The rapid advancement of electric vehicles (EVs) has created a growing demand for high-performance battery technologies capable of delivering longer driving ranges and significantly reduced charging times. Conventional lithium-ion batteries, although widely used in modern EVs, face several limitations such as limited energy density, long charging durations, thermal instability, and gradual capacity degradation over time. To overcome these challenges, researchers and industries across the world are focusing on the development of next-generation EV batteries that can enhance vehicle efficiency, reliability, and sustainability. Next-generation battery technologies primarily aim to increase energy density, improve charging speed, enhance safety, and reduce manufacturing costs. Among the most promising innovations are solid-state batteries, lithium-sulphur batteries, silicon-anode batteries, and sodium-ion batteries. Solid-state batteries replace the liquid electrolyte with a solid electrolyte, thereby improving thermal stability, reducing the risk of fire, and enabling higher energy storage capacity. These batteries are expected to provide driving ranges exceeding 800 kilometres on a single charge while supporting ultra-fast charging within minutes. Similarly, lithium-sulphur batteries offer a much higher theoretical energy density compared to conventional lithium-ion cells, making them suitable for long-range EV applications. However, challenges related to cycle life and material degradation still require further research. Artificial intelligence and smart battery management systems are playing an important role in monitoring battery temperature, charge cycles, and energy consumption patterns. Such technologies help optimize performance, extend battery lifespan, and ensure operational safety. Furthermore, the integration of renewable energy sources with EV charging infrastructure supports sustainable transportation and reduces dependence on fossil fuels. Environmental concerns associated with battery production and disposal have also encouraged the development of recyclable and eco-friendly battery materials. Researchers are exploring sustainable alternatives such as sodium-ion and magnesium-ion batteries, which rely on more abundant and less expensive raw materials than lithium and cobalt. Recycling technologies are also advancing to recover valuable materials from used batteries, thereby minimizing environmental impact and supporting a circular economy. In conclusion, next-generation EV battery technologies have the potential to revolutionize the automotive industry by increasing driving range, reducing charging time, improving safety, and promoting environmental sustainability. As technological innovations continue to evolve, these advanced batteries will play a crucial role in the global transition toward cleaner, more efficient, and sustainable transportation systems.
Operational budgets remain a challenge for healthcare systems around the world, resource allocation continues to be ineffective, and the drive to enhance patient outcomes remains a challenge. Combining intelligent mathematical modelling with machine learning methods has proven to be an effective way of improving healthcare decision-making and operational efficiency. The study explores how machine learning techniques can be used to develop mathematical models for healthcare systems, with a focus on enhancing cost efficiency and optimizing patient outcomes. The research utilizes predictive analytics, statistical modeling, and intelligent optimization techniques to analyze healthcare operational data, patient treatment patterns, and resource utilization metrics. Healthcare datasets and predictive modeling techniques were used to apply a quantitative research methodology. Multiple machine learning algorithms were tested for their ability to predict healthcare costs, risks and treatment outcomes, such as Random Forest, Support Vector Machine (SVM), Linear Regression and Neural Networks. Mathematical optimization frameworks were integrated for the support of hospital resource allocation and planning for operations. The results suggest that intelligent predictive models can greatly enhance diagnostic accuracy, minimize unnecessary healthcare costs, and optimize patient management efficiency. These results also highlight how machine learning can help alleviate healthcare system readmission rates, staffing issues, and clinical decision-making time.The study concludes that Mathematical Modeling coupled with Machine Learning is an effective and scalable framework for modern Healthcare Management. The potential impact of the proposed approach for healthcare providers, policymakers, and hospital administrators is significant, as it would allow for data-driven decisions that would enhance the economic efficiency and quality of patient care. Another key finding from the research is that in the future, technical solutions based on artificial intelligence will have to be explainable and ethical when applied to healthcare settings.
