基于人工神经网络的储油罐用温度传感器非线性补偿技术研究

发布时间：2018-01-21 08:11

本文关键词： BP神经网络储油罐用温度传感器非线性补偿　出处：《东北石油大学》2015年硕士论文　论文类型：学位论文

【摘要】：大型储油罐对国家战略储备具有非常重要的意义,研究储油罐温度场分布规律,能为实际生产中制定安全、合理的储存方案提供科学指导。现阶段,储油罐多采用温度传感器进行温度数据测量,该方法测量性能稳定、示值复现性高,但测量结果存在一定的非线性误差。为提高测量精度,有必要对储油罐用温度传感器进行非线性补偿。本文提出了一种非线性补偿方法,即基于人工神经网络的储油罐用温度传感器非线性补偿。本文首先对储油罐温度场监测系统及其温度测试数据进行初步分析,得出储油罐温度场分布规律,依此建立储油罐内部空间区域划分模型。通过对多组算例进行计算,对影响特征阀值较明显的因素采用了加权平均的方法,得到了罐内空间区域边界特征阀值,将储油罐内部空间划分为罐顶区域、中心区域和罐底区域。在此基础上,根据BP神经网络原理和相应区域内储油罐用温度传感器的测量数据,设计了用于补偿罐顶区域、中心区域和罐底区域温度传感器的BP神经网络的结构、函数等各项参数。其中网络隐含层节点数目采用对比网络输出的均方误差(MSE)的实验方法确定,网络训练用样本数据通过恒温设备的测定获得,并在训练网络前对其进行了预处理。本文利用Matlab软件建立了三个用于储油罐用温度传感器非线性补偿的BP神经网络模型,并分别给出三个BP神经网络训练后的各项性能。对输出结果及温度场监测系统使用数学公式法计算得到的温度测试结果与高精度测温仪采集得到的温度数据进行对比分析,结果表明,网络输出精度较高,性能较好,能够应用在储油罐温度场测试领域中。
[Abstract]:Large-scale oil storage tank is of great significance to the national strategic reserve. Studying the temperature field distribution law of oil storage tank can provide scientific guidance for the formulation of safe and reasonable storage plan in actual production. Most oil tanks use temperature sensor to measure temperature data. This method has stable performance and high reproducibility, but there is a certain nonlinear error in the measurement results, in order to improve the accuracy of measurement. It is necessary to make nonlinear compensation for the temperature sensor used in oil storage tank. In this paper, a nonlinear compensation method is proposed. In this paper, the temperature field monitoring system and temperature testing data of oil storage tank are analyzed preliminarily, and the distribution law of temperature field of oil tank is obtained. According to this, the model of internal space zone division of oil storage tank is established. Through the calculation of multiple examples, the weighted average method is used to calculate the factors that affect the characteristic threshold value, and the boundary characteristic threshold value of the inner space area of the tank is obtained. The inner space of the tank is divided into the top area, the center area and the bottom area. On this basis, according to the principle of BP neural network and the measurement data of the temperature sensor used in the oil storage tank in the corresponding area. The structure of BP neural network is designed to compensate the temperature sensor in the top, center and bottom of the tank. The number of nodes in the network hidden layer is determined by the experimental method of comparing the mean square error (MSE) of the network output, and the sample data for network training are obtained by measuring the constant temperature equipment. In this paper, three BP neural network models for nonlinear compensation of temperature sensor used in oil storage tank are established by using Matlab software. The performance of three BP neural networks after training is given respectively. The temperature test results calculated by mathematical formula method and temperature data collected by high precision thermometer are used to input the output results and temperature field monitoring system. A comparative analysis. The results show that the network has higher output precision and better performance. It can be used in the field of oil tank temperature field measurement.
【学位授予单位】：东北石油大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TE972

【参考文献】