Inter-well connectivity has become one of the important criteria for guiding the development of water drive reservoirs. Traditional methods for predicting inter-well connectivity, such as tracer analysis, well testing analysis, numerical simulation, etc, have problems such as computational difficulties, cumbersome processes, and high costs. However, deep learning based methods suffer from issues such as data sensitivity and poor adaptability. To address the aforementioned challenges and problems, this paper proposes a prediction method for predicting inter well connectivity by fusing numerical models with graph neural networks. On the one hand, this method fully considers the physical parameters related to the injection and production well network in the production process, which derives a numerical model of inter-well connectivity considering multiple factors, and solves the problem of single form in previous numerical models. On the other hand, the similarity between well network structure and graph structure was reasonably utilized, a graph neural network model based on long and short-term memory neural network was designed. Then, a fusion method with deep learning network model was proposed, which was fused with numerical model. This solves the problem of traditional artificial intelligence methods ignoring physical parameters, and introduces self attention mechanism to optimize the model under the long and short-term memory neural network framework. The mechanism model and actual reservoir production data were used to predict inter-well connectivity and fluid production on the established model, and new development plans were formulated based on the predicted results. The model’s performance was verified through several sets of experiments. The results show that when the model established in this paper is used to predict inter-well connectivity, and the liquid production is calculated based on the connectivity prediction results, the prediction accuracy is as high as 98%, indicating that the model’s prediction results are reliable. Using the fusion model to predict inter-well connectivity of different sub-layers in the reservoir, it is found that the model’s prediction accuracy reaches over 95% in well patterns of different scales, showing strong applicability. Finally, the production plan was adjusted based on the connectivity prediction results: liquid reduction was implemented for both wells with high connectivity and those with low connectivity. Comparison shows that the predicted recovery degree after 10 years of the adjusted development plan is 6.8% higher than that of the original plan. This method balances physical interpretability and computational efficiency, providing reliable technical support for judging the development effect of water-flooded reservoirs and designing secondary development plans.

Accurate production forecasting serves as a critical determinant for optimizing extraction strategies, guiding long-term field management in reservoir development. Both conventional methods and deep learning techniques face significant challenges in production forecasting due to the increasing complexities of reservoir extraction. Firstly, traditional production forecasting methods often fail to fully capture the complex reservoir behavior. Finally, these approaches demonstrate suboptimal performance in wells with limited data. These problems can lead to a decrease in prediction accuracy. To address these challenges, this paper introduces the Patching-iTransformer method and applies meta learning. The method improves prediction accuracy and overcomes the problem of few samples in production forecasting. Specifically, we implement a patching mechanism that segments the input time series, thereby converting the univariate time series into a two-dimensional representation. This architectural enhancement significantly strengthens the modelʼs capability to capture latent interdependencies among variables. Currently, we develop a PiAM meta-learning algorithm with domain-specific adaptation for oil field applications by quantitatively assessing individual well contributions to reservoir exploitation. We use time series data from real wells to evaluate the accuracy of multiple wells under the PiAM model. The experimental results demonstrate that Patching-iTransformer achieved better performance improvements than the iTransformer method. R² increased by 0.297, RMSE decreased by 11.64% and MAE decreased by 3.49%. PiAM meta-learning method demonstrated superior performance over the Patching-iTransformer model, showing a 0.535-point improvement in the R² coefficient along with a reduction of 27.54% in RMSE and a decrease of 28.22% in MAE.

Due to the wide application of closely spaced multi-well horizontal pads for developing unconventional gas reservoirs, interference between wells becomes a significant concern. Communication between wells mainly occurs through natural fractures. However, previous studies have found that interwell communication through natural fractures is varied, and non-communication also appears in the mid and late stages of production due to natural fracture closure. This study proposes a boundary element method for coupling multi-connected regions for the first time. Using this method, we coupled multiple flow fields to establish dual-well models with various connectivity conditions of the stimulated reservoir volume (SRV) region. These models also take into consideration of adsorption and desorption mechanism of natural gas as well as the impact of fracturing fluid retention.

The study found that when considering the non-communication of SRV regions between multi-well horizontal pads, the transient behavior of the targeted well exhibits a transitional flow stage occurring before the well interference flow stage. In addition, sensitivity analysis shows that the well spacing and production regime, as well as the connectivity conditions of the SRV region, affect the timing of interwell interference. Meanwhile, the productivity of the two wells, reservoir properties, and fracturing operations affect the intensity of interwell interference.

Eighteen models based on two equations of state (EoS), three viscosity models, and four mixing rules were constructed to predict the viscosities of natural gases at high temperature and high pressure (HTHP) conditions. For pure substances, the parameters of free volume (FV) and entropy scaling (ES) models were found to scale with molecular weight, which indicates that the ordered behavior of parameters of Peng-Robinson (PR) and Perturbed-Chain Statistical Associating Fluid Theory (PC-SAFT) propagates to the behavior of parameters of viscosity model. Predicting the viscosities of natural gases showed that the FV and ES models respectively combined with MIX4 and MIX2 mixing rules produced the best accuracy. Moreover, the FV models were more accurate for predicting the viscosities of natural gases than ES models at HTHP conditions, while the ES models were superior to PRFT models. The average absolute relative deviations of the best accurate three models, i.e., PC-SAFT-FV-MIX4, tPR-FV-MIX4, and PC-SAFT-ES-MIX2, were 5.66%, 6.27%, and 6.50%, respectively, which was available for industrial production. Compared with the existing industrial models (corresponding states theory and LBC), the proposed three models were more accurate for modeling the viscosity of natural gas, including gas condensate.