This scientific investigation elaborates on a paradigm-shifting design and optimization methodology for a decentralized hybrid solar-wind microgrid inherently coupled with a Battery Energy Storage System (BESS). Contemporary power networks grapple with the dual predicaments of maintaining rigorous power quality and ensuring uninterrupted grid resilience amidst extreme load perturbations. This research resolves these bottlenecks through a structurally novel decentralized control architecture, fortified by a computationally efficient hybrid Particle Swarm Optimization (PSO) heuristic dispatch mechanism. Unlike orthodox centralized Energy Management Systems (EMS) that suffer from vulnerability to trans-nodal communication latency, or elementary decentralized droop strategies that fundamentally forfeit economic optimality, the proposed framework structurally integrates localized hardware-level transient suppression with globally aware cost-minimization. Within this context, the study mathematically formulates the nonlinear dynamic behaviors of a 500 kW photovoltaic matrix, a 350 kW wind turbine generation unit, and an 800 kWh capacity BESS. The multi-objective optimization algorithm strategically isolates and suppresses Total Harmonic Distortion (THD) below 2.8%, actively mitigates cyclic battery capacity degradation, and ensures levelized economic dispatch. The analytical models were exhaustively validated using MATLAB/Simulink Simscape libraries subjected to extreme transient load profiling. Results demonstrate an extraordinary voltage settling sequence within 0.15 seconds trailing severe load disturbances—vastly superior to the 0.40 seconds exhibited by standard PI-centric decentralization. Concurrently, the algorithmic dispatch reduced cumulative operational expenditures by 14.6% against benchmark rule-based predictive models, confirming the exceptional efficacy of hybridizing local physical-layer governance with heuristic global optimization in renewable-dense topologies.
This paper aims to enhance students' autonomous learning effectiveness by addressing the actual learning needs of higher vocational students. Focusing on higher vocational English as the research subject and emphasizing the importance of self-learning ability, it first analyzes the value of AI in supporting the development of autonomous learning skills in this context. The study then explores and summarizes the pathways through which AI can facilitate autonomous learning from five key dimensions: listening, speaking, reading, writing, and translation. The objective is to contribute to the effective reconstruction of the higher vocational English education ecosystem and to promote a paradigm shift toward human-computer collaboration in learning.
Addressing false positives and false negatives caused by biased AI training data requires a comprehensive solution. At the data level, broadening data sources ensures diversity; weighted sampling, oversampling, and synthetic data are used to enhance balanced distribution; biased samples are eliminated, and labeling standards are standardized. During training, a fairness loss function and adversarial training are introduced. Using a fixed random seed ensures stability, and domain knowledge graphs are used to supplement context. In the iteration phase, a false positive case knowledge base is built, and samples are continuously collected through log analysis and user feedback. Root cause analysis is used to guide model fine-tuning. Simultaneously, a dynamic evaluation and closed-loop iteration mechanism is established to monitor fairness metrics and ensure the model’s generalization ability, comprehensively reducing the impact of bias from source to application.
Aiming at the deficiencies of computational domain modelling in the current numerical simulation of hydraulic performance and anti-clogging performance of labyrinth-type irrigators, this study takes the actual trapezoidal tooth two-return irrigator as the object, establishes an overall computational physical model including the outlet chamber, and simulates and analyses the velocity and pressure distributions in the flow channel under seven different flow conditions based on the FLUENT software and carries out the structural optimization of the tooth-tip region of the flow channel by replacing the tooth-tip part with a rounded structure instead of a chamfered structure. The tooth tip part is changed to rounded structure instead of chamfered structure. The results show that when the inlet flow rate exceeds 0.4 L/h, the water flow in the flow channel can be regarded as turbulent; there are obvious velocity and pressure high value zones at the top of the teeth in the region of flow on the turn; The flow behavior of the emitter is described by a power-law relationship, exhibiting a flow exponent of about 0.5; This study proposes an innovative flow channel tip structure optimization scheme. By adopting an original micro-rounded design strategy—replacing the traditional chamfered structure with a precision round corner of 0.100 mm—the anti-clogging performance is significantly enhanced, reducing the maximum particle residence time to 23.5 s. This structural optimization provides a clear original contribution to improving flow uniformity and reducing particle retention, offering a new technical pathway for handling fluids with high solid content.
