Energy and AI最新文献

The energy sector faces rapid decarbonisation and decision-makers demand reliable assessments of the security of electricity supply. For this, detailed simulation models with a high temporal and technological resolution are required. When confronted with increasing weather-dependent renewable energy generation, probabilistic simulation models have proven. The significant computational costs of calculating a scenario, however, limit the complexity of further analysis. Advances in code optimization as well as the use of computing clusters still lead to runtimes of up to eight hours per scenario. However ongoing research highlights that tailor-made approximations are potentially the key factor in further reducing computing time. Consequently, current research aims to provide a method for the rapid prediction of widely varying scenarios. In this work artificial neural networks (ANN) are trained and compared to approximate the system behavior of the probabilistic simulation model. To do so, information needs to be sampled from the probabilistic simulation in an efficient way. Because only a limited space in the whole design space of the 16 independent variables is of interest, a classification is developed. Finally it required only around 35 min to create the regression models, including sampling the design space, simulating the training data and training the ANNs. The resulting ANNs are able to predict all scenarios within the validity range of the regression model with a coefficient of determination of over 0.9998 for independent test data (1.051.200 data points). They need only a few milliseconds to predict one scenario, enabling in-depth analysis in a brief period of time.

Building geometry data is crucial for detailed, spatially-explicit analyses of the building stock in energy systems analysis and beyond. Despite the existence of diverse datasets and methods, a standardized and validated approach for creating a nation-wide unified and complete dataset of German building heights is not yet available. This study develops and validates such a methodology, combining different data sources for building footprints and heights and filling gaps in height data using an XGBoost machine learning algorithm. The XGBoost model achieves a mean absolute error of 1.78 m at the national level and between 1.52 m and 3.47 m at the federal state level. The goal is proving the applicability of the methodology at a large scale and creating a useful dataset. The resulting dataset is thoroughly evaluated on a building-by-building level and spatially resolved statistics on the quality of the dataset are reported. This detailed validation found that the building number and footprint area of German building stock is 90.31 % and 94.84 % correct, respectively, and the building height accuracy is 0.59 m at the national level. However, errors are not homogeneous across Germany and further research is needed into the impact of including additional datasets, especially for regions and building types with lower accuracies. This study proves that the chosen methodology is useful for generating a building height dataset and the workflow, with some modifications for regional data availability, can be transferred to other countries. The generated building dataset for Germany constitutes a valuable data basis for the research community in fields such as energy research, urban planning and building decarbonization policy development.

Proton exchange membrane (PEM) based electrochemical systems have the capability to operate in fuel cell (PEMFC) and water electrolyser (PEMWE) modes, enabling efficient hydrogen energy utilisation and green hydrogen production. In addition to the essential cell stacks, the system of PEMFC or PEMWE consists of four sub-systems for managing gas supply, power, thermal, and water, respectively. Due to the system's complexity, even a small fluctuation in a certain sub-system can result in an unexpected response, leading to a reduced performance and stability. To improve the system's robustness and responsiveness, considerable efforts have been dedicated to developing advanced control strategies. This paper comprehensively reviews various control strategies proposed in literature, revealing that traditional control methods are widely employed in PEMFC and PEMWE due to their simplicity, yet they suffer from limitations in accuracy. Conversely, advanced control methods offer high accuracy but are hindered by poor dynamic performance. This paper highlights the recent advancements in control strategies incorporating machine learning algorithms. Additionally, the paper provides a perspective on the future development of control strategies, suggesting that hybrid control methods should be used for future research to leverage the strength of both sides. Notably, it emphasises the role of artificial intelligence (AI) in advancing control strategies, demonstrating its significant potential in facilitating the transition from automation to autonomy.

