Search Results (3,476)

Precisely predicting photovoltaic (PV) output is crucial for reliable grid integration; so far, most models rely on site-specific sensor data or treat large meteorological datasets as black boxes. This study proposes an explainable machine-learning framework that simultaneously ranks the most informative weather parameters and reveals their physical relevance to PV generation. Starting from 27 local and plant-level variables recorded at 15 min resolution for a 1 MW array in Çanakkale region, Türkiye (1 August 2022–3 August 2024), we apply a three-stage feature-selection pipeline: (i) variance filtering, (ii) hierarchical correlation clustering with Ward linkage, and (iii) a meta-heuristic optimizer that maximizes a neural-network R² while penalizing poor or redundant inputs. The resulting subset, dominated by apparent temperature and diffuse, direct, global-tilted, and terrestrial irradiance, reduces dimensionality without significantly degrading accuracy. Feature importance is then quantified through two complementary aspects: (a) tree-based permutation scores extracted from a set of ensemble models and (b) information gain computed over random feature combinations. Both views converge on shortwave, direct, and global-tilted irradiance as the primary drivers of active power. Using only the selected features, the best model attains an average R² ≅ 0.91 on unseen data. By utilizing transparent feature-reduction techniques and explainable importance metrics, the proposed approach delivers compact, more generalized, and reliable PV forecasts that generalize to sites lacking embedded sensor networks, and it provides actionable insights for plant siting, sensor prioritization, and grid-operation strategies. Full article

This study proposes a hybrid approach to analyze and forecast non-stationary financial time series by combining statistical models with deep neural networks. A model is introduced that integrates three key components: the Generalized Autoregressive Score (GAS) model, which captures volatility dynamics; an attention mechanism (ATT), which identifies the most relevant features within the sequence; and a Long Short-Term Memory (LSTM) neural network, which receives the outputs of the previous modules to generate price forecasts. This architecture is referred to as GAS-ATT-LSTM. Both unidirectional and bidirectional variants were evaluated using real financial data from the Nasdaq Composite Index, Invesco QQQ Trust, ProShares UltraPro QQQ, Bitcoin, and gold and silver futures. The proposed model’s performance was compared against five benchmark architectures: LSTM Bidirectional, GARCH-LSTM Bidirectional, ATT-LSTM, GAS-LSTM, and GAS-LSTM Bidirectional, under sliding windows of 3, 5, and 7 days. The results show that GAS-ATT-LSTM, particularly in its bidirectional form, consistently outperforms the benchmark models across most assets and forecasting horizons. It stands out for its adaptability to varying volatility levels and temporal structures, achieving significant improvements in both accuracy and stability. These findings confirm the effectiveness of the proposed hybrid model as a robust tool for forecasting complex financial time series. Full article

This paper presents an advanced time-series forecasting methodology that integrates multiple machine learning models to improve data prediction in enhanced long-range navigation (eLoran) systems. The analysis evaluates five forecasting approaches: multivariate linear regression, long short-term memory (LSTM) networks, random forest (RF), a fusion model combining LSTM and RF, and a dynamic weighting (DW) model. The results demonstrate that the DW model achieves the highest prediction accuracy while maintaining strong computational efficiency, making it particularly suitable for real-time applications with stringent performance requirements. Although the LSTM model effectively captures temporal dependencies, it demands considerable computational resources. The hybrid model utilises the strengths of LSTM and RF to enhance the accuracy but involves extended training times. By contrast, the DW model dynamically adjusts the relative contributions of LSTM and RF on the basis of the data characteristics, thereby enhancing the accuracy while significantly reducing the computational demands. Demonstrating exceptional performance on the ASF2 dataset, the DW model provides a well-balanced solution that combines precision with operational efficiency. This research offers valuable insights into optimising additional secondary phase factor (ASF) prediction in eLoran systems and highlights the broader applicability of real-time forecasting models. Full article

The rising frequency of heatwaves, combined with the urban heat island effect, increases the population’s exposure to high temperatures, significantly impacting the health of vulnerable groups and the overall well-being of residents. While mesoscale meteorological models can reliably forecast temperatures across urban neighbourhoods, dense networks of in situ measurements offer more precise data at the street scale. In this work, the Random Forest technique was used to predict street-scale temperatures in the downtown area of Thessaloniki, Greece, during a prolonged heatwave in July 2021. The model was trained using data from a low-cost sensor network, meteorological fields calculated by the mesoscale model MEMO, and micro-environmental spatial features. The results show that, although the MEMO temperature predictions achieve high accuracy during nighttime compared to measurements, they exhibit inconsistent trends across sensor locations during daytime, indicating that the model does not fully account for microclimatic phenomena. Additionally, by using only the observed temperature as the target of the Random Forest model, higher accuracy is achieved, but spatial features are not represented in the predictions. In contrast, the most reliable approach to incorporating spatial characteristics is to use the difference between observed and mesoscale temperatures as the target variable. Full article

