In complex urban environments, the design of urban gardens increasingly faces the challenge of balancing spatial efficiency, ecological function, and diverse human behavioural patterns. To address multi-objective optimisation scenarios, this study uses publicly available datasets: OpenStreetMap and GeoLife trajectory data. These datasets help extract urban garden structures and real-world human activity trajectories. Together, they form the basis for constructing spatial scenes and modelling behavioural constraints. Building on the classical Ant Colony Optimisation (ACO) framework, a multi-objective Pareto-front guidance mechanism is introduced to adaptively steer the search process. From a methodological perspective, this study proposes a dynamic heuristic function to enhance the algorithm's ability to address complex layout goals, including path connectivity, functional zoning diversity, and visual aesthetic harmony. This approach enables the algorithm to respond simultaneously to static spatial structures and dynamic behavioural patterns, achieving a synergistic optimisation between spatial layout and user needs.
Aggregate full-cycle water conservation data from 47 500-kV transmission and sub-station projects in the Jiangsu Plain Water Network Area from 2015 to 2024. Project characteristics, including project scale and disturbed soil-and-rock area, were extracted to construct a database for statistical analysis. The results show that drilled piles accounted for 87.75% of foundations, while mud generated during single-foundation excavation accounted for 53.96% of the total excavation material. Disposal was predominantly conducted through in-situ landfill (98.93%). Based on these findings, it is recommended that landfill disposal be restricted; measures for dewatering, volume reduction, and harmless treatment be implemented; resource utilisation be promoted; and whole-process supervision with clear responsibility mechanisms be established.
This study aims to address the limitation of traditional soil and water conservation measures that rely on empirical judgement. By integrating Geographic Information System (GIS) technology with deep learning, the study constructs an intelligent decision support system for watershed soil and water conservation and ecological restoration. This study first develops a GIS-based automatic water environment monitoring system. It integrates multi-source geographic data to realise real-time monitoring and early warning of water quality. Further, it constructs a multi-scale soil and water environmental quality evaluation model. By using principal component analysis (PCA), one-way analysis of variance (ANOVA) and pollution index evaluation method, the study comprehensively evaluates the soil and water environmental quality and reveals the key driving factors at different spatial scales. In the field of deep learning, the study optimises the backbone model of Convolutional Neural Network (CNN). This study innovatively integrates GIS and deep learning technologies.
In order to promote the consumption and utilisation of clean energy and reduce the tax burden of benchmark electricity pricing policy, China has implemented a nationwide trial of green power certificate issuance and subscription systems. To further advance the adoption of clean energy and mitigate the fiscal impact of the benchmark tariff policy, this paper constructs a multi-policy coupling model with interaction terms to assess the fiscal relief effect of green certificate trading under the coexistence of benchmark electricity pricing policies and the green certificate system in China. Additionally, the model explores the effects of different trading mechanisms and quota schemes (free market for green certificates vs. mandatory market for electricity quotas) on the pricing strategies of heterogeneous sellers and compliance decisions of heterogeneous quota holders, quantifying synergistic or offsetting effects among different policy instruments in the Chinese context.
Aiming at the real-time optimisation problem of AC/DC hybrid distribution network with high proportion of new energy access, a 'wind-solar-load-storage' collaborative scheduling framework based on multi-agent reinforcement learning (MARL) is proposed. Firstly, the Markov game model is constructed, and wind power, photovoltaic (PV), energy storage and flexible load are modelled as heterogeneous agents, and a mixed action space integrating DQN (Deep Q-Network) and Actor-Critic is designed, and the federated-edge collaborative mechanism is introduced to realise the privacy protection training of 'data-fixed model moving'. The single step decision-making time is less than 70 ms, and the voltage fluctuation is strictly controlled within +/- 5%. It achieves the coordinated optimisation of economy, safety, and privacy, providing a new paradigm for real-time scheduling of high proportion new energy distribution networks.
The high proportion of renewable energy integration endows the power grid with prominent heterogeneous features, including expanded node capacity differences, discrete line impedance, and dynamic power flow variations. Traditional models based on homogeneity assumptions are incompetent to accurately characterise the cascading failure mechanisms of such grids. To address this gap, this paper proposes a cascading failure probability model that integrates heterogeneous topology, dynamic power flow, and cyber-physical coupling. Firstly, a multi-dimensional heterogeneity indicator system, encompassing node capacity heterogeneity, electrical betweenness Gini coefficient, and bidirectional power flow index, is established to quantify the structural, operational, and coupling heterogeneity of green power grids. The results provide theoretical basis and quantitative tools for planning and real-time prevention and control of highly elastic green power grid.
