The crustal thickening due to India-Asia collision was previously believed to play a key role in uplift of plateau. However, recent paleoaltimetry data indicates parts of the plateau reached high elevations earlier than previously thought. Accurate determination of crustal thickening history is critical to further constrain this issue. We applied machine learning to model crustal thickness using refined global geochemical datasets, addressing limitations in earlier approaches. The crustal thickness evolution history of Gangdese terrane in southern Tibet was recovered by the new machine learning model and compared with the paleoaltimetry data. Results show that at the initial stage of collision, the crust thickness was normal and decoupled with high paleoaltimetry, while the subsequent crustal thickening occurred after 41 Ma and kept the same path with the uplift of plateau. This challenges the notion of rapid uplift driven by crustal thickening and instead suggests a protracted uplift process, offering new perspectives on the geodynamics of southern Tibetan Plateau.

Horizontal salt caverns are ideal places to store energy. The distribution of concentration fields and flow fields is critical for the shape prediction of horizontal salt caverns. However, the area between the deviated well and the injection segment is regarded as an inactive segment in the current simplified model. The simplified model assumes that there is no freshwater inflow and that rock salt is not dissolved in the inactive segment, which is quite different from the engineering site. To improve the accuracy and efficiency of the leaching simulation, a program is developed using a modified flow field model. The concentration fields and flow fields are simulated by solving the Navier-Stokes equations and convective diffusion equations. Simulation results show that injected freshwater continuously mixes with brine and forms a buoyant jet during the upward floating. After reaching the cavern top, the mixed flow first flows along the wall and then gradually flows downward to the left and right sides of the inlet. The concentration is the lowest at the cavern top above the inlet and increases from the inlet to the left and right sides. To describe these patterns, the modified flow field model divides the flow field into four segments and considers the dissolution of the "inactive segment". The simulation program is written with the dynamic mesh method and leaching experiments are simulated for verification. Simulated cavern shapes coincide well with actual shapes, indicating that the modified model and the program are reliable.

Horizontal salt caverns are widely used for oil and gas storage, hydrogen storage, compressed air energy storage, and carbon dioxide geological storage in bedded salt. Accurate modeling of solution mining processes is critical for predicting and controlling the shape and capacity of the caverns. To improve the accuracy and efficiency of such predictions, a novel construction prediction model is proposed using a simplified concentration field distribution model. The dissolution of salt and the convection of brine are simulated by solving the Navier-Stokes equations and the convective diffusion equation. Simulation results indicate that the injected freshwater mixes with the brine and forms an upward buoyant jet. Except in the buoyant jet region, there is a vertical concentration stratification. The brine concentration is the lowest from the apex of the buoyant jet to the cavern top, and gradually increases in the downward direction until the saturated zone below the injection port. In this regard, the simplified concentration field distribution model introduces a distribution factor /i to describe these distribution patterns. A C++ simulation program, named "Horizontal Salt Cavern Construction Prediction", is developed combining the simplified concentration model, the dissolution model and the grid system based on the Volume of Fluid Method. The construction of Volgograd Gas Storage is simulated for verification. Simulated brine concentration and cavern shapes coincide well with the actual cases, indicating that the proposed model is reliable. Work in this paper can serve as a new construction design tool and facilitate renewable energy storage in salt caverns.

Construction simulation can help design and control an energy storage salt cavern with a regular shape, which benefits the storage capacity and the long-term operational safety. However, the conventional grid discretization methods using elastic grid concepts faced challenges in accurately tracking intricate boundary movements. This paper introduced a simulation model of salt cavern construction based on the structural grid of the Volume-of-fluid method, and realized the tracking of the salt caverns’ moving boundary surface. The model was validated using the results of indoor experiment and Volgograd horizontal cavern. In the simulation of indoor experiment, the lateral expansion at the top of the inlet, which was difficult to reproduce in previous models, was successfully simulated. The volume of the simulation chamber was 392.8ml with an error of 1.9%. The simulation results validate the capability of the proposed method in free boundary tracking. Furthermore, the Volgograd horizontal cavern construction project is simulated. The cavern shape is close to the sonar detection, with an average error in radius about 3.6%. And the brine-discharge concentration is consistent with the site monitoring, with an average error about 4.5%. The proposed model offers a dependable tool for simulating the leaching process of horizontal salt cavern.

