Hu Huang

Dr. Hu Huang


Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
7 July 2023 in SPE Journal
Verified
Journal Article
Closed access
PDF file
HHHu Huang
Hu Huang
BGBin Gong
Bin Gong
WSWenyue Sun
Wenyue Sun
Summary: A new deep-learning-based surrogate model is developed and applied for predicting dynamic oil rate and water rate with different well controls. The surrogate model is based on the graph neural networks (GNNs) and long-short-term memory (LSTM) techniques. The GNN models are used to characterize the connections of injector-producer pairs and producer-producer pairs, while an LSTM structure is developed to simulate the evolution of the constructed GNN models over time. In this way, we use geological attributes at wells and well controls with different times as input data. The oil rates and water rates at different times are generated. In this study, the GNN-LSTM surrogate model is applied to a high dimensional oil-gas-water field case with flow driven by 189 wells (i.e., 96 producers and 93 injectors) operating under time-varying control specifications. A total of 500 high-fidelity training simulations are performed in the offline stage, out of which 450 simulations are used for training the GNN-LSTM surrogate model, which takes about 150 minutes on an RTX2060 GPU. The trained model is then used to provide production forecasts under various well control scenarios, which are shown to be consistent with those obtained from the high-fidelity simulations (e.g., around 4.8% and 4.3% average relative errors for water production rates and oil production rates, respectively). The online computations from our GNN-LSTM model take about 0.3 seconds per run, achieving a speedup of over a factor of 1,000 relative to the high-fidelity simulations, which takes about 363 seconds per run. Overall, this model is shown to provide reliable and fast predictions of oil rates and water rates with a large level of perturbations in the well controls. Finally, the proposed GNN-LSTM model, in conjunction with the particle swarm optimization (PSO) technique, is applied to optimize the field oil production by varying the well control schedule of all injectors. Due to the significant speedup and high accuracy of the proposed surrogate model, the improved well-control strategies can be efficiently obtained.

5 September 2023 in SPE Journal
Verified
Journal Article
Closed access
PDF file
HHHu Huang
Hu Huang
BGBin Gong
Bin Gong
WSWenyue Sun
Wenyue Sun
FQFeng Qin
Feng Qin
STShenglai Tang
Shenglai Tang
HLHui Li
Hui Li
Summary: The embed-to-control (E2C) framework provides a new deep-learning-based reduced-order modeling framework for much faster subsurface flow predictions than traditional simulation. However, the previous E2C model entails a large number of model parameters, which limits its applicability to large-scale cases. In addition, the previous E2C model has not been applied to a gas-driven subsurface system or well-control optimization. In this work, we make several improvements to the previous E2C framework for more complex and larger-scale problems. First, we reduce the output dimension of the middle layers by increasing the number of downsampling layers and using the depth-wise separable (DWS) convolution techniques in the deconvolution operation. Second, we use the global average pooling (GAP) technique to reduce the model parameters. Third, we apply an “add” operation in the skip connection to fuse the features. The improved E2C surrogate model is applied to a high-dimensional gas system with flow driven by six wells operating under time-varying control specifications. In this case, we can reduce the graphics processing unit (GPU) memory usage from 19.22 GB to 2.57 GB. In the training process, a total of 160 high-fidelity simulations are performed offline, out of which 130 simulation results with partial time sequence are used for training the E2C surrogate model, which takes about 46 hours on an RTX 3090 GPU. The trained model is shown to provide accurate production forecasts under various well control scenarios during the prediction period. The online computations from our E2C model are about 6.5 seconds per case, which achieves a speedup of more than 500 factors to corresponding full-order simulations, which take about 1 hour per run. Finally, the improved E2C model, in conjunction with a particle swarm optimization (PSO) technique, is applied to optimize the injection well strategies of an oil-gas-water field case with 189 wells (i.e., 96 producers and 93 injectors). Due to the significant speedup and high accuracy of the improved surrogate model, it is shown that improved well-control strategies can be efficiently obtained.

