Forecasting inflow after hydraulic fractuaring of oil formation using artificial intelligence

Relevance. The effectiveness of hydraulic fracturing is determined by many features that are currently not taken into consideration in traditional formulas for calculating hydraulic fracturing. Thus, in traditional calculation methods, which are based on one-dimensional models of crack propagation (the work considered the Khristianovich–Girtsma–de Klerk model, the radial model and the Perkins–Kern–Nordgren model), assumptions are made to calculate various parameters, which in some cases will affect the accuracy of the calculation. Because of this, the effectiveness and necessity of hydraulic fracturing might under question. Aim. To predict hydraulic fracturing using artificial intelligence technologies such as machine learning techniques and then compare the accuracy of the prediction with the accuracy of traditional calculations using one-dimensional models to identify the most accurate technique. Object. Model of oil production growth after hydraulic fracturing. Methods. Development of special software for predicting oil production after hydraulic fracturing using traditional methods based on one-dimensional models and using machine learning techniques. Results. Traditional univariate models were 84% accurate, while machine learning models were 87% accurate. Conclusion. The use of artificial intelligence technologies, namely machine learning, provides more accurate hydraulic fracturing prediction compared to traditional methods based on one-dimensional models.