Economic justification of the choice of development of an explored oil and gas field

Abstract

The paper presents a study that combines two methods: clustering of fields using machine learning algorithms and building a regression model for forecasting the cost and economic efficiency of an oil and gas field. In the course of this study, a neural network of direct propagation was developed, which is used to forecast the cost of developing wells in oil and gas fields, taking into account all technical parameters. The resulting neural network, formed on the basis of algorithms of input data, forms output signals when any set of input signals of the training set is applied to the input of the network. The resulting neural network expresses patterns that are present in the input data. This network turns out to be the functional equivalent of some model of dependencies between variables. Indicators of 15 wells were used to create the ANN model. The main task of the model is to determine the cluster of a new deposit. Conventional designations (x - for exogenous (explored factors and y - for actual calculated cost data). Input data (x) for neural network training were: The smallest thickness of oil-bearing formations; The largest thickness of oil-bearing formations; Gas factor; Reservoir temperature; Porosity; Permeability ; Oil and gas content in the formation; Occurrence depth; Water flow rate; Oil flow rate; Gas flow rate; Volume of extracted oil; Volume of extracted gas; Design depth of the well. After dividing the educational sample into classes, a model of the dependence of the following factors on the input was created for each cluster: Approximate cost of well construction without VAT (thousands of dollars); Estimated cost of well construction including VAT (thousands of dollars); Profit (thousands of dollars); Cost of well construction (thousands of dollars); The cost of 1 m of penetration without VAT (dollars); The cost of 1 m of penetration including VAT (dollars). Data, on the training sample for which these indicators were known, the accuracy of the forecast was checked. The error did not exceed 5%. Then, calculations were made for explored wells, but those where the economic indicators are unknown. Based on the calculated well development cost values, the efficiency factor was calculated as a fraction of the predicted development cost divided by the explored reserves. And it is recommended for development that explored field, which has the lowest indicator.