Regenerative fuel cells can operate alternately as an electrolyzer and as a fuel cell, frequently involving water as a reactant or product. Modifying the electrode surface to manipulate water can prevent electrode flooding and enhance the electrode's mass transfer efficiency by facilitating better contact with gaseous reactants. However, conventional electrodes face difficulties in allowing water droplets to penetrate in a single direction leaving electrodes. In this work to address this issue, a wettability gradient electrode is designed and fabricated for efficient water manipulation in regenerative fuel cells. The findings demonstrate that the water removal strategy in the electrolyzer mode yields the highest ammonia yield and Faradaic efficiency of 3.39 × 10^-10 mol s^-1 cm^-2 and 0.49 %, respectively. Furthermore, in the fuel cell mode, the discharging process sustains for approximately 20.5 h, which is six times longer than the conventional strategy. The ability to sustain the discharging process for extended periods is particularly advantageous in regenerative fuel cells, as it enables the cells to operate for longer periods without the need for regeneration.

The durability of proton exchange membrane fuel cells (PEMFC) is an important issue that restricts their large-scale application. To improve their reliability during use, this paper proposes a short-term performance degradation prediction model using particle swarm optimization (PSO) to optimize the gate recurrent unit (GRU). After training using only the data from the first 300 h, good prediction accuracy can be achieved. Compared with the traditional GRU algorithm, the proposed prediction method reduces the root mean square error (RMSE) and mean absolute error (MAE) of the prediction results by 44.8 % and 35.1 %, respectively. It avoids the problem of low accuracy in predicting performance during the temporary recovery phase in traditional GRU models, which is of great significance for the health management of PEMFC.

The industrial metallization of Si solar cells predominantly relies on screen printing, with silver as the preferred electrode material. However, the design of commercial screens often leads to suboptimal silver usage and increased electrical resistance due to print-related inhomogeneities like mesh marks, constrictions and spreading. Real-time monitoring of quality parameters during production has thus become increasingly critical. Current inline optical quality control systems usually only include 2D visualizations of the printed layout, which limits their effectiveness in quality control. Options that allow 3D measurements are usually slow, expensive, and therefore not worth considering in most cases. This research focuses on the development of a model that can estimate the three-dimensional shape of printed contact fingers from a single 2D image without the need of additional hardware using deep learning. Furthermore, a workflow for the generation of training data, which involves the creation of image pairs from a 2D microscope and a 3D confocal laser scanning microscope (CLSM) to accurately represent solar cell fingers, is presented. After model training, the predicted height maps are compared with the ground truth height maps, and the robustness of the model with respect to a paste variation and screen parameter variation is examined. The results confirm the feasibility and reliability of deep learning-based 3D shape estimation, extending its applicability to new, previously unseen data from screen-printed contact fingers. With a structural similarity index (SSIM) score of 0.76, a strong correlation between the estimated and ground truth height maps is established. In summary, our deep learning-based approach for height map estimation offers an effective and reliable solution for fast inline detection and analysis of the cross-sectional area of the printed contact fingers.

Lithium-ion (Li-ion) batteries have emerged as a cornerstone of electric vehicles (EVs), enabling the road transportation towards net zero. The success of electric vehicles largely hinges on the battery performance and safety. It is challenging to test and predict battery performance and safety issues by conventional methods, which are usually time-consuming and expensive, involving significant human and measurement errors. To enable the quick estimation of battery performance and safety, we developed three data-driven machine learning (ML) models, namely a convolutional neural network (CNN), a long short-term memory (LSTM), and a CNN-LSTM to predict battery discharge curves and local maximum temperature (hot spot) under various operating conditions. The developed ML models mitigated data scarcity by employing a three-dimensional multi-physics Li-ion battery model to generate enormous and diverse high-quality data. It was found the CNN-LSTM model outperforms the others and achieved high accuracy of 98.68% to learn discharge curves and battery maximum temperature, owing to the integration of spatial and sequential feature extraction. The battery safety can be improved by comparing the predicted maximum battery temperature against safe temperature threshold. The proposed data development and data-driven ML models are of great potential to provide digital tools for engineering high-performance and safe EVs.

