Optimal Hydrate Inhibition Policies with the Aid of Neural Networks

Abstract

Hydrates are known to occur in a variety of natural-gas handling facilities and processing equipment in oil fields, refineries, and chemical plants when natural gas and water coexist at elevated pressure and reduced temperature. Prevention of hydrate formation costs large amounts of capital and results in large operating expenses. Hydrate inhibition using chemical inhibitors is still the most widely used method. Accurate prediction of hydrate inhibition is required for cost-effective design and operation. Available models have limitations in ranges of application and types and compositions of the fluids and inhibitors used. This paper describes the development and application of neural networks for the prediction and optimization of naturalgas hydrate inhibition. Neural network models have been used to accurately determine the temperature depression of gas hydrates for a variety of types and concentrations of inhibitors. Experimental data covering wide ranges of hydrate formation conditions, gas compositions, and concentrations of various types of inhibitors have been used in model validation. The factors that may affect the inhibition process, such as gas gravity and pressure, were investigated. An optimization study has been carried out on the selection of inhibitor type and concentration using the developed neural network models. Optimization was based on economical and technical performance considerations concerning inhibitor losses in vapor and liquid hydrocarbons. The results indicate that optimal design depends on water content, operating conditions of pressure and temperature, and gas composition. Optimized hydrate inhibition strategies have been recommended for various gas composition systems.

Artificial Neural Network-based detection of gas hydrate formation

“…The model takes the evaporation of the inhibitor into consideration. The devised method suggests inhibitor injection ratios for gases of various composition [15].…”

Experimental Simulation of the Exploitation of Natural Gas Hydrate

“…When the concentration of methanol and ethanol is below 5 wt%, their injection could promote the formation of hydrate. Elgibaly and Elkamel [100] pointed out that, compared to ethanol, ethylene glycol has lower volatility and stronger hydrogen bonding with water, therefore, ethylene glycol is more beneficial for recycling than ethanol. They also believed that, to some extent, the inhibiting effect of electrolyte on hydrate is different with ethanol and ethylene glycol.…”

Toward a Robust, Universal Predictor of Gas Hydrate Equilibria by Means of a Deep Learning Regression

“…Models have been developed to predict hydrate equilibrium conditions where thermodynamic inhibitors and electrolytes are present. Elgibaly and Elkamel 20 developed models with up to 16 inputs, including gas composition, hydrogen sulfide, and inhibitor and electrolyte concentration, later expanded on in Elgibaly and Elkamel, 21 accounting for inhibitor costs. Employing a train-test-validate split, Chapoy et al 15 developed a neural network with 19 inputs from a dataset of over 3000 equilibrium points, including hydrate structure, gas compositions, and inhibitor and electrolyte concentration, while excluding methane molar gas-phase compositions less than 50%.…”