Abstract

Global floor area is increasing every year which is subsequently leading to an increase in electricity and heating demand in buildings. Residential buildings have huge potential for energy savings and there is an immediate need to decarbonize them by the end of 2050. Machine learning is finding application across all fields and will thus have an important role to play in the building sector also. One of the important challenges that building owners need to tackle is occupancy detection in residential apartments which can help save considerable amounts of energy and costs. However, occupancy is highly variable, and it is difficult to quantify and predict occupancy because of the random and individualistic nature of humans. In addition, scalable approaches for occupancy detection should prioritize data from common and cost-effective sensors like temperature sensors. In contrast to existing literature which has stated that occupancy detection based on the data from a single environmental sensor is not appropriate for obtaining good results, this paper aims to detect occupancy in a real residential building using only indoor temperature as the feature to train the model. Different machine learning models and techniques are studied and tested to understand how the accuracy of occupancy detection can be increased. With the right techniques, it has been possible to obtain promising results in the form of an accuracy of 95% using machine learning models and only indoor temperature to train it.

Abstract

Resources and raw materials are key enablers for the energy and industrial transformation needed to com-bat current climate and socio-ecological crises. Important substitution processes, such as defossilisation, shape the trajectory of the new raw material supply chain. Whenever raw materials or resources are sourced from the ground, processed, used and recycled, they can create unintended environmental, social and justice issues, leading to conflicting sustainability goals. Due to the complexity of interlinked effects, identifying and addressing these conflicts requires interdisciplinary collaboration. The aim of this paper is to propose sustainability risks of raw material supply as an innovative topic, explored from four different perspectives: (1) geo- and environmental science, 2) engineering, (3) economics and social science, and (4) material science, informed by sustainability science. As a method, we propose a five-step approach towards a comprehensive research agenda. Our approach establishes and builds on a common un-derstanding of materials. As a result, we introduce three material categories, namely (1) fossil energy re-sources, (2) base or mass materials, (3) critical raw materials (step 1) and go through all five steps. Our results enable an extended understanding of sustainability risks of raw materials and resources, informed by relevant disciplinary aspects (step 2), normative insights from sustainability science (step3), and guiding questions for an expert assessment of blind spots and links to other disciplines and knowledge types (step 4). We discuss future research directions (step 5) and the need for accompanying research. In conclusion, an interdisciplinary approach is essential to understand sustainability risks of materials and resources and enable science to contribute meaningfully and help guide efforts to combat current socio-ecological crises.

Abstract

In power systems with high shares of variable renewable energy sources (vRES), balancing generation and demand while maintaining frequency stability becomes increasingly challenging. Reduced system inertia and renewable variability require flexible resources capable of providing fast reserves and absorbing surplus energy. This study assesses the role of Battery Energy Storage Systems (BESS) in supporting the Italian power system in 2030 through two services: energy time-shifting and frequency reserve provision. Future reserve needs are estimated using two different approaches, formula-based and convolution-based, applied to historical forecast errors for load, solar, and wind generation. Both methods indicate a marked increase in reserve requirements from 2023 to 2030: under the formula approach, upward and downward reserves rise by 20% and 62%, while the convolution method estimates increases of 12% and 31%, respectively. Building on these results, the analysis simulates a BESS fleet, sized according to Italy’s NECP targets, operating under a price-driven dispatch strategy to assess its contribution to system balancing. Results show around 214 full cycles annually, delivering 11 TWh of peak-shaving energy and covering up to 50% of upward and 79% of downward reserves. Reserve provision peaks in spring and autumn, reaching up to 90% downward participation on weekdays, driven by strong renewable output and price signals. An economic evaluation suggests that the total reserve provision costs amount to approximately €185 million per year, with BESS accounting for €66 million per year. Overall, the findings highlight the strategic role of BESS as a dual-purpose resource delivering both energy arbitrage and ancillary services and offer insights to support future market design and grid stability in high-renewable systems.