Carbon auditing is an essential approach to monitoring progress towards achieving “dual carbon” strategic objectives. In keeping with China’s “dual carbon” strategy, the incorporation of machine learning methods into the development of carbon audit evaluation models has represented a significant innovation that has improved audit efficiency. This study constructed a carbon audit evaluation model by employing machine learning techniques to evaluate and analyze its effects on three provinces and one municipality in southwestern China from two dimensions: economic benefits and low-carbon performance. The results indicated divergent aspects relating to economic growth and low-carbon benefits in the studied regions. The proposed model not only assessed the current economic and low-carbon performance of those provinces, but also predicted future trends, that could facilitate the identification of audit priorities, and, thus, provide valuable references for subsequent carbon auditing research.
This paper proposes an artificial neural network (ANN) framework utilizing a multilayer feed-forward structure to evaluate the static security of a representative 380 kV Saudi transmission grid. The proposed model estimates a composite Static Security Index (SSI) that accounts for both line loading violations and bus voltage deviations. The ANN, designed with one hidden layer of ten neurons, receives normalized active and reactive power demands from load buses as inputs, while producing the SSI corresponding to each contingency as the output. Training is conducted using a back-propagation learning algorithm, where datasets are derived from Newton-Raphson load-flow simulations at various loading levels. Comparative analysis confirms that the proposed ANN approach matches the accuracy of the conventional NRLF-based evaluation while achieving substantially faster computation. The obtained performance suggests that the model can serve as an effective real-time decision-support tool for online contingency ranking and system security assessment in control centers.
With the rapid development of generative artificial intelligence (AIGC), it is quickly reshaping knowledge production and dissemination methods, effectively addressing the pain points and challenges in vocational education. Therefore, studying the application logic and implementation paths of AIGC in higher vocational accounting programs is not only a response to technological changes but also a critical breakthrough for promoting high-quality development in vocational education. This paper explores the application scenarios of AIGC in higher vocational accounting programs from three aspects: teaching content generation and resource construction, practical skills cultivation, and personalized learning and career development. It analyzes the advantages and challenges of its application, proposing optimization paths such as building a "human-machine collaborative" teaching model, balancing efficiency with educational essence, strengthening technical support and standardized management, constructing a secure and trustworthy AI education ecosystem, improving the faculty training system, and perfecting policy and standard construction. The study points out that the development of AIGC has injected new momentum into higher vocational accounting education, enhancing teaching effectiveness and efficiency. It systematically reconstructs the closed-loop path of "theoretical teaching—practical training—professional skills," achieving a reshaping of the vocational education system empowered by AI. However, it emphasizes that the essence of education should be the premise, preventing skill alienation and ethical risks. It stresses that higher vocational accounting education needs to transform at three levels: institutional construction, resource investment and utilization, and faculty. Through the organic integration of technological innovation and educational principles, it aims to achieve sustainable empowerment of AIGC in higher vocational accounting education, providing a systematic solution for the high-quality development of vocational education.
Aiming at the problems of low precision of learning portraits, inadequate push and early warning intervention in vocational education, this study integrated multimodal data and AI technologies to construct a "feature-level + decision-level" two-layer fusion model and a four-dimensional learning portrait model of "knowledge-skill-behavior-style". It also designed a DQN adaptive push algorithm, an XGBoost-LSTM risk early warning model and a human-intelligence collaborative hierarchical intervention closed-loop mechanism. A 16-week empirical study on intelligent manufacturing majors showed that the proposed models and system significantly improved students' learning effect, efficiency and initiative, with core model performance indicators all exceeding 0.85 and good intervention effects. This research provides technical support for the digitalization and personalized education of vocational education, and also points out the limitations in samples and data as well as future research directions.
