Electricity consumption prediction is crucial for the operation, strategic planning, and maintenance of power grid infrastructure. The effective management of power systems depends on accurately predicting electricity...
详细信息
Electricity consumption prediction is crucial for the operation, strategic planning, and maintenance of power grid infrastructure. The effective management of power systems depends on accurately predicting electricity usage patterns and intensity. This study aims to enhance the operational efficiency of power systems and minimize environmental impact by predicting mid to long-term electricity consumption in industrial facilities, particularly in forging processes, and detecting anomalies in energy consumption. We propose an ensemble model combining Extreme Gradient Boosting (XGBoost) and a long short-term memory autoencoder (LSTM-AE) to accurately forecast power consumption. This approach leverages the strengths of both models to improve prediction accuracy and responsiveness. The dataset includes power consumption data from forging processes in manufacturing plants, as well as system load and System Marginal Price data. During data preprocessing, Expectation Maximization Principal Component Analysis was applied to address missing values and select significant features, optimizing the model. The proposed method achieved a Mean Absolute Error of 0.020, a Mean Squared Error of 0.021, a Coefficient of Determination of 0.99, and a Symmetric Mean Absolute Percentage Error of 4.24, highlighting its superior predictive performance and low relative error. These findings underscore the model's reliability and accuracy for integration into Energy Management Systems for real-time data processing and mid to long-term energy planning, facilitating sustainable energy use and informed decision making in industrial settings.
In this work, we introduce MOLA, a multi-block orthogonal long short-term memory autoencoder paradigm, to conduct accurate, reliable fault detection of industrial processes. To achieve this, MOLA effectively extracts ...
详细信息
In this work, we introduce MOLA, a multi-block orthogonal long short-term memory autoencoder paradigm, to conduct accurate, reliable fault detection of industrial processes. To achieve this, MOLA effectively extracts dynamic orthogonal features by introducing an orthogonality-based loss function to constrain the latent space output. This helps eliminate the redundancy in the features identified, thereby improving the overall monitoring performance. On top of this, a multi-block monitoring structure is proposed, which categorizes the process variables into multiple blocks by leveraging expert process knowledge about their associations with the overall process. Each block is associated with its specific orthogonal long short-term memory autoencoder model, whose extracted dynamic orthogonal features are monitored by distance-based Hotelling's T2 statistics and quantile-based cumulative sum (CUSUM) designed for multivariate data streams that are nonparametric and heterogeneous. Compared to having a single model accounting for all process variables, such a multi-block structure significantly improves overall process monitoring performance, especially for large-scale industrial processes. Finally, we propose an adaptive weight-based Bayesian fusion (W-BF) framework to aggregate all block-wise monitoring statistics into a global statistic that we monitor for faults. Fault detection speed and accuracy are improved by assigning and adjusting weights to blocks based on the sequential order in which alarms are raised. We demonstrate the efficiency and effectiveness of our MOLA framework by applying it to the Tennessee Eastman process and comparing the performance with various benchmark methods.
Sodium-cooled fast reactors (SFR), which use high temperature fluid near ambient pressure as coolant, are one of the most promising types of GEN IV reactors. One of the unique challenges of SFR operation is purificati...
