In the exploration and exploitation of oil and gas, artificial intelligence models are extensively employed in the prediction of formation pore pressure. Among them, single models tend to encounter problems such as overfitting or unstable prediction outcomes, leaving room for improvement in aspects like prediction accuracy and generalization ability. To enhance the prediction accuracy of formation pore pressure, a CNN-Attn neural network-based formation pore pressure prediction model was established by virtue of deep learning technology. In this research, five types of logging and while-drilling data were optimally selected, and the linear correlation between the data and formation pore pressure was verified using the Pearson correlation coefficient method. Through the optimization of the structure of the one-dimensional CNN, the model can effectively capture the local characteristics of the data and, when combined with the self-attention mechanism, strengthen the model’s ability to capture global dependencies, thereby elevating the model’s expressiveness and comprehension. To validate the prediction accuracy of this model, two wells in the Bayan block were subjected to prediction. The average absolute errors of the prediction results were both less than 1 MPa, the root mean square errors were both less than 1 MPa, the average relative errors were both less than 1.3%, and the determination coefficients were both greater than 0.9, with higher accuracy compared to the BP, CNN, and LSTM models. This model has improved the prediction accuracy of formation pore pressure and provided data support for drilling safety.

Fire flooding is a primary method used to enhance heavy oil recovery, often replacing steam stimulation. However, it encounters challenges such as low sweep efficiency and delayed effective times in heavy oil development. A new fire flooding PGI (pulse gas injection) technology was used to solve these problems. The combustion sweep effect of fire flooding can be improved by adjusting the working system of the gas injector. Based on the geological characteristics of the Hongqian 1 heavy oil reservoir in Xinjiang, the feasibility and mechanism of PGI were elucidated through numerical simulations. The influence of geological and engineering factors was studied on the development effect of PGI, and the consequences are applied to the fire flooding field. The combustion front can be controlled by the peak-valley value stage of PGI, accelerating the uniform movement of the combustion front and improving the oil displacement efficiency. Specifically, PGI can reduce the effective time by approximately 600 days and increase combustion sweep by more than 30%. This technique is particularly suitable for medium heterogeneous reservoirs with a vertical permeability contrast of less than 15 and crude oil viscosities ranging from 2 000 to 10 000 mPa·s. The optimal pulse amplitude ranges between 1.5 and 2, with a recommended step length of 30 days. When applied to the Hongqian 1 fire flooding industrial area in Xinjiang, daily oil production increased by 0.5 to 2.8 t for the well group, and the air-oil ratio decreased by 34%. PGI can achieve a better production increase effect for field production.

To investigate the spontaneous imbibition mechanism of shale reservoirs to fracturing fluid, the marine shale of Longmaxi Formation in Sichuan Basin and the continental shale of Yanchang Formation in Ordos Basin were taken as the research object, and the actual fracturing fluid was taken as the imbibition fluid. The spontaneous imbibition characteristics and control mechanism of marine and continental shales on fracturing fluid were studied by the spontaneous imbibition test system and the characterization methods of mineral composition and microscopic pore. The results show that the spontaneous imbibition curves of the two groups of shale increase rapidly and then decrease slowly with the increase of imbibition time, and the decrease of imbibition is mainly due to the decrease of shale weight caused by the dissolution of soluble minerals. TOC (total organic carbon) and quartz content of the two groups of shale exhibit a negative correlation with the maximum imbibition, while clay minerals, carbonate and pore volume exhibit a positive correlation with the maximum imbibition. Although the clay mineral content of Yanchang Formation shale is higher than that of Longmaxi Formation shale, its maximum imbibition is relatively low, which may be due to the more developed pore structure of Longmaxi Formation shale. The adsorption of hydrophilic minerals in shale and the capillary action of microscopic pores are the main driving forces for the spontaneous imbibition of fracturing fluid, and the influence of microscopic pores is stronger than that of hydrophilic minerals. For the flowback problem of fracturing fluid in shale gas wells, a better fracturing fluid ratio scheme can be proposed by adjusting the chemical composition of fracturing fluid and combining the spontaneous imbibition test results of fracturing fluid to meet the actual engineering needs. The research results can provide theoretical references for efficient development of shale gas.

Aiming at the thermoacoustic oscillation in the combustion process of a new generation of gas turbine, several time-domain analysis methods of high frequency data signals were retested and compared by constructing a complex network model. The results show that the two complex network models, node strength and network diameter, can give an earlier warning of thermoacoustic oscillations than the traditional time-domain analysis methods(root mean square and time kurtosis). Coupling prediction effect and data processing time, the node strength is preferred to construct the complex network model. Finally, the method was applied to the experimental data analysis of multi-nozzle micro-mixing burners, and it was found that the characteristic turning point of the method is about 2.3 s ahead of limit cycle analysis and the characteristic point of statistical analysis.