The relentless push to integrate inverter-based Renewable Energy Sources (RES) into weak transmission grids creates a difficult balancing act: how to minimize costs without risking the grid's dynamic stability. This paper presents a new solution for the Western Kenya 132/220kV grid: a Tri-Hybrid Artificial Intelligence Framework. This system is designed to bridge the gap between accurate forecasting and secure power dispatch. It combines a Long Short-Term Memory (LSTM) network for precise weather prediction, physics-based models for realistic solar and wind simulation, and a Hybrid Particle Swarm Optimization - Grey Wolf Optimizer (PSO-GWO) to make dispatch decisions. Unlike standard approaches that treat stability as an afterthought, the proposed framework embeds non-linear checks—specifically for Frequency Nadir (>49.5Hz), Voltage Recovery, and Critical Clearing Time—directly into the decision-making loop. When tested on the Western Kenya network, this approach showed clear advantages over standard Genetic Algorithms (GA) and PSO. It cut operational costs by 7.0% ($5,810/hr versus $6,250/hr), reduced carbon emissions by 17.7%, and converged to a solution 60% faster. More importantly, the sensitivity analysis pinpointed a hard limit for Solar Hosting Capacity at 65 MW; going beyond this point compromises system inertia. N-1 contingencies were effectively handled, keeping the frequency nadir safely at 49.65 Hz and fixing long-standing voltage issues at Kisumu. These results prove that it is possible to maintain a resilient grid even with high renewable penetration, offering a practical roadmap for other emerging economies facing similar grid weaknesses.
In order to explore new ideas of safety management of nitrification process device, a safety toughness evaluation method is proposed. Firstly, the safety toughness evaluation index system of nitrification process device is constructed from three aspects of pressure, state and response. Secondly, based on the entropy weight method and DEMATEL, objective weight and subjective weight are calculated respectively, and the combination weight is derived by using the combination assignment method, and each safety toughness characteristic parameter is obtained through the cloud model theory. Finally, this method is applied to the safety evaluation of nitrification process device in a certain place, and the overall safety toughness grade is “Good (Grade Ⅳ)”, the result is in line with the actual situation, and the application of this method for safety evaluation has certain scientific and reference significance for improving the safety status of nitrification process device.
The traditional control method of the bored pile drilling machine has problems such as high labor intensity and high risk, which have long restricted the construction efficiency and safety guarantee. This paper designs a remote electro-hydraulic control system based on Siemens S7-1200 PLC and Juguang 550 IoT module, combined with the MODBUS RTU communication protocol, which enables construction personnel to wirelessly remotely control the bored pile drilling machine robot, monitor its real-time status, collect key parameters, and manage them in the cloud. To verify the system performance, a dedicated experimental platform and actual engineering construction verification were set up. After strict testing, it was proved that this control system can maintain excellent stability and practicability in complex environments with long-distance communication and multi-sensor collaborative collection. In the actual pilot construction of a project in a mountainous area in Southwest China, this drilling machine can effectively increase construction efficiency, save 200,000 yuan in project costs, greatly reduce on-site labor intensity, significantly improve on-site safety, and optimize the construction schedule. In conclusion, this remote electro-hydraulic control system provides effective technical support and theoretical evidence for the intelligence and informatization of the bored pile drilling machine, and has good promotion and application value.
This paper focuses on the research of collaborative defense mechanisms for campus network security, addressing the challenges of virus invasion and data leakage faced by campus networks, and proposes the construction of a dynamic joint defense system that spans across systems and platforms. Although universities have made progress in the field of collaborative defense, they still face problems such as insufficient management collaboration, complex attack methods, and prominent data security risks, which require optimization from aspects of dynamic defense and standardized construction. The collaborative defense architecture designed in this study integrates core components such as dynamic network camouflage, multi-level intrusion detection, and distributed intelligent firewalls. The dynamic camouflage system confuses attackers through IP hopping technology; the intrusion detection module combines traffic analysis and behavior modeling to achieve precise threat recognition; the distributed firewall implements dynamic policy control based on SDN architecture. Each component operates independently and is closely linked, forming a "monitor-respond-recover" closed loop. The system construction needs to strengthen security awareness, innovate management mechanisms, integrate cutting-edge technologies, build a collaborative ecosystem, and establish a verification mechanism that includes technology optimization, strategy iteration, and process improvement. Practical cases show that this system has significantly improved the network attack blocking rate of a certain university and significantly shortened the residence time of APT attacks. Future research will focus on optimizing the dynamic collaboration algorithm of subsystems, conducting empirical evaluations in real network environments, continuously verifying and improving defense mechanisms through ongoing experiments, promoting the evolution of campus network security towards intelligence and adaptability, and providing reliable support for educational informatization.
