نوع مقاله : مقاله پژوهشی
نویسندگان
1 دانشکده مهندسی مکانیک، دانشگاه علم و صنعت ایران
2 دانشگاه علم و صنعت ایران، دانشکده مهندسی شیمی، نفت و گاز
چکیده
کلیدواژهها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
Estimating the flow rate in oil wells of a field is a vital and practical process. However, the flows extracted from oil wells are multiphase, and their accurate estimation is highly challenging and costly. Virtual flow meters, compared to multiphase flow meters and well-testing methods, are an economically viable option that can accurately predict future flow rates by leveraging existing data and artificial intelligence algorithms. Therefore, data-driven virtual flow meters have recently received significant attention. This paper estimates the production flow rate of a well using three machine learning algorithms: 1- k-nearest neighbors; 2- gradient boosting; and 3- decision tree, using pump data. Pearson and Spearman statistical analyses were used to select appropriate features as the algorithm inputs. The dataset under investigation pertains to one of the wells of a southern oil field in Iran. The available dataset has a small volume and insufficient diversity, but despite this, the results show that the proposed algorithms perform well. The k-NN method, with an accuracy of 0.9494, performed better than the other two methods in estimating oil flow rate. To examine the performance of the algorithms against noisy data, one percent of standard deviation noise was added to the input data. The investigations showed that the k-NN model, with an accuracy of 0.9257, performed better than the other two methods and was least affected by the noise.
کلیدواژهها [English]