Abstract

The energy transition is a complex socio-technical transformation shaped not only by techno-economic factors, but also by geopolitical and social dynamics. However, these dynamics are hardly covered by quantitative energy system models. Qualitative scenario development enriched with a new AI-based narrative approach offers a promising way to communicate geopolitical dynamics, such as global power dynamics, while generating consistent storylines. Therefore, this paper aims to explore the potential of AI-based scenario narrative generation in the context of the European Energy – Geopolitics – Society nexus and outlines complementary pathways for future quantification. We employ a qualitative foresight approach based on the Cross-Impact Balance (CIB) method and a new AI-based narrative approach, which constitutes the methodological innovation of this study. As a result, we present five consistent future scenarios for Europe in 2050 and corresponding storylines, produced through a novel AI-driven storytelling approach, as well as a numerical scenario comparison. Our findings reveal practical learnings from the use of AI in scenario development, including prompting strategies, input preparation, and output evaluation. We discuss opportunities and challenges of the AI-based method and future research directions for quantifying qualitative narratives, proposing GENeSYS-MOD as a suitable global energy system model. In conclusion, new geopolitical dynamics of the energy transition can be effectively communicated – particularly through AI-generated scenario narratives – when their limitations are carefully considered and they are complemented by links to quantitative energy system models.

Abstract

Facing the challenge of growing heating and cooling demand, more heat recovery materials used to capture and transfer heat should be developed. Melting point prediction is one of the most essential steps of machine-learning-driven material discovery. Existing methods rely on large datasets, often with the help of density function theory and molecular dynamics simulations, or complex deep learning methods. Here, we developed a multilayer perceptron model for predicting the melting point of the materials using a small but reliable dataset. By comparing with classical machine learning models, including ridge regression, random forest, XGBoost, and support vector machines, the multilayer perceptron model with transfer learning from a much larger dataset not dedicated to heat recovery materials demonstrated a flexible compromise balancing robustness and data size, providing a promising yet simple solution for melting point prediction.

Abstract

The European Energy Exchange (EEX) is the leading electricity trading platform in Europe, covering spot, intraday, and futures markets. Owing to the non-storability of electricity, high renewable penetration, frequent price spikes, and evolving regulatory frameworks, EEX provides a representative setting for electricity market modeling. This paper reviews major modeling approaches applied to the EEX. For spot markets, econometric time-series models, jump-diffusion processes, regime-switching frameworks, and machine learning methods are compared in terms of their ability to capture seasonality, volatility, price spikes, and negative prices, while for futures markets, studies on forward pricing, risk premiums, and derivative valuation are examined. The impacts of renewable energy are highlighted, showing that wind and solar generation reduce average prices, increase volatility due to forecast uncertainty, and stimulate demand for renewable-linked products. Overall, the synthesis indicates that hybrid and policy-sensitive models are essential for forecasting, risk management, and policy evaluation in electricity markets under energy transition.

Abstract

Economic incentive-driven demand response is key to shaping residential electricity consumption patterns, yet current strategies often lack appliance-level incentive models for household energy systems, coupled with difficulties in collecting behavioral data. Therefore, an interactive gamified simulation platform has been developed, where users can independently manage appliances, weighing economic rewards against discomfort from adjusted usage. A System Usability Scale (SUS) evaluation confirms the platform’s usability and reliability. Experimental findings highlight a non linear association between economic incentives and the acceptable adjustment range of common appliances: the adjustable range expands sharply when incentives increase from 1 to 3 yuan, but further increments yield diminishing returns in flexibility. Notably, air conditioners and water heaters demonstrate the highest flexibility potential, with 0.5 kWh and 1.5 kWh of regulatable capacity respectively under a 2-yuan incentive. Overall, the quantitative models obtained in this study can provide data support for appliances adjustable potential, and help to realize more humanized home energy management which meet user needs and preferences.