Trucks consume a lot of energy. Hybrid technology maintains a long range while realizing energy savings. Hybrid is therefore an effective energy-saving technology for trucks. Recovery of engine waste heat through the organic Rankine cycle further enhances engine efficiency and provides effective thermal management. However, the powertrain greatly increases the complexity of energy management system. In order to design an energy management system with high efficiency and robustness, this study proposes a deep reinforcement learning embedded rule-based energy management system. This method optimises the key parameters of the rule-based energy management system by inserting deep reinforcement learning into it. Therefore, this scheme combines the good optimization effect of deep reinforcement learning and the excellent robustness of rule. In order to verify the feasibility of this scheme, this study builds the system dynamic model and carries out a simulation study. Subsequently, a hybrid powertrain semi physical experimental bench was constructed and a rapid control prototype experimental study was carried out. The simulation results show that the deep reinforcement learning embedded rule-based energy management system can reduce the energy consumption by 4.31 % compared with the rule-based energy management system under the C-WTVC driving cycle. In addition, energy saving and safe operation can also be achieved under other unfamiliar untrained driving cycles. The rapid control prototype experimental study shows that the deep reinforcement learning embedded rule-based energy management system has good agreement in experiment and simulation, which demonstrates the potential for real vehicle engineering applications and promotes the engineering application of deep reinforcement learning.

The computational demands of 3D continuum models for proton exchange membrane fuel cells remain substantial. One prevalent approach is the hierarchical model combining a 2D/3D flow field with a 1D sub-model for the catalyst layers and membrane. However, existing studies often simplify the 1D domain to a linearized 0D lumped model, potentially resulting in significant errors at high loads. In this study, we present a computationally efficient neural network driven 3D+1D model for proton exchange membrane fuel cells. The 3D sub-model captures transport in the gas channels and gas diffusion layers and is coupled with a 1D electrochemical sub-model for microporous layers, membrane, and catalyst layers. To reduce computational intensity of the full 1D description, a neural network surrogates the 1D electrochemical sub-model for coupling with the 3D domain. Trained by model-generated large synthetic datasets, the neural network achieves root mean square errors of less than 0.2%. The model is validated against experimental results under various relative humidities. It is then employed to investigate the nonlinear distribution of internal states under different operating conditions. With the neural network operating at 0.5% of the computing cost of the 1D sub-model, the hybrid model preserves a detailed and nonlinear representation of the internal fuel cell states while maintaining computational costs comparable to conventional 3D+0D models. The presented hybrid data-driven and physical modeling framework offers high accuracy and computing speed across a broad spectrum of operating conditions, potentially aiding the rapid optimization of both the membrane electrode assembly and the gas channel geometry.

The poppet valves two-stroke (PV2S) aircraft engine fueled with sustainable aviation fuel is a promising option for general aviation and unmanned aerial vehicle propulsion due to its high power-to-weight ratio, uniform torque output, and flexible valve timings. However, its high-altitude gas exchange performance remains unexplored, presenting new opportunities for optimization through artificial intelligence (AI) technology. This study uses validated 1D + 3D models to evaluate the high-altitude gas exchange performance of PV2S aircraft engines. The valve timings of the PV2S engine exhibit considerable flexibility, thus the Latin hypercube design of experiments (DoE) methodology is employed to fit a response surface model. A genetic algorithm (GA) is applied to iteratively optimize valve timings for varying altitudes. The optimization process reveals that increasing the intake duration while decreasing the exhaust duration and valve overlap angles can significantly enhance high-altitude gas exchange performance. The optimal valve overlap angle emerged as 93 °CA at sea level and 82 °CA at 4000 m altitude. The effects of operating parameters, including engine speed, load, and exhaust back pressure, on the gas exchange process at varying altitudes are further investigated. The higher engine speed increases trapping efficiency but decreases the delivery ratio and charging efficiency at various altitudes. This effect is especially pronounced at elevated altitudes. The increase in exhaust back pressure will significantly reduce the delivery ratio and increase the trapping efficiency. This study demonstrates that integrating DoE with AI algorithms can enhance the high-altitude performance of aircraft engines, serving as a valuable reference for further optimization efforts.