Article
Computer Science, Information Systems
Tallataf Rasheed, Abdul Rauf Bhatti, Muhammad Farhan, Akhtar Rasool, Tarek H. M. El-Fouly
Summary: This research presents an Improved Supervised Learning (ISL)-based Deep Neural Network (DNN) that accurately forecasts Electric Vehicles (EVs) load demand. The ISL technique is used to improve prediction performance by refining the training process with additional information and features. The proposed models with ISL show consistent improvements in several architectures and reduce error values, making it significant for the planning and management of EV charging stations.
Article
Energy & Fuels
Xin Tong, Jingya Wang, Changlin Zhang, Teng Wu, Haitao Wang, Yu Wang
Summary: A LSTM-Autoencoder model is proposed to address the weak representation ability and severe loss of time series features in traditional methods for large-scale and complex power load forecasting tasks. Experimental results show that the method outperforms many existing mainstream methods.
Article
Computer Science, Artificial Intelligence
Zhen Fang, Xu Ma, Huifeng Pan, Guangbing Yang, Gonzalo R. Arce
Summary: This paper proposes a long-term forecasting method based on a modified adaptive LSTM model, which improves the prediction performance by capturing important profit points and considering the temporal correlation of FTS.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
K. Venkatachalam, Pavel Trojovsky, Dragan Pamucar, Nebojsa Bacanin, Vladimir Simic
Summary: Weather forecasting plays a crucial role in various aspects of modern society, and this study proposes a deep learning model called LSTM and T-LSTM for accurate weather prediction. Evaluation metrics demonstrate the effectiveness and reliability of the T-LSTM method.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Energy & Fuels
Alessandro Brusaferri, Matteo Matteucci, Stefano Spinelli, Andrea Vitali
Summary: This work presents a novel approach to improve the trustworthiness of neural network based load forecasting systems by integrating predictive distributions and uncertainty sources. Experimental results demonstrate significant performance improvements in load forecasting.
Article
Computer Science, Information Systems
Keon-Jun Park, Sung-Yong Son
Summary: The feasibility of utilizing electricity consumption data is increasing with the widespread use of advanced metering infrastructure among consumers. Load forecasting technology using deep learning has been applied extensively in recent years, but the available data may be limited by privacy regulations. This study introduces a modified federated learning algorithm to address this problem and proposes a method to forecast residential loads. The experimental results show that the proposed method improves forecasting performance and convergence.
Article
Energy & Fuels
Zahra Fazlipour, Elaheh Mashhour, Mahmood Joorabian
Summary: This paper presents an innovative DLSTM-SAE model for short-term load forecasting, equipped with a MSAM to solve the random initial weight problem. Comparative tests using actual data demonstrate the superiority and robustness of the proposed model.
Article
Computer Science, Artificial Intelligence
Donlapark Ponnoprat
Summary: Short-term precipitation forecasting is crucial for human activity planning, and a seasonally-integrated autoencoder (SSAE) model is proposed in this study to handle nonlinearity and detect seasonality in time series. Experimental results show that SSAE outperforms other models in various climates, and the seasonal component helps improve the correlation between forecast and actual values.
APPLIED SOFT COMPUTING
(2021)
Article
Engineering, Civil
Sepideh Emami Tabrizi, Kai Xiao, Jesse Van Griensven The, Muhammad Saad, Hani Farghaly, Simon X. Yang, Bahram Gharabaghi
Summary: In cold climates, road authorities apply salt on roads during winter to ensure public safety. Predicting pavement temperature can optimize road salt application, reduce costs, improve public safety, and decrease environmental impacts. This research developed a reliable and accurate pavement surface temperature prediction tool using machine learning techniques.
JOURNAL OF HYDROLOGY
(2021)
Article
Thermodynamics
Wenyu Zhang, Qian Chen, Jianyong Yan, Shuai Zhang, Jiyuan Xu
Summary: This study proposes a novel asynchronous deep reinforcement learning model for short-term load forecasting to address the challenges of high temporal correlation and high convergence instability. By introducing new methods to disrupt temporal correlation, adaptively judge training situation, and stabilize model training convergence by considering action trends, the proposed model achieves higher forecasting accuracy, less time cost, and more stable convergence compared to eleven baseline models.
Article
Chemistry, Multidisciplinary
Paweena Suebsombut, Aicha Sekhari, Pradorn Sureephong, Abdelhak Belhi, Abdelaziz Bouras
Summary: Water is essential for crop production, but becoming scarce. Proper irrigation scheduling can improve crop yield and quality. Soil Moisture (SM) is a key irrigation parameter. Predicting future soil moisture using machine learning is valuable for water optimization and crop yield.