In the era of ubiquitous connectivity, cybersecurity confronts an asymmetrically escalating threat landscape. Malware, as a primary attack vector in cyberspace, has evolved into a highly specialized black-market ecosystem. According to Kaspersky's 2023 report, 487 new polymorphic variants emerge every second globally, with 73% employing obfuscation techniques to evade conventional detection. This study focuses on constructing an AI-powered malware identification framework, proposing a Multimodal Feature Fusion-based Collaborative Detection framework (MMF-MalDet). The framework innovatively integrates static syntax tree analysis with dynamic memory behavior tracking, overcoming the limitations of single-dimensional detection approaches. By introducing a federated transfer learning mechanism, cross-institutional threat intelligence sharing is achieved while ensuring data privacy. Experimental results demonstrate a 98.6% detection accuracy for novel Emotet variants, representing a 41.5% improvement over traditional methods. The research concurrently reveals the dynamic equilibrium between adversarial robustness and model interpretability, providing methodological foundations for establishing proactive immune cybersecurity ecosystems.
Generative AI technology, as a product of the further development of information technology, is undoubtedly a driving force for educational transformation. Based on an in-depth exploration of the technical characteristics of this technology and its impact on education, and by examining the current status of vocational education teachers' professional skills and teaching methods, identifying the problems and needs they face, this study explores strategies for AI technology to enhance teachers' personal abilities. Building on the research of the current status of vocational education teachers, the paper proposes a research approach for teacher professional development strategies integrated with generative AI technology. Using case studies, it points out the implementation effects of this technology. The study believes that integrating AI technology into teacher professional development strategies can promote the professionalization of teachers and has reference value for improving the quality of education and teaching as well as future development trends.
The Nuan Gong Qi Wei Pill (NGQW) is a classic prescription that holds a significant position and is widely used by several generations of Mongolian physicians, especially in the traditional treatment of the "He Yi" disease in gynecologic conditions. Varying perspectives exist among Mongolian medical practitioners regarding its origin, making it crucial to explore its historical evolution, identify the formula's source, and analyze the inheritance and development of this classic prescription and ancient literature. A literature review was conducted to trace the origin and clinical application evolution of NGQW, examining its dosages, formulations, and administration methods. Based on the literature reviewed, the NGQW was initially documented in the "Yan Fang Bai Pian" (ཉམས་ཡིག་བརྒྱ་རྩ།, nyam yig brgya rtsa) authored by Gongman Gongjuepengda, compiled during 1550-1577. It was written at least 120 years earlier than the present account. It consists of seven ingredients—cardamom (Amomum kravanh pierre ex Cagnep, སུག་སྨེལ།, sug smel), clove (Eugenia caryophyllata Thunb, ལི་ཤི, li shi), agilawood (Aquiria agallocha Roxb, ཨ་ཀ་རུ།, a ka ru), Myristica fragrans Houtt (ཛྰ་ཏི།, dzva ti), Solomon's seal (Polygonatum sibricum delar ex Redoute, ར་མཉེ།, ra mnye), asparagus cochinchinensis (AsparaguscochinchinensisLour. Merr, ཉེ་ཤིང།, nye shing), and conic gymnadenia tuber (GymnadeniaconopseaR.Br, དབང་ལག, dbang lag)—it primarily addresses “He Yi” disease. Commonly treating gynecological blood stasis. While the specific dosage was unrecorded, the prescription's composition and therapeutic functions have endured through various Mongolian medical practitioners' clinical applications. The formulation and dosages are specific and distinct, Specifically, 300 g of cardamom, 50 g each of clove, agilawood, Myristica fragrans Houtt, Solomon's seal, asparagus cochinchinensis, and conic gymnadenia tuber. Among the ingredients in the NGQW formulation, Solomon's seal, asparagus cochinchinensis, and conic gymnadenia tuber require pre-processing before use. Nowadays, leading to two developed forms: pill and powder, with the pill being the more widely used form. Dosage range among 2-3 g (10 pills equals to 2 grams), 1-2 times a day, to be taken with warm water. After more than 400 years, the NGQW maintains widespread utilization in treating heart and kidney "He Yi" disease, commonly treating gynecological blood stasis, just like waist pain, cold abdominal pain, irregular menstruation, excessive leukorrhea, dysmenorrhea, female infertility ailments within Mongolian medicine. Its definitive therapeutic efficacy and distinctive ethnic characteristics render it a valuable asset for the development of Mongolian traditional medicine culture.
