Casing Damage Forecasting Method And Its Application In The Process Of Oilfield Development

With the continuous development of oilfield, casing damage is becoming more and more serious gradually around the world, which has caused enormous economic loss and hindered the routine production severely. Therefore, a reasonable method of forecasting casing damage early and the appropriate preventive measures should be adopted to extend casing life, reduce the rate of casing damage and then to achieve efficient production. So what this paper researched is of great significance.Casing damage problem is a highly complex large-scale system, which is characterized by uncertainty, ambiguity and time-varying. The basic theory of forecasting and the common forecasting method are introduced from the perspective of system theory firstly. The method of support vector machine is optimally selected by comparative analysis of common forecasting methods, which is particularly suitable to solve the small sample, nonlinear and high dimension problem. Then the underlying factors causing casing damage in the process of oilfield development have been studied systematically and comprehensively and casing damage mechanism is analyzed from different perspectives of casing damage based on field practice. The results indicate that casing damage is the combined effect of many factors in the process of oilfield development. The main mechanism is the change of development parameters, such as local injection pressure, injection-production ratio, water injection strength and the injection-production corresponding situation causing asymmetric pore pressure, which makes the force status of casing changing and then causes different degrees of casing damage. The sensitive factors causing casing damage were identified in the process of oilfield development after the attribute reduction to the influencing factors of casing damage were conducted by the rough set theory on the ROSETTA platform and the weight coefficients of the attributes were figured out by the concept of the significance of attributes. Taking the reduced attributes as the input variables of support vector machine, aiming at limitations of traditional support vector machine to solve the time-varying problem, dynamic forecasting methods based on time series integrated with support vector machines was put forward and casing damage dynamic forecasting model was established in the development process. The impact of kernel function, kernel parameters, sample numbers and sample dimension on forecasting accuracy was studied. Aiming at corrosion problem of shallow casing, simulating formation water of different components were prepared and the influence of calcium concentration, magnesium ion concentration, bicarbonate ion concentration, chlorine ion concentration, salinity, dissolved oxygen, pH value and temperature on the corrosion of casing steel were investigated. Gray correlation analysis method was used to study the impact degree of various factors on corrosion of casing steel and the casing corrosion forecasting model was established based on support vector machine regression algorithm. The results indicate that the method of casing corrosion forecasting can obtain good results, and its forecasting accuracy is superior to that of regression analysis and neural network methods.On the basis of theoretical research, a practical and reliable casing damage dynamic forecasting analysis software system was developed with SQL Server database and Csharp language. The application results in Yushulin oilfield and Gangxi oilfield showed that this system has the advantages of simplicity, convenience and high reliability, which is worth widely popularizing and applying.