Cyclic steam stimulation (CSS) is successfully applied to increase heavy oil recovery in heavy oil reservoirs in Bohai Bay, China. However, during the CSS processes, hydrogen sulfide (H₂S) was detected in some heavy oil reservoirs. The existing literature mainly focused on the H₂S generation of onshore heavy oil. There is no concrete experimental data available, especially about the level of H₂S generation during CSS of offshore heavy oil. In addition, there is still a lack of effective reaction kinetic models and numerical simulation methods to simulate H₂S generation during the CSS of offshore heavy oil. Therefore, this paper presents a case study from Bohai Bay, China. First, the laboratory aquathermolysis tests were conducted to simulate the gases that are produced during the CSS processes of heavy oil. The effects of the reaction temperature and time on the H₂S generation were studied. Then, a one-dimensional CSS experiment was performed to predict H₂S generation under reservoir conditions. A kinetic model for the prediction of H₂S generation during the CSS of heavy oil was presented. The developed model was calibrated with the experimental data of the one-dimensional CSS experiment at a temperature of 300 °C. Finally, a reservoir model was developed to predict H₂S generation and investigate the effects of soaking time, steam quality, and steam injection volume on H₂S generation during CSS processes. The results show that the H₂S concentration increased from 0.77 ppm in the first cycle to 1.94 ppm in the eighth cycle during the one-dimensional CSS experiment. The average absolute error between the measured and simulated H₂S production was 12.46%, indicating that the developed model can accurately predict H₂S production. The H₂S production increase with soaking time, steam quality, and steam injection volume due to the strengthened aquathermolysis reaction. Based on the reservoir simulation, the H₂S production was predicted in the range of 228 m³ to 2895 m³ within the parameters of this study.

Supercritical multi-source, multi-component thermal fluid is a heavy oil thermal recovery method independently developed by China National Offshore Oil Co., Ltd (Beijing, China). It uses waste liquid at the production end of the production well as the water source, the injection medium temperature exceeds 374 °C, 22.1 MPa, and all the produced flue gas is re-injected. Compared with steam huff and puff technology, supercritical technology has the advantages of high enthalpy value, high heat utilization rate, good oil displacement effect, and being green and pollution-free. In addition, its oil–water treatment cost is low, it can realize the reuse of organic matter, it has a good cost advantage of water treatment under the background of low carbon, and it is a thermal recovery method with great application potential for offshore heavy oil. Therefore, it is necessary to carry out research on the adaptability and application potential of supercritical multi-source, multi-heat flow thermal recovery technology in the sea. Based on the laboratory one-dimensional displacement experiment, this paper reveals the mechanism of heavy oil supercritical multi-source multi-component thermal fluid displacement and the contribution of supercritical components to the displacement effect, and establishes the supercritical multi-source, multi-component thermal fluid numerical simulation characterization method. Combined with the characteristics of offshore heavy oil reserves, the main control factors affecting supercritical multi-source, multi-component thermal fluid development were established by numerical simulation and orthogonal test methods, and the adaptive screening method of offshore supercritical technology was established. The application potential of 670 million tons of offshore heavy oil reserves was evaluated and sorted, and KL 10-2 oilfield was selected as the pilot test oilfield. The results show that supercritical technology has great advantages in oil displacement and water treatment cost reduction, and the results play an important guiding significance for the development of offshore heavy oil technology system and the iteration of new technology.

Cyclic steam stimulation (CSS) is one of the main offshore heavy oil recovery methods used. Predicting the production of horizontal CSS wells is significant for developing offshore heavy oil reservoirs. Currently, the existing reservoir numerical simulation and analytical models are the two major methods to predict the production of horizontal CSS wells. The reservoir numerical simulation method is tedious and time-consuming, while the analytical models need many assumptions, decreasing models’ accuracy. Therefore, in this study, a novel methodology combining the particle swarm optimization algorithm (PA) and long short-term memory (LM) model was developed to predict the production of horizontal CSS wells. First, a simulation model was established to calculate the cumulative oil production (COP) of horizontal CSS wells under different well, geological, and operational parameters, and then the correlations between the calculated COP and parameters were analyzed by Pearson correlation coefficient to select the input variables and to generate the initial data set. Then, a PA-LM model for the COP of horizontal CSS wells was developed by utilizing the PA to determine the optimal hyperparameters of the LM model. Finally, the accuracy of the PA-LM model was validated by the initial data set and actual production data. The results showed that, compared with the LM model, the mean absolute percentage error (MAPE) of the testing set for the PA-LM model decreased by 4.27%, and the percentage of the paired points in zone A increased by 2.8% in the Clarke error grids. In addition, the MAPEs of the training set for the PA-LM and LM models stabilized at 267 and 304 epochs, respectively. Therefore, the proposed PA-LM model had a higher accuracy, a stronger generalization ability, and a faster convergence rate. The MAPEs of the actual and predicted COP of the wells B1H and B5H by the optimized PA-LM model were 8.66% and 5.93%, respectively, satisfying the requirements in field applications.

The discrete fracture model (DFM) and the embedded discrete fracture model (EDFM) are both the most widely used methods to simulate fractured wells’ production. In general, DFM represents fractures using an unstructured grid, and EDFM represents fractures using 2D virtual polygon grids embedded into 3D corner point grids. However, both EDFM and DFM suffer from non-orthogonal grid effects. Based on the non-linear flux approximation, this work presents a new flux scheme for DFM and EDFM that is capable of simulating fractured reservoir production in second-order accuracy. Two test cases are set to verify the accuracy of this method. Finally, this work is used to simulate the production of one fractured horizontal well belonging to the Y-oil field. Orthogonal design is used to optimize the parameters. The results show that this method is capable of improving accuracy while keeping the solution effective. This method exhibits good practicability and has practical significance for fractured reservoir development.