In the numerical simulations of multi-phase flow using Volume-of-Fluid (VOF) method, the calculation of the interface normal is a crucial point. In this paper, a machine learning method is used to develop an artificial neural network (ANN) model to make more accurate prediction of the local normal vector from neighboring volume fractions. Spherical surfaces with different radii are intersected with a structural background grid to generate 84328 groups of data: 3×3×3 neighboring volume fractions are used as input, and normal vector as output. Using 90% data as training dataset, the ANN model is well trained by optimizing the number of hidden layers and the number of neurons on each layer. Using the remaining 10% data, normal predictions are made using ANN-VOF and the most used YOUNG and HEIGHT-FUNCTION methods. The RMSE of the ANN-VOF/YOUNG/ HEIGHT-FUNCTION methods are 0.008/0.022/0.045 respectively. In the reconstruction of a sinusoidal surface, the MSE of the ANN-VOF/YOUNG/ HEIGHT-FUNCTION methods are 0.008/0.018/0.041. It is demonstrated that the ANN-VOF method has better performance for interface normal prediction. The proposed method has a simple computational logic and does not need to deal with complex geometric topology, which lays the foundation for application in other more complex grids.

In recent years, fluid convection has played an increasingly important role in environmental problems, which has attracted increasing attention. Scaled physical modeling is an important approach to understand the behavior of fluid convection in nature. However, a common source of errors is conflicting similarity criteria. Here, we present using hypergravity to improve the scaling similarity of gravity-dominated fluid convection, e.g., natural convection and multiphase flow. We demonstrate the validity of the approach by investigating water-brine buoyant jet experiments conducted under hypergravity created by a centrifuge. Considering the influence of the Coriolis force, an evaluation and correction method is presented. Results show that the scaling similarity increases with the gravitational acceleration. In particular, the model best represents the prototype under N(3)g with a spatial scale of 1/N and a timescale of 1/N-2 by simultaneously satisfying the Froude and Reynolds criteria. The significance of centrifuge radius and fluid velocity in determining the accuracy of the scaled model is discussed in light of the Coriolis effect and turbulence. This study demonstrates a new direction for the physical modeling of fluid subject to gravity with broad application prospects.

A multi-step horizontal salt cavern (MSHSC) for energy storage has been drawing more attention, using retractable water-injection tubes to mine larger storage spaces between two boreholes. The retract distance per step best determines the shape of the cavern roof, and thus has an important influence on the stability of MSHSCs. This paper discusses the impact of the step distance on the stability and economy of MSHSCs. The geometry of MSHSCs with different step distances is obtained using our previously developed solution mining model. Geo-mechanical models are developed based on the geometry and the geo-conditions of Jintan salt mine in China. Then a series of static creep analyses are performed using FLAC3D software. Results show that smaller step distances mean higher stability risks during long-term operation. As the step distance is reduced, both the volume shrinkage and plastic zone volume increase. In the displacement contours after 30 years of operation, the large displacement zone increases significantly with decreasing step distance. However, the cavern volume and expected revenue increase with decreasing step distance, contrary to the stability parameters. Economic analysis with integrated stability considerations is conducted to give a balanced step distance. Volume shrinkage is considered in the calculation of effective cavern volume. Plastic zone volume is reflected in the space cost of MSHSCs, since it influences the thickness of the cavern top plate and the pillar width between adjacent caverns. Further, a comprehensive indicator of economic cost/expected revenue per unit volume of rock salt versus time is proposed, calculated, and compared. According to the comparison results, a medium step distance of 100 m or 125 m is recommended for the design of MSHSCs in a given salt layer, with the stability and economic considerations best balanced.

The construction design and control of energy storage salt caverns is the key to ensure their long-term storage capacity and operational safety. Current experimental and numerical design/optimizing methods are time-consuming and rely heavily on engineering experience. This paper proposes a machinelearning-based method for the rapid capacity prediction and construction parameter optimization of energy storage salt caverns. We propose a data generation method that uses 1253 sets of random construction parameters as input. The resulting capacity/efficiency-concerned effective volume (V) and maximum radius (rmax) obtained by our numerical program are the output. A back-propagation artificial neural network model for salt cavern construction prediction (BPANN-SCCP) is trained on the dataset. The cross-validated mean absolute percentage error (MAPE) of the BPANN-SCCP predicted Vis 1.838%, that of the predicted rmax is 3.144%. This accuracy meets the engineering design requirements, and the prediction efficiency is improved by about 6 x 107 times. Using this model, a design parameter optimization method is devised to optimize 3 sets of design parameters from a million random ones. The resulting caverns are regular in shape with larger capacity ratio than 3 field caverns in Jintan Salt Cavern Gas Storage, verifying the reliability of the proposed optimization method. (c) 2022 Elsevier Ltd. All rights reserved.