Accurate forecasting of river water flow helps to warn of floods and droughts in advance, provides a basis for the rational allocation of water resources, and at the same time, offers important support for the safe operation of hydropower stations and water conservancy projects. Water flow is characterized by time series, but the existing models focus on the positive series when LSTM is applied, without considering the different contributions of the water flow series to the model at different moments. In order to solve this problem, this study proposes a river water flow prediction model, named AT-BiLSTM, which mainly consists of a bidirectional layer and an attention layer. The bidirectional layer is able to better capture the long-distance dependencies in the sequential data by combining the forward and backward information processing capabilities. In addition, the attention layer focuses on key parts and ignores irrelevant information when processing water flow data series. The effectiveness of the proposed method was validated against an actual dataset from the Shizuishan monitoring station on the Yellow River in China. The results confirmed that compared with the RNN model, the proposed model significantly reduced the MAE, MSE, and RMSE on the dataset by 27.16%, 42.01%, and 23.85%, respectively, providing the best predictive performance among the six compared models. Moreover, this attention mechanism enables the model to show good performance in 72 h (3 days) forecast, keeping the average prediction error below 6%. This implies that the proposed hybrid model could provide a decision base for river flow flood control and resource allocation. Full article

Temporal domain generalization is crucial for the temporal forecasting of sensor data due to the non-stationary and evolving nature of most sensor-generated time series. However, temporal dynamics vary in scale, semantics, and structure, leading to distribution shifts that a single model cannot easily generalize over. Additionally, conflicts between temporal domain-specific patterns and limited model capacity make it difficult to learn shared parameters that work universally. To address this challenge, we propose an ensemble learning framework that leverages multiple domain-specific models to improve temporal domain generalization for sensor data forecasting. We first segment the original sensor time series into distinct temporal tasks to better handle the distribution shifts inherent in sensor measurements. A meta-learning strategy is then applied to extract shared representations across these tasks. Specifically, during meta-training, a recurrent encoder combined with variational inference captures contextual information for each task, which is used to generate task-specific model parameters. Relationships among tasks are modeled via a self-attention mechanism. For each query, the prediction results are adaptively reweighted based on all previously learned models. At inference, predictions are directly generated through the learned ensemble mechanism without additional tuning. Extensive experiments on public sensor datasets demonstrate that our method significantly enhances the generalization performance in forecasting across unseen sensor segments. Full article

Energy storage technologies are becoming increasingly important when it comes to maintaining the balance between electricity generation and consumption, especially with the increasing share of variable renewable energy sources (VRES). Pumped storage hydropower plants (PSHs) are currently the largest form of energy storage at the grid level. The aim of this study is to investigate the importance and prospects of using PSHs as part of the energy transition to decarbonize energy sources. A comparison was made between PSHs and battery energy storage systems (BESSs) in terms of technical, economic, and ecological aspects. To identify the key factors influencing the wider adoption of PSHs, a combined approach using SWOT analysis (which assesses strengths, weaknesses, opportunities, and threats) and the Analytical Hierarchy Process (AHP) as a decision support tool was applied. Regulatory and market uncertainties (13.54%) and financial inequality (12.77%) rank first and belong to the “Threats” group, with energy storage capacity (10.11%) as the most important factor from the “Strengths” group and increased demand for energy storage (9.01%) as the most important factor from the “Opportunities” group. Forecasts up to 2050 show that the capacity of PSHs must be doubled to enable the integration of 80% of VRES into the grids. The study concludes that PSHs play a key role in the energy transition, especially for long-term energy storage and grid stabilization, while BESSs offer complementary benefits for short-term storage and fast frequency regulation. Recommendations to policymakers include the development of clear, accelerated project approval procedures, financial incentives, and support for hybrid PSH systems to accelerate the energy transition and meet decarbonization targets. Full article

