Multifactor Evaluation Technology for Casing Damaged Wells

Abstract

Abstract Being a maturing field that has been developed for over 40 years, Daqing Oilfield has nearly 40 years of water flooding development history. One of the major problems that Daqing Oilfield is currently faced with is severe casing damage of both producers and injectors. Factors influencing casing damage are numerous, heightening difficulties in prediction and comprehensive analysis. In the past, casing damage mechanism research emphasizes on casing strength in respect of well drilling, fault and shale in respect of geology, and water injection pressure in respect of reservoir. Yet the accuracy for trouble shooting based on single factor analysis is quite low. To conduct casing damage mechanism research, the influence of multi-factors on casing damage should be considered comprehensively to draw a correct conclusion. Therefore, research on multi-factor evaluation technology and its application in casing damaged wells is implemented. Three-layer feed forward neural network model is used to make the casing damage formation factors corresponding to the input vectors of the neural network. Factors causing casing damage such as water invasion in mudstone, pressure performance, geological features, casing performance, and stress field etc. are described as a series of real numbers between 0 and 1. More often than not, these factors are difficult to be described through mathematical methods, but a blurry value can be given by experts. Thus, a blurry vector for casing damage factor is composed of multi-factors, that is to say, the output of neural network is correspondent to the present working status of casings, which can be given by experts based on their experiences, and the output value is a description of the present working status of the casings. The given casing damage data is used as instructor signal to train the network, and following training, the network can perform the prediction of the working status of casings.