Pub Date : 2022-08-24DOI: 10.1109/IAI55780.2022.9976863
Mingwei Jia, Danya Xu, Tao Yang, Y. Yao, Yi Liu
Most data-driven soft sensor methods can model nonlinear time-varying characteristics of biochemical processes. However, the intrinsic relationship between variables, which is helpful for understanding model behavior, has rarely been investigated in existing data-driven methods. In this work, a novel soft sensor model of time-attention graph convolutional network (TA-GCN) is proposed, which jointly leverages variable relationships and long-term temporal dependencies to improve interpretability and prediction accuracy. This model first uses the maximum information coefficient to construct a topology graph and trains edge strengths end-to-end. The data are then encoded in the spatial-temporal dimension based on GCN and attention mechanism. Finally, the empirical knowledge that analyzes the operating state of the process and graph are combined to explain the model behavior. In comparison to existing soft sensors, TA-GCN enables efficient and scalable training for long-term spatial-temporal dependencies. Experimental results on InPenSim dataset demonstrate that TA-GCN is competitive with state-of-the-art methods.
{"title":"Time-Attention Graph Convolutional Network Soft Sensor in Biochemical Processes","authors":"Mingwei Jia, Danya Xu, Tao Yang, Y. Yao, Yi Liu","doi":"10.1109/IAI55780.2022.9976863","DOIUrl":"https://doi.org/10.1109/IAI55780.2022.9976863","url":null,"abstract":"Most data-driven soft sensor methods can model nonlinear time-varying characteristics of biochemical processes. However, the intrinsic relationship between variables, which is helpful for understanding model behavior, has rarely been investigated in existing data-driven methods. In this work, a novel soft sensor model of time-attention graph convolutional network (TA-GCN) is proposed, which jointly leverages variable relationships and long-term temporal dependencies to improve interpretability and prediction accuracy. This model first uses the maximum information coefficient to construct a topology graph and trains edge strengths end-to-end. The data are then encoded in the spatial-temporal dimension based on GCN and attention mechanism. Finally, the empirical knowledge that analyzes the operating state of the process and graph are combined to explain the model behavior. In comparison to existing soft sensors, TA-GCN enables efficient and scalable training for long-term spatial-temporal dependencies. Experimental results on InPenSim dataset demonstrate that TA-GCN is competitive with state-of-the-art methods.","PeriodicalId":138951,"journal":{"name":"2022 4th International Conference on Industrial Artificial Intelligence (IAI)","volume":"3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115482785","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-08-24DOI: 10.1109/IAI55780.2022.9976505
Di An, Michael Difrieri, Yangquan Chen
The foundation of soil carbon management is the measurement of soil carbon content, which potentially enables many carbon-negative or carbon-neutral technologies for fighting climate change and improving soil health for greater crop yield. Several researchers used a non-intrusive method to quantify soil organic carbon content using ground penetrating radar (GPR) with a fixed sensor configuration. The sensor we used in this study, however, is compactly comprised of an array of 18 radar transmitter (TX) and receiver (RX) pairs. It is necessary to propose an assessment of sensing performance which can avoid possible failure in identifying the correct soil carbon spatial-temporal changes. In this paper, we provide a comprehensive assessment of the evaluation of non-intrusive methods for sensing soil carbon content when a radar array is used. Specifically, our proposed evaluation score utilizes explicit physical knowledge as a data-driven metric to find the optimal antenna pair combination for our radar array sensor under different sensing tasks and environments. We evaluated our soil carbon sensing score (SCSS) using the data collected from real-world soil sample experiments. The results show that the optimal antenna pair has the greatest sensing ability to measure soil carbon content in a variety of sensing environments and sensing distances, with a 36% increase in classification accuracy.
