The refining industry's substantial hydrogen demand relies on high-carbon-emission production methods, facing dual challenges of reducing costs and achieving net-zero emissions. This study proposes a renewable ene...
详细信息
The refining industry's substantial hydrogen demand relies on high-carbon-emission production methods, facing dual challenges of reducing costs and achieving net-zero emissions. This study proposes a renewable energy-powered water electrolysis system integrated with seasonal hydrogen storage to address these challenges. A multi-objective capacity optimization model is developed to minimize costs and carbon emissions, ensuring reliable hydrogen supply. To manage computational complexity, time variational autoencoder is applied to extract patterns from high-dimensional data for scenario generation and typical day selection. A refinery case study validates the system's capacity configuration. Pareto analysis reveals a tradeoff between costs and emissions, necessitating large-scale seasonal hydrogen storage to balance renewable energy fluctuations, while battery storage manages short-term fluctuations. Sensitivity analysis shows exceeding a 0.38 electrolyzer minimum load reduces the economic viability of renewable hydrogen production. The system achieves an LCOH of 2.28 USD/kg and annual carbon emissions of 361,139 tons, offering cost-effective and sustainable hydrogen production.
Fault classification is a common problem in industrial fault diagnosis. Usually, classifiers are built assuming an equal amount of data across different classes. However, the amount of normal data and fault data colle...
详细信息
ISBN:
(纸本)9798350321050
Fault classification is a common problem in industrial fault diagnosis. Usually, classifiers are built assuming an equal amount of data across different classes. However, the amount of normal data and fault data collected from industrial processes is often imbalanced. Intrinsically, the fault classification is also an imbalanced data classification problem. To address this issue, data augmentation methods can be used to effectively generate more data and achieve data balance. While the performance of classification is greatly influenced by the generated data. The quality of the generated data can greatly impact the classification performance. To ensure the stability of the generated data, this paper extends the generation of single data to the generation of time series data using a time variational autoencoder. Using the generated time series data, a new classifier called the time Series Data Augmentation Classifier (TSDAC) is proposed to solve the imbalanced fault classification problem. After that, the TSDAC is applied to Tennessee Eastman (TE) benchmark process. The results show that the TSDAC is recommended for imbalanced fault classification.
暂无评论