5 January 2023 in Geoenergy Science and Engineering
Verified
Journal Article
Closed access
HHHu Huang
Hu Huang
BGBin Gong
Bin Gong
YLYimin Liu
Yimin Liu
WSWenyue Sun
Wenyue Sun

10 February 2026 in Energies
Verified
Journal Article
Open access
PDF file
HHHu Huang
Hu Huang
GBGong Bin
Gong Bin
ZLZhengkai Lan
Zhengkai Lan
SCShaoying Chang
Shaoying Chang
The advancement of Industry 5.0 hinges on the deep integration of artificial intelligence (AI) with domain expertise to foster sustainable industrial development. This study proposes a deep learning-based surrogate modeling framework that integrates reservoir production requirements with AI technologies, providing intelligent decision support for production optimization and enhanced efficiency. To evaluate AI’s effectiveness in complex industrial scenarios, we conduct an integrated analysis encompassing model construction, dynamic prediction, and production optimization using a real-world oilfield case. This oilfield features a dynamically increasing number of wells and requires dynamic adjustments to injection–production relationships. To address this challenge, we enhance the Embed-to-Control model by improving the nonlinear representation capability within its decoder structure. Subsequently, we construct a high-fidelity dataset containing 300 samples for model training and testing. The experimental results demonstrate that the proposed improved model achieves a high accuracy in predicting key state variables (pressure and saturation) and oil production. Regarding computational efficiency, a single model run requires only approximately 17.3 s, achieving an over 200× speedup relative to traditional numerical simulators. Finally, we coupled the trained surrogate model with the particle swarm optimization algorithm to optimize the injection well control strategy. The optimized scheme increases daily oil production by 13.84%, boosting economic benefits. This study demonstrates a practical technological pathway to accelerate the oil and gas industry’s transition toward Industry 5.0.

14 March 2026 in International Journal of Modern Physics C
Verified
Journal Article
Closed access
HHHu Huang
Hu Huang
MZMingzhen Zheng
Mingzhen Zheng
LDLi Dinggen
Li Dinggen
Lei WangLei Wang
Lei Wang

1 October 2026 in SPE Journal
Verified
Journal Article
Closed access
HWHaoqiang Wu
Haoqiang Wu
HHHu Huang
Hu Huang
BGBin Gong
Bin Gong
CLChen Liu
Chen Liu
LZLili Zuo
Lili Zuo
Summary: History matching of continuous 3D geological models is limited by the dimensionality of the parameter field, which drives both the computational cost of the inversion and its ill-posedness. This paper develops an efficient data assimilation framework in which the geomodel is represented through an orthogonal spectral decomposition and updated with the ensemble smoother with multiple data assimilation (ES-MDA). The heterogeneous 3D property field is transformed into the frequency domain, where the macroscopic geological trend and the spatial continuity are carried by a very small fraction of low-frequency coefficients of the order of 1% of the spectrum. Because the transform basis is analytically predefined rather than inferred from the ensemble, the parameterization is constructed once, at negligible cost, and remains fixed throughout assimilation. A Pearson correlation screening is then applied to the retained coefficients to isolate the subspace to which the dynamic response is most sensitive, which limits overparameterization and the associated noise-induced ensemble collapse. The framework is evaluated on two contrasting real-field models: a tight sandstone gas reservoir matched on bottomhole pressure (BHP), and a faulted oil reservoir in which three fault transmissibility multipliers are inverted jointly with the property field. In both cases, an effective subspace of only five to seven degrees of freedom is sufficient to reproduce the observed production history. A hold-out experiment on the first case, in which the final quarter of the record is withheld from assimilation and predicted by the matched ensemble, reduces the mean relative pressure error over the unobserved period from 69.0% for prior to 20.8% for the posterior, confirming that the calibrated models retain predictive skill; the same experiment shows that the posterior ensemble spread is narrower than the residual forecast error warrants. Benchmarking against the higher-order singular value decomposition (HOSVD), an ensemble-derived tensor basis, yields comparable energy compaction and inversion accuracy while avoiding the tensor decomposition itself, giving a 13.6-fold speedup in parameterization and a 6.4-fold speedup in single-model reconstruction. The resulting workflow provides a practical, scalable, and physically interpretable route to efficient 3D reservoir characterization.