{"title":"Optimal Antenna Pairing of A Miniaturized Radar Array for Smart Sensing of Soil Carbon Content","authors":"Di An, Michael Difrieri, Yangquan Chen","doi":"10.1109/IAI55780.2022.9976505","DOIUrl":"https://doi.org/10.1109/IAI55780.2022.9976505","url":null,"abstract":"The foundation of soil carbon management is the measurement of soil carbon content, which potentially enables many carbon-negative or carbon-neutral technologies for fighting climate change and improving soil health for greater crop yield. Several researchers used a non-intrusive method to quantify soil organic carbon content using ground penetrating radar (GPR) with a fixed sensor configuration. The sensor we used in this study, however, is compactly comprised of an array of 18 radar transmitter (TX) and receiver (RX) pairs. It is necessary to propose an assessment of sensing performance which can avoid possible failure in identifying the correct soil carbon spatial-temporal changes. In this paper, we provide a comprehensive assessment of the evaluation of non-intrusive methods for sensing soil carbon content when a radar array is used. Specifically, our proposed evaluation score utilizes explicit physical knowledge as a data-driven metric to find the optimal antenna pair combination for our radar array sensor under different sensing tasks and environments. We evaluated our soil carbon sensing score (SCSS) using the data collected from real-world soil sample experiments. The results show that the optimal antenna pair has the greatest sensing ability to measure soil carbon content in a variety of sensing environments and sensing distances, with a 36% increase in classification accuracy.","PeriodicalId":138951,"journal":{"name":"2022 4th International Conference on Industrial Artificial Intelligence (IAI)","volume":"200 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115720815","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-08-24DOI: 10.1109/IAI55780.2022.9976773
Yang Ji
This paper mainly focuses on the problem of target tracking trajectory generation for quadrotors in static complex environments. In this paper, a trajectory planning framework is proposed that can be used to address the occlusion problem caused by obstacles in target tracking situations. The main innovations are: First, a visibility function is introduced that is constructed based on a signed Euclidean distance field (ESDF) map. Based on this function, the favorable view direction is derived to track the target distinctly. Second, the distance between the target and chaser is optimized as one of the cost functions, resulting in a smooth and comfortable distance between them. Finally, the effectiveness of the algorithm is verified on the Rviz platform.
{"title":"Target tracking trajectory generation for quadrotors in static complex environments","authors":"Yang Ji","doi":"10.1109/IAI55780.2022.9976773","DOIUrl":"https://doi.org/10.1109/IAI55780.2022.9976773","url":null,"abstract":"This paper mainly focuses on the problem of target tracking trajectory generation for quadrotors in static complex environments. In this paper, a trajectory planning framework is proposed that can be used to address the occlusion problem caused by obstacles in target tracking situations. The main innovations are: First, a visibility function is introduced that is constructed based on a signed Euclidean distance field (ESDF) map. Based on this function, the favorable view direction is derived to track the target distinctly. Second, the distance between the target and chaser is optimized as one of the cost functions, resulting in a smooth and comfortable distance between them. Finally, the effectiveness of the algorithm is verified on the Rviz platform.","PeriodicalId":138951,"journal":{"name":"2022 4th International Conference on Industrial Artificial Intelligence (IAI)","volume":"37 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124994243","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-08-24DOI: 10.1109/IAI55780.2022.9976524
Yuhong Ren, Jiafu Tang
We present the flexibility of a human-centered production system called the seru production system (SPS). A theoretical framework for analyzing the flexibility of an SPS is proposed based on the input-process-output (IPO) system view. The enabling effect of workforce configuration on the flexibility of an SPS is explained. The flexibility of SPS is identified to be the capability of an SPS to have inclusiveness and variability. The inclusiveness shows the capability of an SPS to control-variability-with-stability, and variability presents its ability to control-variability-with-variability, which correspond to structural flexibility (SF) and reorganization flexibility (RF), respectively. We reveal that the SPS adopts SF as the main strategy to satisfy most demands and uses RF as an auxiliary means to capture unforeseen demands. In addition, our work reports the strategies for implementing SPS flexibility including structural flexibility strategy, reorganization flexibility strategy, and hybrid flexibility strategy.