Accurately identifying the sources of fine particulate matter (PM_2.5) pollution is crucial for pollution control and public health protection. Taking the PM_2.5 pollution event that occurred in Suzhou in June 2023 as a typical case, this study analyzed the characteristics and components of PM_2.5, and quantified the contributions of meteorological conditions, regional transport, and local emissions to the summertime PM_2.5 surge in a typical Yangtze River Delta (YRD) city. Chemical composition analysis highlighted a sharp increase in nitrate ions (NO₃⁻, contributing up to 49% during peak pollution), with calcium ion (Ca²⁺) and sulfate ion (SO₄²⁻) concentrations rising to 2 times and 7.5 times those of clean periods, respectively. Results from the random forest model demonstrated that emission sources (74%) dominated this pollution episode, significantly surpassing the meteorological contribution (26%). The Weather Research and Forecasting model combined with the Community Multiscale Air Quality model (WRF–CMAQ) further revealed that local emissions contributed the most to PM_2.5 concentrations in Suzhou (46.3%), while external transport primarily originated from upwind cities such as Shanghai and Jiaxing. The findings indicate synergistic effects from dust sources, industrial emissions, and mobile sources. Validation using electricity consumption and key enterprise emission data confirmed that intensive local industrial activities exacerbated PM_2.5 accumulation. Recommendations include strengthening regulations on local industrial and mobile source emissions, and enhancing regional joint prevention and control mechanisms to mitigate cross-boundary transport impacts. Full article

The fast expansion of cloud computing has raised carbon emissions and energy usage in cloud data centers, so creative solutions for sustainable resource management are more necessary. This work presents a new algorithm—Carbon-Aware, Energy-Efficient, and SLA-Compliant Virtual Machine Placement using Deep Q-Networks (DQNs) and Agglomerative Clustering (CARBON-DQN)—that intelligibly balances environmental sustainability, service level agreement (SLA), and energy efficiency. The method combines a deep reinforcement learning model that learns optimum placement methods over time, carbon-aware data center profiling, and the hierarchical clustering of virtual machines (VMs) depending on resource constraints. Extensive simulations show that CARBON-DQN beats conventional and state-of-the-art algorithms like GRVMP, NSGA-II, RLVMP, GMPR, and MORLVMP very dramatically. Among many virtual machine configurations—including micro, small, high-CPU, and extra-large instances—it delivers the lowest carbon emissions, lowered SLA violations, and lowest energy usage. Driven by real-time input, the adaptive decision-making capacity of the algorithm allows it to dynamically react to changing data center circumstances and workloads. These findings highlight how well CARBON-DQN is a sustainable and intelligent virtual machine deployment system for cloud systems. To improve scalability, environmental effect, and practical applicability even further, future work will investigate the integration of renewable energy forecasts, dynamic pricing models, and deployment across multi-cloud and edge computing environments. Full article

This study investigates the evaluation of housing policies for migrants in China, focusing on the interplay between rental and purchase decisions under the rent-and-purchase policy (RPP) framework. Employing a system dynamics model, we simulate migrant housing choices from 2001 to 2023 and forecast market trends from 2024 to 2030. The results indicate that RPPs significantly improve housing quality and reduce costs for migrants by mitigating institutional disparities and market distortions. Scenario analyses demonstrate that a coordinated approach combining supply-side interventions (e.g., affordable housing expansion) with rights-based policies (e.g., equalizing renter and buyer rights) effectively balances affordability and demand stability. The findings emphasize the critical role of addressing rights inequalities and advocate for a holistic policy framework to tackle migrant housing challenges, offering actionable insights for policymakers in system science and urban planning. Full article

Spatiotemporal correlations between meteo-inputs and wind–solar outputs in an optimal regional scale are crucial for developing robust models, reliable in mid-term prediction time horizons. Modelling border conditions is vital for early recognition of progress in chaotic atmospheric processes at the destination of interest. This approach is used in differential and deep learning; artificial intelligence (AI) techniques allow for reliable pattern representation in long-term uncertainty and regional irregularities. The proposed day-by-day estimation of the RE production potential is based on first data processing in detecting modelling initialisation times from historical databases, considering correlation distance. Optimal data sampling is crucial for AI training in statistically based predictive modelling. Differential learning (DfL) is a recently developed and biologically inspired strategy that combines numerical derivative solutions with neurocomputing. This hybrid approach is based on the optimal determination of partial differential equations (PDEs) composed at the nodes of gradually expanded binomial trees. It allows for modelling of highly uncertain weather-related physical systems using unstable RE. The main objective is to improve its self-evolution and the resulting computation in prediction time. Representing relevant patterns by their similarity factors in input–output resampling reduces ambiguity in RE forecasting. Node-by-node feature selection and dynamical PDE representation of DfL are evaluated along with long-short-term memory (LSTM) recurrent processing of deep learning (DL), capturing complex spatio-temporal patterns. Parametric C++ executable software with one-month spatial metadata records is available to compare additional modelling strategies. Full article

