Abstract: Under the smart grid paradigm, the surging penetration of distributed energy resources and deep integration of demand response mechanisms have led to complex characteristics in power load sequences, including strong nonlinearity, multi-scale coupling, and stochastic abrupt changes, posing significant challenges to traditional forecasting methods. This paper proposes an enhanced Temporal Convolutional Network (TCN) framework featuring an innovative multi-channel spatiotemporal feature decoupling architecture. Through synergistic optimization of dynamic feature weighting modules and temporal attention mechanisms, effective fusion of multi-source heterogeneous features is achieved. The method employs a hierarchical dilated causal convolution structure to preserve temporal causality while enhancing long-range dependency capture capabilities. A gated mechanism-based dynamic allocation strategy for feature contribution levels is designed to precisely quantify the spatiotemporal coupling relationships among meteorological factors, date types, and historical load data. Empirical studies demonstrate that compared with baseline models like LSTM and GRU, the proposed model achieves 23.7% and 18.4% improvements in MAE (15.8MW) and RMSE (21.3MW) metrics respectively, with 67.5% higher training efficiency and 32.1% reduction in prediction error for load mutations. This research provides a reliable solution for high-precision load forecasting in smart grid environments. Read More

Abstract: Lithium-ion batteries have been widely used in the field of energy storage such as electric vehicles by virtue of their high energy density, long cycle life and environmental protection characteristics. In order to enhance the precision of battery State of Charge (SOC) estimation, this paper proposes a Thevenin model as the equivalent circuit model. Utilizing a Genetic Algorithm (GA), the model parameters are optimized to ensure the accuracy of the model. The identification results are effectively verified. On this basis, three SOC estimation algorithms, GA-EKF, GA-AEKF and GA-AUKF, were designed in this paper, and simulations and error analyses were carried out based on the UDDS operating conditions data. The results indicate that the GA-AUKF algorithm demonstrates superior accuracy and stability in terms of SOC estimation accuracy, exhibiting a significantly higher level of precision than the GA-EKF and GA-AEKF algorithms. Read More

Abstract: The performance and lifetime of proton exchange membrane fuel cells (PEMFC) are highly dependent on the wetting state of the membrane, and the phenomenon of "membrane drying" can lead to serious problems such as decreased proton conductivity, increased heat production and even membrane tearing. In order to reduce the dependence on external humidification system, self-humidification technology at the battery level has become a research hotspot in recent years. This paper systematically reviews the methods of self-humidification by adjusting the structure of internal components (flow field, membrane, catalytic layer, gas diffusion layer) and material modification, including the key progress in the past five years. The results show that the water distribution uniformity can be significantly improved by optimizing the flow field design (such as bionic flow field and porous metal foam flow field). The introduction of inorganic/organic additives (such as CeO₂, MOF) or the use of bipolar membrane design can effectively improve water retention and proton conductivity; The functional modification of the catalytic layer and gas diffusion layer (such as oxide load and gradient pore structure) further enhances the operating stability under low humidity conditions. Although self-humidifying technology has advantages in reducing system volume and cost, it still faces challenges such as material compatibility, long-term durability, and inadequate water management for high current density. Future research should focus on multi-scale collaborative optimization, the development of new materials and the performance verification under actual working conditions to promote the wide application of PEMFC in portable devices and new energy vehicles. Read More

Abstract: By predicting and classifying relative poverty, we can spot and tell the difference between potentially impoverished groups early on. This allows for early intervention and efficient resource allocation, aiding long - term poverty governance. Given the lack of algorithmic research in relative poverty identification using multi - year data, this paper proposes the RP - DCSA model. It blends deep learning (DNN) with the interpretable SHapley Additive exPlanation (SHAP) model. The 2020 China Family Panel Studies (CFPS) survey data form the research base. Spearman correlation coefficients are applied for feature selection to eliminate redundant ones. Next, the DNN - based RP - DCSA model is built and compared experimentally with LR, RF, etc. Finally, SHAP is used for interpretable analysis to identify key features affecting relative poverty classification and assess their impact on results. The RP-DCSA model achieves an 89.55% classification accuracy on the CFPS2020 dataset, outperforming other algorithms in various indicators. Read More

Abstract: Developmental dyslexia is a common neurodevelopmental disorder that significantly affects children' normal learning and life. Early identification and intervention are crucial for patients. However, current models for classifying dyslexia fail to automatically extract features based on patient data and overlook the interrelationships between brain nodes in patients. Therefore, this paper proposes a graph convolutional neural network model that constructs a brain network from patients' fMRI data as an adjacency matrix, calculates node feature matrices, trains GCN models for classification, and achieves the diagnosis of dyslexia patients. Experimental results show that the model has an identification accuracy of 94%, precision of 95%, recall of 94%, and F1 score of 94%. This study provides a new approach for identifying and diagnosing dyslexia, which is beneficial for early intervention in dyslexia patients. Read More

Abstract: This article focuses on the MHPA-PV/T heat pump system and studies the energy matching relationship between the power generation and heat generation of the combined heat and power system and the user's electrical and thermal load requirements. Reasonable operation control strategies are proposed to meet the user's demand for electricity and suitable temperature water. Selecting the energy consumption period from 20:00 to 21:00 for testing, the feasibility of the selected control strategy was verified, while maximizing the energy utilization efficiency of the system. Read More