May 27, 2022

pSeven at NAFEMS UK Conference 2022

pSeven will be presented at NAFEMS UK Conference 2022 by Applied Computing & Engineering Ltd

The NAFEMS UK Conference 2022, a regional conference for simulation engineers, will take place in Milton Keynes on June 7-8. It will bring end-users, developers, academics and industrialists together in one, independent gathering, the conference will feature several outstanding invited presentations, technical tracks, training courses and workshops under one roof.

Anthony Mosquera, Manager of Design Services at Applied Computing & Engineering Ltd, will give a talk on “Reliable Forecast of Gas Properties in Underground Storage” on 8 June at 16:10 (Local Time) and present pSeven based solution for Storengy, the European leader in industrial gas storage that allows monitoring of gas properties in the UGS is therefore needed to prevent injection of gas not responding to the specifications in the transportation network.

Abstract

Together with Liquefied Natural Gas (LNG) terminals and gas transportation network, Underground Gas Storages (UGS) are key modulation assets to satisfy the global Natural Gas demand in Europe, especially during winter time. Three types of UGS – depleted field, aquifer, and salt cavern – are exploited by Storengy, the European leader in industrial gas storage. We focus here on the aquifer type, which physics is fluid flow in porous media.

During production phase of those assets, the gas injected back from the UGS to the transportation network must fulfill some quality specifications (Yi properties) like minimal calorific value and low-enough H2S concentration level for instance. In aquifer UGS, some physical and chemical reactions might happen to the injected gas stream in the underground porous media, which finally modify the Yi properties of the UGS produced gas. A rigorous monitoring of gas properties in the UGS is therefore needed to prevent injection of gas not responding to the specifications in the transportation network. Moreover, gas storage volumes in UGS are time dependent and cyclic, with high levels in the summer (due to gas injection and low gas demand), and lower gas stock in the winter (due to gas production to address increasing gas demand). For a proper management of the UGS portfolio and to anticipate corrective actions, it is crucial to have reliable forecasts over time of gas properties produced from the UGS.

The physics behind gas properties alteration in aquifer UGS is well understood but associated numerical models are not available as such. We can however identify a set of Xi parameters (e.g. CO2 gas concentration levels, number of active production wells, water production rate) that determine the Yi properties. As obtaining an estimate of a Yi property based on some Xi observations is not possible by simply solving a numerical model, we propose here an alternative method. It capitalizes on raw measurement data on which we perform dependency analysis and approximation model training in the pSeven integration platform to:

• Compute the Sobol indices and assess the impact of each Xi parameter on a Yi property
• Build a reliable predictive model, able to forecast any Yi property over a period of interest

This approach implements agnostic and bias-free machine learning techniques to raw data that come directly from UGS measurements. Therefore, it is complementary to other Storengy internally developed approaches and by cross-checking, it allows to reinforce our confidence in gas properties forecasts in UGS.

To NAFEMS UK Conferencenavigate_next

About AC&E