Italy strives to meet its renewable energy targets for 2030 and 2050, with photovoltaic (PV) technology playing a central role. However, the push for increased solar adoption, spurred by past incentive schemes such as “Conto Energia” and “Superbonus 110%”, raises long-term challenges related to PV waste management. In this study, we present a multi-scale approach to forecast End-of-Life (EoL) PV waste across Italy’s 20 regions, aiming to support national circular economy strategies. Historical installation data (2008–2024) were collected and combined with socio-economic and energy-related indicators to train a Backpropagation Neural Network (BPNN) for regional PV capacity forecasting up to 2050. Each model was optimised and validated using R² and RMSE metrics. The projections indicate that current trends fall short of meeting Italy’s decarbonisation targets. Subsequently, by applying a Weibull reliability function under two distinct scenarios (Early-loss and Regular-loss), we estimated the annual and regional distribution of PV panels reaching their EoL. This analysis provides spatially explicit insights into future PV waste flows, essential for planning regional recycling infrastructures and ensuring sustainable energy transitions. Full article

Due to its rising carbon footprint, new paradigms for carbon-efficient computing are needed. For distributed computing systems, one option is to shift computing loads in space or time to take advantage of low-carbon electricity, a paradigm known as carbon-aware computing. We present a literature review of carbon-aware scheduling techniques, which shows that most of the literature carried out either spatial or temporal shifting but not both. Of the 28 analyzed studies, 11 considered both spatial and temporal shifting, and only 2 developed a combined optimization algorithm. Additionally, existing approaches typically focus on operational electricity alone. With the growing decarbonization of electricity, however, device production (which involves various industrial processes and cannot be easily decarbonized) is bound to become more relevant and needs to be considered. We thus suggest a novel spatio-temporal scheduling algorithm for cloud and edge computing. Our algorithm performs simultaneous spatio-temporal shifting while taking into consideration both device production and operation. As temporal shifting requires forecasts of future workloads, we also put forward a workload predictor. Although not fully implemented yet, we bring various theoretical arguments in support of our proposed algorithm. Full article

Previous work used machine learning (ML) to estimate past and current alfalfa yields and showed that domain adaptation (DA) with data synthesis shows promise in classifying yields as high, medium, or low. The current work uses similar techniques to forecast future alfalfa yields. A novel technique is proposed for forecasting alfalfa time series data that exploits stationarity and predicts differences in yields rather than the yields themselves. This forecasting technique generally provides more accurate forecasts than the established ARIMA family of forecasters for both univariate and multivariate time series. Furthermore, this ML-based technique is potentially easier to use than the ARIMA family of models. Also, previous work is extended by showing that DA with data synthesis also works well for predicting continuous values, not just for classification. The novel scale-invariant tabular synthesizer (SITS) is proposed, and it is competitive with or superior to other established synthesizers in producing data that trains strong models. This synthesis algorithm leads to R scores over 100% higher than an established synthesizer in this domain, while ML-based forecasters beat the ARIMA family with symmetric mean absolute percent error (sMAPE) scores as low as 12.81%. Finally, ML-based forecasting is combined with DA (ForDA) to create a novel pipeline that improves forecast accuracy with sMAPE scores as low as 9.81%. As alfalfa is crucial to the global food supply, and as climate change creates challenges with managing alfalfa, this work hopes to help address those challenges and contribute to the field of ML. Full article

The ability to predict capacity retention is critical for ensuring the long-term reliability of supercapacitors in energy storage systems. This study presents a comprehensive machine learning framework that integrates both electrical cycling data and experimentally derived material and structural features to forecast the degradation behavior of commercial supercapacitors. A total of seven supercapacitor samples were tested under various current and voltage conditions, resulting in over 70,000 charge–discharge cycles across three case studies. In addition to electrical measurements, detailed physical and material characterizations were performed, including electrode dimension analysis, Scanning Electron Microscopy (SEM), Energy Dispersive X-ray Spectroscopy (EDS), and Thermogravimetric Analysis (TGA). Three machine learning models, Linear Regression (LR), Random Forest (RF), and Multi-Layer Perceptron (MLP), were trained using both cycler-only and combined cycler + material features. Results show that incorporating material features consistently improved prediction accuracy across all models. The MLP model exhibited the highest performance, achieving an R² of 0.976 on the training set and 0.941 on unseen data. Feature importance analysis confirmed that material descriptors such as porosity, thermal stability, and electrode thickness significantly contributed to model performance. This study demonstrates that combining electrical and material data offers a more holistic and physically informed approach to supercapacitor health prediction. The framework developed here provides a practical foundation for accurate and robust lifetime forecasting of commercial energy storage devices, highlighting the critical role of material-level insights in enhancing model generalization and reliability. Full article