{"title":"Flexibility of Seru Production System: An Input-Process-Output System View*","authors":"Yuhong Ren, Jiafu Tang","doi":"10.1109/IAI55780.2022.9976524","DOIUrl":"https://doi.org/10.1109/IAI55780.2022.9976524","url":null,"abstract":"We present the flexibility of a human-centered production system called the seru production system (SPS). A theoretical framework for analyzing the flexibility of an SPS is proposed based on the input-process-output (IPO) system view. The enabling effect of workforce configuration on the flexibility of an SPS is explained. The flexibility of SPS is identified to be the capability of an SPS to have inclusiveness and variability. The inclusiveness shows the capability of an SPS to control-variability-with-stability, and variability presents its ability to control-variability-with-variability, which correspond to structural flexibility (SF) and reorganization flexibility (RF), respectively. We reveal that the SPS adopts SF as the main strategy to satisfy most demands and uses RF as an auxiliary means to capture unforeseen demands. In addition, our work reports the strategies for implementing SPS flexibility including structural flexibility strategy, reorganization flexibility strategy, and hybrid flexibility strategy.","PeriodicalId":138951,"journal":{"name":"2022 4th International Conference on Industrial Artificial Intelligence (IAI)","volume":"28 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116447399","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In ultra-narrow gap welding, it is necessary to detect the aspect ratio parameters of arc shape in real time and efficiently. However, the existing arc shape method can not realize on-line detection. To solve this problem, this paper proposes a lightweight arc detection network AD-YOLOV5 based on YOLOv5s network model. To reduce the complexity of YOLOv5s network, the Repvgg Block module is used to replace the CONV module in Backbone network, and the coordinate attention mechanism is introduced in Neck network to guarantee the lightness of YOLOv5s network and improve the precision of model. The experimental results show that the model size is reduced by 65% and the detection speed is increased by 50% while the detection accuracy of aspect ratio remains unchanged. The implementation of the method in this paper provides a reference for the online monitoring of ultranarrow gap welding quality.
{"title":"Detection of arc shape in ultra-narrow gap welding based on improved YOLOv5s","authors":"Weilong He, Ping Wang, A. Zhang, Jing Ma, Shengming Ma, Yanpeng Feng","doi":"10.1109/IAI55780.2022.9976556","DOIUrl":"https://doi.org/10.1109/IAI55780.2022.9976556","url":null,"abstract":"In ultra-narrow gap welding, it is necessary to detect the aspect ratio parameters of arc shape in real time and efficiently. However, the existing arc shape method can not realize on-line detection. To solve this problem, this paper proposes a lightweight arc detection network AD-YOLOV5 based on YOLOv5s network model. To reduce the complexity of YOLOv5s network, the Repvgg Block module is used to replace the CONV module in Backbone network, and the coordinate attention mechanism is introduced in Neck network to guarantee the lightness of YOLOv5s network and improve the precision of model. The experimental results show that the model size is reduced by 65% and the detection speed is increased by 50% while the detection accuracy of aspect ratio remains unchanged. The implementation of the method in this paper provides a reference for the online monitoring of ultranarrow gap welding quality.","PeriodicalId":138951,"journal":{"name":"2022 4th International Conference on Industrial Artificial Intelligence (IAI)","volume":"196 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122341311","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-08-24DOI: 10.1109/IAI55780.2022.9976653
Linsong Yuan, Shenwei Chen, Guanglun Liu
Due to variational environmental conditions and varied adaptive control strategies, the operation states of wind turbines are continuously changing, leading to diverse types of samples in the power curve. Different kinds of samples may contain noises or valuable information for specific downstream tasks and thus need to be correctly identified and labeled. To this end, this paper proposes a spatial-temporal pattern recognition algorithm for data identification and tagging. According to spatial distribution and temporal characteristics, all data points are divided into four groups including normal samples, isolated outliers, change points, and faulty samples. Then, some distances based on the dynamic time warping method are defined to make evaluations and then serve as indicators for achieving precise tagging of each category. Case studies and comparative experiments are conducted to verify the effectiveness and superiority of the proposed method.