Abstract

Cold storage Carnot batteries (CSCBs) have emerged as a promising solution for long-duration energy storage by integrating ice-based cold storage. However, existing architectures are constrained by limited efficiency caused by suboptimal discharge cycles. In this study, two modified discharge cycles, recuperated and regenerative configurations, are investigated within the CSCB system framework. Additionally, two low-GWP working fluids, R1234ze(E) and R1234yf, are evaluated. Results show that the regenerative cycle provides a higher efficiency improvement than the recuperated cycle (7.3% versus 4.2%), and R1234ze(E) outperforms R1234yf across all discharge configurations. Increasing the cold storage temperature can enhance the overall system round-trip efficiency (RTE). With ice/water as a storage medium, the proposed CSCB system achieves an RTE of approximately 78.3%. This study provides insights into modifying the CSCB system to achieve high-efficiency, sustainable, long-duration energy storage.

Abstract

Efficient development of shale oil and gas in China relies on factory operations and large-scale fracturing technology. Large-scale fracturing of shale oil and gas requires a long time and numerous equipment and facilities, with frequent and severe incidents of fracturing sand blockage. The research on early warning research in these incidents is crucial for the safety of shale oil and gas fracturing operations. However, the effective methods for analyzing the main control factors of fracturing sand blockage and predicting the pump pressure during operations are lacked. To study this issue, considering the fracturing mechanism and pump pressure variation characteristics, a method for real-time prediction of pump pressure during fracturing operations has been established to conduct sand blockage early warning research here. First, a fracturing simulator was used to simulate the entire process of pump pressure changes during fracturing. By altering different fluid properties and formation parameters, the main control factors of pump pressure variation were analyzed, and the grey correlation analysis method was used to rank these factors. Secondly, based on fracture mechanics, proppant transport theory, and the Long Short-Term Memory (LSTM) neural network model, a framework and model for predicting pump pressure during operations was established, forming a method for early warning of fracturing sand blockage under the integration of mechanism and intelligence. Finally, the early warning method for sand blockage was applied to actual field fracturing operations. Results indicate that the factors affecting the pump pressure of a typical well, from most to least significant, are discharge rate, fluid viscosity, differential principal stress, sand concentration, number of fracture clusters, and number of perforations. When other parameters remain constant, as fluid viscosity, differential principal stress, and discharge rate increase, the pump pressure increases; as the number of fracture clusters, perforations, and sand concentration increase, the pump pressure decreases. This method can be used for the identification and early warning of fracturing sand blockage incidents in the actual field operations, which is 19 seconds earlier than on-site manual identification, with a relative error of about 6.8%. The predicted pump pressure is friendly matched with the actual field one, which is helpful in accurate early warning of fracturing sand blockage.

Abstract

The accurate measurement of downhole CO₂ injection rate is of great practical significance for optimizing enhanced oil recovery (EOR) effects, evaluating reservoir performance, and designing and adjusting development plans. To improve the accuracy of CO₂ injection well flow rate measurement, a downhole CO₂ injection flow rate measurement method integrating mechanism and data was proposed. Data were acquired and preprocessed using an indoor experimental platform for CO₂ flow rate measurement, and a mechanistic model, a data-driven model, and a hybrid-driven model were constructed. The accuracies of the three models were compared and analyzed. The research results show that: the mechanistic model can be used for downhole CO₂ flow rate measurement, with most errors ranging from 9% to 20%. Relying on its flexible application of multi-source data, relatively simple construction process, and strong adaptability, the data-driven model exhibits higher accuracy in CO₂ flow rate measurement than the mechanistic model, with an accuracy of 87.3%. The hybrid-driven model integrates the in-depth interpretability of the mechanistic model and the efficient learning ability of the data-driven model for massive data. Its comprehensive performance in accuracy, adaptability, and interpretability surpasses that of single models, with an accuracy of 94.7%. The application of this hybrid-driven model in an oil well in southern China shows that its accuracy is 92.3%, which can meet the actual on-site measurement requirements.