Against the backdrop of the era of big data and globalization, cross-border e-commerce platforms have gradually gained attention, and efficient recommendation is the key to the development of cross-border e-commerce platforms. This article focuses on the application of AIGC in the recommendation system of cross-border e-commerce, through crawling and processing user data, historical transaction data, and product information, using NLP and collaborative filtering algorithms to describe the recommended products, aiming to propose a personalized recommendation system that can collect user preference information, generate diverse recommended content and shopping suggestions. On the premise of ensuring the source of data, after verification, the recommendation system has achieved good results in terms of recommendation rate, recall rate, and user satisfaction level, meeting the diverse and personalized needs, enhancing user stickiness and platform retention rate, increasing product monetization, proposing new ideas for personalized recommendation marketing using AIGC technology in e-commerce platforms, and providing new solutions for optimizing recommendation systems of similar platforms.
Financial statement fraud detection is a challenging computational problem because confirmed fraud cases account for well under one percent of firm-year observations and the discriminative signal resides in nonlinear interactions among financial indicators rather than in isolated ratios. This study proposes CSFL-Stack, a cost-sensitive focal-loss, imbalance-aware stacking ensemble for financial fraud detection. The framework combines in-fold SMOTE oversampling, a self-implemented focal-loss neural network, Random Forest, Gradient Boosting, and a logistic-regression meta-learner, augmented by Bayesian cost-sensitive threshold optimization and SHAP-based interpretability. Experiments are conducted on a large U.S. firm-year panel (146,045 observations, 1990–2014, 964 confirmed fraud cases) under a strict time-aware protocol: training on 1990–2009, validation on 2010–2011, and out-of-time testing on 2012–2014. Among the six models compared, Logistic Regression achieved the highest test AUC of 0.7495, while the proposed stacking framework attained 0.7047. Results show that under severe temporal distribution shift, architectural complexity alone does not guarantee superior generalization; probability calibration, cost-sensitive threshold design, and explanation quality are equally critical. SHAP analysis highlights equity issuance, soft assets, long-term debt issuance, and tax-related variables as the most influential fraud predictors, consistent with interaction-driven fraud signals. The study contributes a reproducible, leakage-controlled, and explainable computational pipeline that advances fraud-detection methodology beyond single-classifier benchmarks.
Early detection of diseases plays a crucial role in reducing mortality rates, improving patient outcomes, and lowering healthcare costs. With the rapid growth of artificial intelligence (AI) and machine learning (ML) techniques, data-driven diagnostic systems have emerged as powerful tools for supporting clinical decision-making. This research paper presents a quantitative AI-based framework for early disease detection using basic medical data such as age, gender, body mass index (BMI), blood pressure, blood glucose level, cholesterol, and heart rate. The proposed system utilizes supervised machine learning algorithms, including Logistic Regression, Support Vector Machine (SVM), Random Forest, and Artificial Neural Networks (ANN), to classify individuals into healthy or disease-prone categories. A structured dataset consisting of 5,000 patient records was used for training and testing, with data preprocessing steps such as normalization, missing value handling, and feature selection applied to enhance model performance. Quantitative evaluation was conducted using accuracy, precision, recall, F1-score, and receiver operating characteristic (ROC) curve analysis. Experimental results demonstrate that the Random Forest model achieved the highest accuracy of 92.6%, followed by ANN with 91.3%. The findings confirm that AI-based analysis of basic medical data can provide reliable early disease prediction, enabling proactive healthcare interventions and improved population health management.
