METHOD FOR DETERMINING RESERVOIR PRESSURE IN VOLUME OF FIELD BASED ON ARTIFICIAL NEURAL NETWORKS Russian patent published in 2023 - IPC E21B47/06 G06N3/02 

FIELD: neural computing.

SUBSTANCE: claimed group of inventions is related to methods and systems for determining reservoir pressure in the volume of a deposit based on artificial neural networks, and to computer processing of the obtained data for modelling the physical properties of rocks. The method for determining reservoir pressure involves obtaining data including at least well trajectory data, logging data, physical properties of rocks along the wellbore and forming, based on the obtained data, a basic model that describes a preliminary assessment of the distribution of reservoir pressure along the wellbore. Next, a set of calibration data is prepared by direct or indirect measurements of reservoir pressure at specific points in the well trajectories and scaling of the input data. Then the structure of an artificial neural network is formed, consisting of at least data on the number of layers, the number of neurons in the layers, the number and values of input physical properties and the activation function. Next, an artificial neural network is pre-trained using scaled input data and a basic model that describes a preliminary assessment of the distribution of reservoir pressure along the wellbores. Then transfer training of the artificial neural network is carried out using scaled input data and calibration data obtained as a result of direct or indirect measurements of reservoir pressure at specific points in the well trajectories. Next, reservoir pressure values are obtained using a trained neural network, which have a minimum standard deviation from the calibration values in the delayed sample. The formation pressure determination system includes at least one processor, random access memory, and machine-readable instructions for executing the formation pressure determination method.

EFFECT: improves the accuracy of determining reservoir pressure in rocks depending on their physical properties.

19 cl, 6 dwg