The voltage fluctuations in the high-voltage DC bus of the train traction converter have a significant impact on the output power quality of the traction system. Therefore, it is necessary to improve the response speed of the intermediate stage isolated DC/DC converter to reduce the power coupling between the high-voltage stage and the low-voltage stage. Taking the isolated DC/DC converter as the research object, an unbiased model predictive control and sampling noise suppression strategy was proposed to address its inherent problems of high sensitivity to circuit parameters and susceptibility to sampling noise. Firstly, the operation principle of the dual-bridge series resonant converter and the causes of the steady-state errors were analyzed, and a feedback correction method based on recursive least square algorithm was designed to eliminate the steady-state error. Then, the introduction of noise suppression coefficient reduces the sensitivity of the control variable to the control target through a simple and effective method. Furthermore, the virtual current was utilized in predictive model instead of the actual current sampling value, and it further reduces the system costs. Finally, an experimental platform was built to verify the improvement of the proposed strategy in both steady-state and dynamic performance.

The issue of distributed PV(photovoltaic) integration capacity allocation in distribution networks was addressed, which focuses on the photovoltaic integration capacity configuration based on load-storage coordination optimization. A coordinated regulation model for load-storage systems, incorporating energy storage, dispatchable load, and interruptible load, was first established. Based on this model, the constraints of load-storage regulation capabilities were considered. The optimization model for the configuration of distributed PV integration in the distribution network was developed with the objective of maximizing the capacity of distributed PV and the net investment and operational profit of the distribution network. Simulation results show that through the coordinated regulation of distributed PV with energy storage, dispatchable loads, and interruptible loads, significant improvements in the integration of distributed PV into the grid can be achieved.

To address the issues of slow convergence speed and susceptibility to local optima in ELM (electromagnetism-like mechanism) algorithm for fault location problems in distribution networks, BELM (binary electromagnetism-like mechanism) algorithm was proposed. First, the Sobol sequence was introduced to initialize the population to ensure the quality of the initial population. Second, based on the fitness value, the population was divided into a high-quality population and an ordinary population, and an optimal particle guidance strategy and a local search strategy based on the XOR operation were adopted for these two sub-populations, respectively, the former guides the high-quality particles to the potentially optimal location to accelerate the convergence speed, the latter performs global exploration and enriches the diversity of the population by exchanging information with the elite particles. Finally, the search efficiency of the algorithm was further improved by the improvement of the combined force calculation and particle movement rules. The simulation results show that compared with other algorithms, the proposed algorithm demonstrates superior accuracy and rapid convergence in locating faults within distribution networks.

To address the issue of fault signal transient characteristics being easily affected by noise, leading to misidentification of feeders in single-phase ground faults within resonant grounded systems, a line selection method was proposed that combines parameter-optimized VMD (variational mode decomposition) and improved D-S (dempster-shafer) evidence theory for fault feature fusion. First, to tackle the challenge of selecting the penalty factor Alpha and decomposition level K parameters in VMD, NRBO (Newton-Raphson-based optimizer) is introduced to adaptively determine these parameters under different noise environments. Next, three fault features—kurtosis, polarity, and transient energy—was fused, and the Jousselme distance was incorporated into D-S evidence theory to prevent conflicting results caused by noise interference on fault features. This approach provides the probability of fault occurrence on each feeder, allowing for accurate fault feeder identification. Finally, Simulink simulation results demonstrate that the method can accurately identify the fault feeder across various noise levels and fault scenarios. Compared to other parameter optimization algorithms, it achieves faster convergence, and the introduction of Jousselme distance further enhances the reliability of fault feeder identification.

In order to improve the prediction accuracy of lithology affected by imbalanced geological data, an ECA-MSCB ResNet model was proposed. The model integrates ECA (efficient channel attention) and MSCB (multi-scale convolutional block) into the traditional ResNet architecture to achieve efficient extraction and representation of lithological data features. For the issue of imbalanced lithology categories, prior probability-balanced logit bias was introduced during model training, and the focal loss function was modified to enhance the recognition of minority lithology classes. Experimental results show that the model based on ECA-MSCB ResNet performs well on the imbalanced geological lithology dataset, achieving an average prediction accuracy improvement of approximately 7.45% compared to the original ResNet model and 27.33% compared to the random forest method. Notably, the recognition of minority lithology classes improves by an average of 17.9%. Furthermore, the model demonstrates strong lithology classification ability on public datasets, achieving an F₁-score of 75.77%. In addition, the recognition accuracy of the proposed model outperformed both traditional and mainstream methods. The ECA-MSCB ResNet method holds significant application value in the field of imbalanced geological lithology recognition.

Dual-mode communication based on wireless and PLC (power line communication) can complement each other, and is widely used in smart metering and power Internet of Things. For the application of OFDM (orthogonal frequency division multiplexing) dual-mode communication system in new energy low-latency service access and other applications, an OFDM subcarrier diversity combination and power adaptive allocation algorithm considering the non-ideal channel estimation was proposed. Firstly, an adaptive optimal power allocation model was established with the constraints of service data volume and transmission power and the goal of minimizing the total delay. Then, on the basis of diversity grouping, whale optimization algorithm and threshold recovery were used to optimize power allocation, to achieve the compromise between algorithm complexity and performance. The simulation results show that the proposed algorithm can reduce the average transmission delay and stability while satisfying constraints such as rate, providing enhanced performance guarantees for real-time data acquisition in applications such as distribution equipment status monitoring.