To address the challenges of energy inefficiency and delayed response in ventilation management of energy-intensive underground infrastructures, this study proposes a novel digital technology named Spatial Temporal Attention-based Back Propagation (STABP), which integrates coupled simulation and BP algorithm. It aims to achieve accurate prediction and dynamic regulation of ventilation demand in underground spaces, thereby optimising energy use. This study constructs a cross-disciplinary technological framework that integrates physical process simulation with data-driven algorithms, representing a novel approach to managing technological complexity in building energy systems. Firstly, based on Kriging interpolation method, spatial reconstruction is performed on limited sensor data to generate high-resolution gridded pollutant concentration fields, which serve as input boundary conditions for coupled simulation. Then, the BP algorithm is used for rapid dimensionality reduction and error compensation of high-dimensional spatiotemporal fields. Time series data is processed using Long Short-Term Memory (LSTM).
With the rapid development of science, technology, and the economy, the Earth's ecosystem has suffered severe damage. The Guangxi Beibu Gulf Marine Region (GBGMR) is China's extremely important ecological barrier. This work aims to help formulate scientific and effective governance of the GBGMR and achieve Common Prosperity of the GBGMR to enable the high-quality and sustainable development of the GBGMR. This is achieved by constructing an integrated Ecological Carrying Capacity (ECC) model and introducing footprint breadth and depth analysis. The results show that the carbon and water ecological environments in the GBGMR show significant temporal and spatial differences. The ecological sustainability of Ningxia province along the GBGMR is the worst, while that of Henan is the strongest.
Combined with the bidirectional long short-term memory network, a temporal prediction model is constructed to characterise the dynamic evolution characteristics of carbon emissions and environmental comfort. On this basis, a multi-objective optimisation framework is established. The non-dominated sorting genetic algorithm II is adopted to solve the optimal Pareto frontier, thus realising the coordinated trade-off and dynamic regulation of energy consumption and comfort. On the premise of maintaining the indoor thermal-humidity environment within the optimal comfort range, the energy consumption of lighting and Heating, Ventilation, and Air Conditioning (HVAC) systems is successfully reduced by 21.4%. The optimisation of environmental quality significantly improves the cognitive status of researchers, with an estimated 11.5% increase in innovative work efficiency. The research findings confirm that reducing the carbon footprint of campuses can effectively empower scientific research and innovative productivity, providing a scientific paradigm for the refined management of green and smart parks.
At present, energy network security threat identification still faces the problem that temporal and network relationships are difficult to fuse. To address this issue, this study proposes a fusion model using Graph Neural Network (GNN) and Transformer model. This model mainly includes the following parts: using Graph Attention Network (GAN) to mine the spatial relationships between energy nodes and control entities; and using Multi-Head Self-Attention (MHSA) to extract long-range time series of energy regulation data. By combining the above two methods, the model well completes end-to-end threat detection for energy communication networks. The above research results verify that the method of joint modelling of spatial and temporal information has certain effectiveness in the field of energy network security, which provides a new idea for constructing adaptive threat identification methods in localised energy regulation networks.
This study focuses on the application research of the intelligent Backpropagation (BP) algorithm in promoting regional new energy dissemination within international new energy teaching, exploring the practical value and mechanism of the algorithm from multiple dimensions. Based on the dataset of the Chinese Bridge Chinese Proficiency Competition for Foreign College Students and the learning data from the Chinese International Education Online platform, the study selects ten core features as input variables. They include learners' regional new energy cognitive basis, learning behaviour characteristics, and regional energy demand matching degree, while taking regional new energy dissemination effectiveness (covering knowledge mastery, dissemination willingness, and cooperative attitude) as the output variable to construct a BP neural network model. The research results enrich the theoretical system of international new energy education, and offer empirical support and practical guidance for designing regionally adaptive teaching programs and promoting the collaborative development of cross-border new energy technologies.
This paper used the power Internet of Things (IoT) technology to monitor the status of the intelligent DR and give early warning of faults. First, the existing problems of the current DR were introduced. This paper then analysed the status monitoring requirements and system configuration of the DR and then discussed the monitoring technology of the intelligent DR to deal with the corresponding data faults. At the end of this paper, the effect of condition monitoring and fault early warning of intelligent DR was analysed, and finally, the conclusion was drawn. After adopting the IoT, the monitoring accuracy of each intelligent DR has improved compared with the previous one. The timeliness of fault early warning in DR has been greatly improved after the adoption of power distribution network technology, and the timeliness of fault early warning in DR 5 can reach 92%.
