Advances in Virtual Flow Metering Using Deep Composite Lstm-Autoencoder Network for Gas-Condensate Wells

Abstract

AbstractIn terms of cost and execution time, data-driven Virtual Flow Meters (VFM) are alternative solutions to traditional well testing (WT) and physical multiphase flow meters (MPFM) for production rate determination which is needed for critical decisions by operators but faced with the challenge of low accuracy due to the transient and dynamic state of multiphase flow systems. Recently, some progress has been recorded by training steady-state feed-forward neural networks to learn to approximate production rate based on a certain number of input features (e.g., choke opening, pressure, temperature, etc.) without any recursive feedback connection between the network outputs and inputs. This disconnection has impacted their accuracy. Dynamic artificial neural network, for example, the recurrent neural networks (RNN), e.g., LSTM has shown good performance as their architecture allows for the usage of data from the past time step to predict the current time step. Forecast accuracy for RNNs is limited to a short period due to their inherent vanishing gradient issues. While a majority of VFM applications have been developed for oil and gas systems, little or non is applied to gas condensate systems.In this project, a sequence-to-sequence deep composite LSTM-Autoencoders neural network (LSTM-A-NN) was explored and used to demonstrate the ability to leverage its architecture to accurately predict multiphase flow rate for highly dynamic multiphase flow phenomenon associated with retrograde condensate reservoirs. Data used in training and validating the LSTM-A-NN were generated from simulations. First, a 3D compositional simulator (ECLIPSE 300) was used to simulate, as close as possible, a realistic case of a compositional reservoir with flow from the subsurface to the wellhead to generate production rate data. Secondly, an integrated production system was built using GAP to simulate a coupled material-balanced-based inflow with wells and a surface separator. The production output data in this case includes production rates, wellhead pressure, bottom-hole pressure, temperatures, condensate-gas ratio, etc. For both cases, the LSTM-A-NN performance was impressive (mean square error less than 1) and demonstrated its flexibility and scalability with an increasing number of input features (production data). The LSTM-A-NN learns the physics of complex fluid flow through non-linear dimensionality reduction while passing the temporal sequence of production data through its encoder network. The encoded data representation is thereafter decoded and reconstructed such that the output is in the same dimension as the input.The ability to leverage some advanced artificial intelligence frameworks such as a composite LSTM-A-NN has proven that it is possible to achieve the desired accuracy needed in data-driven VFM to meet the requirement of low cost, low execution time, and high accuracy.This project has also demonstrated the ability of the data-driven model to learn the complex dynamics within the temporal ordering of input sequences of production data, with an internal memory adapted to remember or use information across long input sequences, hence, yielding longer and more reliable forecasts, unlike other networks.