APPLIED SCIENCES-BASEL
(2021)
Article
Computer Science, Information Systems
Nastaran Gholizadeh, Petr Musilek
Summary: This paper evaluates the performance of federated learning for short-term forecasting of individual house loads as well as the aggregate load, and discusses its advantages and disadvantages by comparing it to centralized and local learning schemes. By proposing a new client clustering method, the convergence time of federated learning can be reduced.
INTERNET OF THINGS
(2022)
Article
Mathematics
M. Zulfiqar, Nahar F. Alshammari, M. B. Rasheed
Summary: Electric vehicles, especially plug-in hybrid electric vehicles (PHEVs), are expected to play a crucial role in future energy systems by assimilating surplus energy from renewable sources. Efforts are being made to develop efficient PHEVs charging solutions to minimize the impact on power infrastructure. Our research proposes a novel machine learning method, specifically Q-learning, which outperforms conventional AI techniques in accurately forecasting PHEV charging loads in different scenarios.
Article
Chemistry, Multidisciplinary
Manuel Lopez-Martin, Antonio Sanchez-Esguevillas, Luis Hernandez-Callejo, Juan Ignacio Arribas, Belen Carro
Summary: This study compares various traditional machine learning and deep learning techniques, as well as new methods for dynamic model analysis and short-term load forecasting. It explores the impact of critical parameters in time series forecasting, including rolling window length, forecast length, and the number/nature of features used.
APPLIED SCIENCES-BASEL
(2021)
Article
Computer Science, Artificial Intelligence
Chao Peng, Yifan Tao, Zhipeng Chen, Yong Zhang, Xiaoyan Sun
Summary: The paper proposes a multi-source transfer learning guided ensemble LSTM method for building multi-load forecasting, which includes a two-stage building matching method and an LSTM modeling strategy to achieve high-precision load forecasting results with limited target building data.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Computer Science, Artificial Intelligence
Lidong Zhang, Tianyu Hu, Zhile Yang, Dongsheng Yang, Jianhua Zhang
Summary: The heat exchanger is an indispensable device in the energy and chemical industry and plays a vital role in optimizing design. This paper proposes a novel algorithm called EDOLSCA for the optimal design of heat exchangers, and its advantages have been validated in practical applications.
NEURAL COMPUTING & APPLICATIONS
(2023)
Article
Energy & Fuels
Amin Amin, Oudom Kem, Pablo Gallegos, Philipp Chervet, Feirouz Ksontini, Monjur Mourshed
Summary: The research team developed a cloud-based framework to optimize electricity consumption and generation in buildings, aiming to improve the participation of EU households in demand response programs. Through electro-thermal simulations and consideration of low-voltage grid constraints, their approach demonstrated success in real-world testing.
Article
Thermodynamics
Xiangfei Liu, Mifeng Ren, Zhile Yang, Gaowei Yan, Yuanjun Guo, Lan Cheng, Chengke Wu
Summary: This paper proposes a novel HVAC control system based on a multi-step predictive deep reinforcement learning algorithm. The system predicts the outdoor ambient temperature using the GC-LSTM algorithm and combines it with the DDPG algorithm to adjust the output power of the HVAC system based on the dynamic changing of electricity prices. Simulation results demonstrate the effectiveness of the system in saving costs while maintaining user comfort.
Article
Construction & Building Technology
Meng Yang, Chengke Wu, Yuanjun Guo, Rui Jiang, Feixiang Zhou, Jianlin Zhang, Zhile Yang
Summary: This paper presents a deep learning model called Spatial Temporal Relation Transformer (STR-Transformer) for automatically identifying risky behaviors in construction sites. By simultaneously extracting and fusing spatial and temporal features from video streams, the STR-Transformer enables more accurate and reliable safety surveillance, with potential to reduce accident rates and management costs.
AUTOMATION IN CONSTRUCTION
(2023)
Review
Construction & Building Technology
Mousa Alrasheed, Monjur Mourshed
Summary: Anthropogenic climate change may lead to overheating of dwellings and increased cooling demand, resulting in higher greenhouse gas emissions from mechanical cooling. This review examines the factors affecting overheating risks in dwellings and explores passive cooling strategies to mitigate their impact on thermal comfort and wellbeing. The effectiveness of passive strategies depends on building design, construction, operation, climate, and occupancy. The study finds that a combination of passive strategies is needed to minimize overheating risks by the 2080s. External solar shading is the most effective method for insulated dwellings, while cool paint is ideal for uninsulated ones. Moreover, occupant interaction is required for optimal air circulation and cooling performance of thermal mass and natural ventilation.