This paper introduced an environmental module for measuring carbon emissions and carbon tax strategies and a dynamic model to the Computable General Equilibrium (CGE) model to transform it into a dynamic model, which is capable of simulating and analysing various low-carbon emission reduction strategies. A case study was then conducted. A social accounting matrix was constructed. Subsequently, the carbon tax and subsidy strategies were dynamically analysed. The scenarios of increasing the carbon tax rate and subsidy were compared with the baseline scenario. The results indicated that a low-carbon emission reduction strategy that increases the carbon tax could effectively reduce oil and coal consumption and carbon emissions. As the carbon tax rate increased, both oil and coal consumption and carbon emissions decreased. Moreover, implementing a carbon tax subsidy could mitigate the negative impacts on economic development caused by the carbon tax.
With the continuous exploitation of fossil fuels, the reserves of this non-renewable energy source are increasingly being consumed. To alleviate the energy crisis and the environmental problems caused by fossil energy, wind power is now forming a boom in the world. However, the distribution of wind turbine units is relatively scattered, with each unit being far apart. Moreover, the cabins of wind turbines are mostly located at a height of several tens of metres, making traditional manual maintenance very difficult. The Artificial Intelligence (AI) wind power equipment Operation and Maintenance (O&M) system can display various technical indicators of the operation of each generator unit in real-time through Big Data (BD) technology, which plays an essential and positive role in reducing the O&M risks of O&M personnel and improving O&M efficiency. This article studied the O&M system and methods of wind power equipment using BD AI technology. The final experimental results showed that the wind power equipment O&M system using BD AI technology had an average maintenance difficulty score of 95.817 points, average maintenance duration of 4.208 hours and an average maintenance cost of 146,300 US dollars, which had significant advantages compared to traditional manual maintenance.
The modified CANDLE (Constant Axial shape of Neutron flux, nuclide densities and power shape During Life of Energy producing reactor) burn-up strategy was utilised effectively in both fast and thermal reactors. In this work, we investigated the neutronic performance of natural uranium-thorium (238U-232Th) composite fuels on a gas-cooled fast reactor employing a modified CANDLE burn-up shuffling in the radial direction. The investigation was carried out on a reactor of 450MWt with natural uranium and natural thorium composite fuels as input for the fuel cycle. The core has been subdivided into ten distinct regions directed radially. Neutronic calculations were carried out using SRAC coding, while JENDL 4.0 was used as a nuclear data library. The various volume fractions of thorium have been mixed with natural uranium as fuel to obtain the impact of natural thorium on the performance of the gas-cooled fast reactor design. The increased volume fraction of thorium caused a decrease in the effective multiplication factor but didn't affect the burn-up level significantly. The active core height has been varied to investigate its impact on reactor performance. The discharge burn-up level for 165 cm active core height is about 345 MWd/ton HM, while 170 cm active core is about 336 MWd/ton HM.
In response to the complex situation where the accuracy of describing the dynamic behaviour of the power system (PS) is low and the control strategy is difficult to cope with dynamic changes, this article combines differential equation models and power data to study the dynamic response analysis and control of the PS. Firstly, electricity data was collected from a certain power company, and the data was cleaned and standardised. Then, differential equation models were constructed for the generators, loads, and transmission lines in the PS, describing their dynamic behaviour and discretising the model. The MPC (Model Predictive Control) algorithm was used to define the objective function, set constraints, and solve the problem. The combination of differential equation modelling and MPC algorithm has improved the accuracy of describing the dynamic behaviour of the PS, and has good adaptability to complex dynamic changes, ensuring the safe and stable operation of the PS.
This paper optimised the power grid equipment fault prediction model based on ML and high-performance computing, analysed the application of high-performance computers in online fault prediction and designed the overall structure of the mechanical equipment fault prediction and detection model. It explains the data classification and prediction in ML, describes how to establish prediction models and applies different ML algorithms to power grid equipment fault prediction models. Through experiments, comparing the optimisation effects of varying ML algorithms on power grid equipment fault prediction models, it was found that the Least Squares Support Vector Machine (LS-SVM) prediction algorithm has the highest accuracy and the best optimisation effect on power grid equipment fault prediction models. After using the LS-SVM prediction algorithm, the entire fault prediction time has been shortened.
