for processing low frequency non-stationary acceleration, and Kalman filter has good processing effect on power signal. The composite filtering algorithm also has a good processing effect on the load signal with relatively stable variation. By comparing the time domain extremum merging method, the frequency domain M-Rife algorithm and the function fitting algorithm, the results show that, Frequency domain M-Rife algorithm is more accurate for power signal cycle calculation; extreme point merging method is more effective for load signal cycle calculation; for fitting method, the original signal is directly fitted by function. It is found that the noise interference is too large and the fitting effect is not good. Different order sinusoidal functions are used to fit the pre-processed signals. By comparing the sum of squared residuals, the coefficients and root-mean-square are determined to evaluate the fit effect. After comparison, it is found that the fitting method is more suitable for calculating the period of acceleration signal. Through many theoretical simulations, and on the basis of a long time, a large number of experimental data are analyzed and accumulated. The acceleration, load and power data of an oil well for 50 successive times in actual production are analyzed and calculated by means of MATLAB platform. In the final data fusion, the advantages and disadvantages of two methods for determining weighting factors are compared. The method of determining weighting factor based on fuzzy exponential trust function is chosen to calculate the weights of three kinds of signals. Finally, the accurate motion period of the pumping unit is calculated, and the period value with higher accuracy can be obtained by the algorithm proposed in this paper, which can achieve the ideal test effect.
【学位授予单位】：河南科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TE933.1

【相似文献】