INDOOR AND BUILT ENVIRONMENT
(2023)
Article
Architecture
Tanvir Morshed, Monjur Mourshed
Summary: Buildings in the UK are responsible for one-third of greenhouse gas emissions. Recent building regulations aimed at reducing heating and energy use may be rendered ineffective due to overheating in highly-insulated, airtight buildings in the projected future climate. This research found that airtight office buildings in London will likely overheat in the 2050s, leading to increased electricity consumption and summertime space conditioning. A mixed-mode ventilation strategy is suggested to achieve energy efficiency and meet overheating and emissions targets. Current heating-focused legislation needs to be re-evaluated to consider the effects of the warming climate and overheating risks.
JAPAN ARCHITECTURAL REVIEW
(2023)
Article
Computer Science, Artificial Intelligence
Junjie Jiang, Zhile Yang, Chengke Wu, Yuanjun Guo, Wei Feng
Summary: Recent advancements in computer vision and augmented reality technology have the potential to enhance human-computer interaction, but their integration is not common in the industry. Efficient object detection models are crucial for accurate object localization in augmented reality devices, but the limited computing capabilities and memory usage of wearable AR devices pose challenges for deploying state-of-the-art detectors.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Engineering, Electrical & Electronic
Jiajing Zhou, Zhao An, Zhile Yang, Yanhui Zhang, Huanlin Chen, Weihua Chen, Yalin Luo, Yuanjun Guo
Summary: This paper proposes PT-Informer, a deep learning framework for fault prediction and detection in nuclear power plants. Unlike traditional approaches, PT-Informer extracts fault features from raw vibration signals and achieves ultra-real-time fault prediction. Experimental results demonstrate that PT-Informer outperforms traditional models in terms of prediction accuracy and fault classification.
Article
Engineering, Electrical & Electronic
Lan Cheng, Zhao An, Yuanjun Guo, Mifeng Ren, Zhile Yang, Sean McLoone
Summary: This article proposes a multimodal few-shot learning method (MMFSL) for unbalanced data modeling of industrial bearings. MMFSL can handle time-series data and images, and evaluate the quality of the generated data. Through experiments, the MMFSL model can significantly improve fault detection accuracy and reduce false alarm rate. Moreover, the fault classification model accuracy is also improved compared to the original datasets.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2023)
Article
Energy & Fuels
Amin Amin, Monjur Mourshed
Summary: In this study, a hybrid bottom-up community energy forecasting framework is developed to accurately estimate domestic electricity demand. The framework considers key factors such as demographic characteristics, occupancy patterns, and appliance features, resulting in highly accurate estimations and more reliable demand profiles.
Article
Engineering, Industrial
Lu Zhang, Yi Feng, Qinge Xiao, Yunlang Xu, Di Li, Dongsheng Yang, Zhile Yang
Summary: This paper investigates the difficulties of Dynamic Flexible Job Shop Scheduling (DFJSP) caused by the uncertainties and complexity in the production process due to customized requirements. A new DFJSP model, VPT-FJSP, is proposed and solved using Markov decision process (MDP) and reinforcement learning methods. The experimental results show that the proposed framework outperforms genetic algorithm and ant colony optimization in most cases, demonstrating its effectiveness and robustness.
JOURNAL OF MANUFACTURING SYSTEMS
(2023)
Review
Environmental Sciences
Kumar Biswajit Debnath, Monjur Mourshed
Summary: Bangladesh, as an emerging economy, is heading towards a coal-intensive electricity generation mix due to limited renewable energy potential and energy and food security challenges.
Article
Engineering, Electrical & Electronic
Zhao An, Lan Cheng, Yuanjun Guo, Mifeng Ren, Wei Feng, Bo Sun, Jun Ling, Huanlin Chen, Weihua Chen, Yalin Luo, Zhile Yang
Summary: In this paper, a deep learning fault detection and prediction framework combining PCA and Informer is proposed to solve the problem of online monitoring of nuclear power valves. The effectiveness of the framework for fault diagnosis and prediction of nuclear power valves is demonstrated, enabling online monitoring and maintenance of important equipment without shutting down the nuclear station.