详细信息
Sodium-cooled fast reactors (SFR), which use high temperature fluid near ambient pressure as coolant, are one of the most promising types of GEN IV reactors. One of the unique challenges of SFR operation is purification of high temperature liquid sodium with a cold trap to prevent corrosion and obstructing small orifices. We have developed a deep learning longshort-termmemory (LSTM) autoencoder for continuous monitoring of a cold trap and detection of operational anomaly. Transient data were obtained from the Mechanisms Engineering Test Loop (METL) liquid sodium facility at Argonne National Laboratory. The cold trap purification at METL is monitored with 31 variables, which are sensors measuring fluid temperatures, pressures and flow rates, and controller signals. Loss-of-coolant type anomaly in the cold trap operation was generated by temporarily choking one of the blowers, which resulted in temperature and flow rate spikes. The input layer of the autoencoder consisted of all the variables involved in monitoring the cold trap. The LSTM autoencoder was trained on the data corresponding to cold trap startup and normal operation regime, with the loss function calculated as the mean absolute error (MAE). The loss during training was determined to follow log-normal density distribution. During monitoring, we investigated a performance of the LSTM autoencoder for different loss threshold values, set at a progressively increasing number of standard deviations from the mean. The anomaly signal in the data was gradually attenuated, while preserving the noise of the original time series, so that the signal-to-noise ratio (SNR) averaged across all sensors decreased below unity. Results demonstrate detection of anomalies with sensor-averaged SNR < 1.
This paper presents a data-driven approach to short-term wind turbine fault prediction and condition monitoring based on a hybrid architecture of recurrent neural network and longshort-termmemory. The proposed archi...
详细信息
This paper presents a data-driven approach to short-term wind turbine fault prediction and condition monitoring based on a hybrid architecture of recurrent neural network and longshort-termmemory. The proposed architecture is established by utilizing time series data from the supervisory control and data acquisition system and a Bladed model of a 5 MW wind turbine to predict faults occurring to the wind generator. The recurrent neural network-longshort-termmemory training procedure is enhanced with self-organizing maps and longshort-termmemory auto encoder so as to describe the complex interaction between the mechanical system and unpredictable wind speed. To verify the performance of the proposed scheme, we conduct in-depth numerical experiments by applying the hybrid architecture to the Bladed 5 MW wind turbine model with rated wind speed of 11.8 m/s. Experimental results confirm that the proposed scheme has superior accuracy and practicality of fault prediction compared with eminent existing machine learning algorithms such as extreme gradient boost and random forest regressor.
The proposed framework consists of three modules as an outlier detection method for indoor air quality data. We first use a long short-term memory autoencoder (LSTM-AE) based reconstruction error detector, which desig...
详细信息
The proposed framework consists of three modules as an outlier detection method for indoor air quality data. We first use a long short-term memory autoencoder (LSTM-AE) based reconstruction error detector, which designs the LSTM layer in the shape of an autoencoder, to build a reconstruction error-based outlier detection model and extract latent features. The latent feature class-assisted vector machine detector constructs an additional outlier detection model using previously extracted latent features. Finally, the ensemble detector combines the two independent classifiers to define a new ensemble-based decision rule. Furthermore, because real-time anomaly detection proceeds with unsupervised learning, more stable and consistent external detection rules are defined than when using a single ensemble model. Laboratory tests with five random cases were performed for objective evaluation. Thus, we propose a framework that can be applied to various industrial environments by detecting and defining stable outlier decision rules.
Demand response (DR) is an economical way of addressing the challenges faced by the massive penetration of distributed energy resources, such as renewable energy. Residential consumers account for a significant propor...
详细信息
Demand response (DR) is an economical way of addressing the challenges faced by the massive penetration of distributed energy resources, such as renewable energy. Residential consumers account for a significant proportion of electricity consumption. However, their behavior is highly random and uncertain, meaning it is difficult to quantify the impact of DR programs in which they participate. This paper presents a two-level optimal bidding strategy framework for load aggregators that combines a data-driven forecasting model and a data-driven agent-based model (D-ABM) to provide a realistic estimate of the impact of DR. First, the aggregated load of all consumers and market prices are predicted via a longshort-termmemory (LSTM) autoencoder forecasting model. Then, the proposed D-ABM estimates and quantifies the difference rate in terms of total load due to DR. Since D-ABM is a bottom-up approach, each consumer can be treated as a heterogeneous agent and changes in individual electricity usage patterns due to DR can be estimated. Changes in collective electricity consumption patterns can also be quantified by considering the estimated individual behavior and the interactions defined by the basic rules. In addition, assumptions about biases and preferences that explain the irrationality of individual decision-making are given to agents, and the uncertainty of DR participation is considered more realistically. Finally, based on these uncertainties addressed at each level, various bidding strategies for load aggregators can be obtained. The numerical simulation results indicate that our framework provides a more realistic estimation of the impact of total load under DR, minimizes any deviations from bidding strategies, and ensures maximum profits for load aggregators.