{"title":"Spatial-temporal Pattern Recognition for Data Identification and Tagging Based on Power Curve in Wind Turbines","authors":"Linsong Yuan, Shenwei Chen, Guanglun Liu","doi":"10.1109/IAI55780.2022.9976653","DOIUrl":"https://doi.org/10.1109/IAI55780.2022.9976653","url":null,"abstract":"Due to variational environmental conditions and varied adaptive control strategies, the operation states of wind turbines are continuously changing, leading to diverse types of samples in the power curve. Different kinds of samples may contain noises or valuable information for specific downstream tasks and thus need to be correctly identified and labeled. To this end, this paper proposes a spatial-temporal pattern recognition algorithm for data identification and tagging. According to spatial distribution and temporal characteristics, all data points are divided into four groups including normal samples, isolated outliers, change points, and faulty samples. Then, some distances based on the dynamic time warping method are defined to make evaluations and then serve as indicators for achieving precise tagging of each category. Case studies and comparative experiments are conducted to verify the effectiveness and superiority of the proposed method.","PeriodicalId":138951,"journal":{"name":"2022 4th International Conference on Industrial Artificial Intelligence (IAI)","volume":"134 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122620983","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-08-24DOI: 10.1109/IAI55780.2022.9976875
Tianwei Dong, C. Zang, Peng Zeng
Elevators play a vital role in people's daily activities as a vehicle. Once the elevator runs in the process, failure will seriously threaten the user's life and property safety, so the corresponding fault diagnosis of the elevator is necessary for the elevator maintenance process. In this paper, the wavelet soft threshold denoising method is used to reduce the influence of external interference on the diagnosis results, and the time domain features of signals are extracted to form the feature vector. The stochastic configuration network is used to classify the feature vector and establish the elevator fault diagnosis model. Finally, the feasibility of the method is verified by experimental comparison. The final experiment shows that this method has good stability and a high fault recognition rate, which is very important for elevator maintenance.
{"title":"Fault diagnosis study of elevator based on stochastic configuration networks","authors":"Tianwei Dong, C. Zang, Peng Zeng","doi":"10.1109/IAI55780.2022.9976875","DOIUrl":"https://doi.org/10.1109/IAI55780.2022.9976875","url":null,"abstract":"Elevators play a vital role in people's daily activities as a vehicle. Once the elevator runs in the process, failure will seriously threaten the user's life and property safety, so the corresponding fault diagnosis of the elevator is necessary for the elevator maintenance process. In this paper, the wavelet soft threshold denoising method is used to reduce the influence of external interference on the diagnosis results, and the time domain features of signals are extracted to form the feature vector. The stochastic configuration network is used to classify the feature vector and establish the elevator fault diagnosis model. Finally, the feasibility of the method is verified by experimental comparison. The final experiment shows that this method has good stability and a high fault recognition rate, which is very important for elevator maintenance.","PeriodicalId":138951,"journal":{"name":"2022 4th International Conference on Industrial Artificial Intelligence (IAI)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124385383","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-08-24DOI: 10.1109/IAI55780.2022.9976558
Tianqun Ren, Xiang Chen
This paper studies control for a class of vehicle systems. Different from the majority of the existing work, two adaptive control laws are proposed to tackle the feedback stability of the nonlinear vehicle dynamics under both position and velocity controls. The results of the paper show that the proposed adaptive control laws are capable of dealing with the nonlinear dynamics in the presence of unknown vehicle parameters in achieving the velocity and position control. In combination with the classic proportional and derivative control, the proposed adaptive control method mitigates parameter uncertainties and model nonlinearities. The control performance of the vehicle systems is illustrated by simulation studies.