Gait analysis plays a crucial role in detecting and monitoring various neurological and musculoskeletal disorders early. This paper presents a comprehensive study of the automatic detection of abnormal gait using 3D v...
详细信息
Gait analysis plays a crucial role in detecting and monitoring various neurological and musculoskeletal disorders early. This paper presents a comprehensive study of the automatic detection of abnormal gait using 3D vision, with a focus on non-invasive and practical data acquisition methods suitable for everyday environments. We explore various configurations, including multi-camera setups placed at different distances and angles, as well as performing daily activities in different directions. An integral component of our study involves combining gait analysis with the monitoring of activities of daily living (ADLs), given the paramount relevance of this integration in the context of Ambient Assisted Living. To achieve this, we investigate cutting-edge Deep Neural Network approaches, such as the Temporal Convolutional Network, Gated Recurrent Unit, and long short-term memory autoencoder. Additionally, we scrutinize different data representation formats, including Euclidean-based representations, angular adjacency matrices, and rotation matrices. Our system's performance evaluation leverages both publicly available datasets and data we collected ourselves while accounting for individual variations and environmental factors. The results underscore the effectiveness of our proposed configurations in accurately classifying abnormal gait, thus shedding light on the optimal setup for non-invasive and efficient data collection.
In online learning, students' learning behavior might change as the course progresses. How students adjust learning behaviors aligned with course requirements reflects their self-regulated learning strategies. Ana...
详细信息
In online learning, students' learning behavior might change as the course progresses. How students adjust learning behaviors aligned with course requirements reflects their self-regulated learning strategies. Analyzing students' learning patterns can help instructors understand how the course design or activities shape students' learning behaviors, including their learning beliefs and motivation, and facilitate teaching decision makings accordingly. This study aims to propose a scientific analytic method to understand students' self-regulated learning (SRL) patterns. The whole process includes the following four steps: (1) encoding behavioral patterns;(2) detecting turning points and chunking behavioral patterns;(3) grouping similar patterns;and (4) interpreting results. A case study with 4604 K-12 students from 476 courses was conducted to validate the proposed method. Five successful patterns, three at-risk patterns, and three average patterns were identified. The case study indicated that successful students showed at least one of the following characteristics: (1) Balanced, (2) Proactive and Balanced, and (3) Balanced with one highly engaged behavior. The at-risk students showed the following characteristics: (1) Oscillatory and (2) Low Engaged. Patterns which led to successful or at-risk conditions are compared and connected with corresponding SRL strategies. Practical and research implications are discussed in the article as well.
In recent years, user behavior anomaly detection has been gaining attention in cybersecurity. A crucial challenge that has been discussed in the literature is that supervised models that use vast amounts of data for t...
详细信息
ISBN:
(纸本)9781665482257
In recent years, user behavior anomaly detection has been gaining attention in cybersecurity. A crucial challenge that has been discussed in the literature is that supervised models that use vast amounts of data for training do not apply to real scenarios for anomaly detection. Within this context, the requirement to gather datasets with labeled behavior anomalies has proven to be a significant limiting factor for evaluating different models. This paper presents WEAPON, an unsupervised learning-based architecture for user behavior anomaly detection that requires a small amount of data for building behavior profiles considering the individuality of each user. WEAPON implements the weak supervision-based behavior anomaly labeling approach using Snorkel. When compared to other approaches, WEAPON proved to be more efficient, surpassing the ROC curve of the second best model by 4.31%. Furthermore, WEAPON outperforms rule-based methods by finding anomalies that an expert would not anticipate.
暂无评论