{"title":"Adaptive Control for A Class of Nonholonomic Vehicle Systems","authors":"Tianqun Ren, Xiang Chen","doi":"10.1109/IAI55780.2022.9976558","DOIUrl":"https://doi.org/10.1109/IAI55780.2022.9976558","url":null,"abstract":"This paper studies control for a class of vehicle systems. Different from the majority of the existing work, two adaptive control laws are proposed to tackle the feedback stability of the nonlinear vehicle dynamics under both position and velocity controls. The results of the paper show that the proposed adaptive control laws are capable of dealing with the nonlinear dynamics in the presence of unknown vehicle parameters in achieving the velocity and position control. In combination with the classic proportional and derivative control, the proposed adaptive control method mitigates parameter uncertainties and model nonlinearities. The control performance of the vehicle systems is illustrated by simulation studies.","PeriodicalId":138951,"journal":{"name":"2022 4th International Conference on Industrial Artificial Intelligence (IAI)","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124133563","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-08-24DOI: 10.1109/IAI55780.2022.9976538
Xingguo Wang, Deqing Huang, Huanlai Xing
This paper mainly focuses on the high-speed train cooperative operation problem. To solve this problem, this paper presents a speed curve optimization method based on improved reinforcement learning algorithm. First, according to the train dynamics system, we build the speed curve optimization object. In order to realize the cooperative operation of trains, we use the artificial potential field method to establish the reward function for train spacing. At the same time, to ensure passenger comfort, train jerk rate also needs to be added into the reward function. And then, agent of improved reinforcement learning is established. The improved reinforcement learning algorithm is different from the general reinforcement learning algorithm in that the observation dimension of policy network is manually reduced compared with that of the Q value network to improve the learning speed of the algorithm. At the same time, in order to reduce the agent's attempts to perform useless actions in some states, a reference controller is added to the system to further accelerate the learning process. In addition, training parameters need to be set, such as training termination conditions, maximum number of steps, desired global reward value, and so on. After the training. The Agent can generate a desirable speed curve of train based on constraints of vehicle output and jerk rate under cooperative operation.
{"title":"A Train Cooperative Operation Optimization Method based on Improved Reinforcement Learning Algorithm*","authors":"Xingguo Wang, Deqing Huang, Huanlai Xing","doi":"10.1109/IAI55780.2022.9976538","DOIUrl":"https://doi.org/10.1109/IAI55780.2022.9976538","url":null,"abstract":"This paper mainly focuses on the high-speed train cooperative operation problem. To solve this problem, this paper presents a speed curve optimization method based on improved reinforcement learning algorithm. First, according to the train dynamics system, we build the speed curve optimization object. In order to realize the cooperative operation of trains, we use the artificial potential field method to establish the reward function for train spacing. At the same time, to ensure passenger comfort, train jerk rate also needs to be added into the reward function. And then, agent of improved reinforcement learning is established. The improved reinforcement learning algorithm is different from the general reinforcement learning algorithm in that the observation dimension of policy network is manually reduced compared with that of the Q value network to improve the learning speed of the algorithm. At the same time, in order to reduce the agent's attempts to perform useless actions in some states, a reference controller is added to the system to further accelerate the learning process. In addition, training parameters need to be set, such as training termination conditions, maximum number of steps, desired global reward value, and so on. After the training. The Agent can generate a desirable speed curve of train based on constraints of vehicle output and jerk rate under cooperative operation.","PeriodicalId":138951,"journal":{"name":"2022 4th International Conference on Industrial Artificial Intelligence (IAI)","volume":"35 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125732252","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-08-24DOI: 10.1109/IAI55780.2022.9976577
Mingyu Li, Jingwen Zhang, Tianyou Chai
This paper proposes a multi-step ahead power demand model for fused magnesia smelting processes (FMSP) which combines a linear model and an unknown nonlinear term to predict the electricity demand and its variation tendency for the next 5 steps. The linear model is identified by the multi-output fast recursive algorithm (MFRA) while the unknown nonlinear term is fitted with a long-short term memory (LSTM) model. The hyperparameters in the LSTM are estimated by the Bayesian optimization (BO) algorithm. Since the sampling period of the power is only 7 seconds, and we have to predict the next 5 steps electricity demand and its tendency within one sampling period, we therefore update parameters of the linear model by the MFRA while parameters of the dense layer of the LSTM are updated by the gradient descent algorithm within the online multi-step demand forecasting framework. The experimental results using the real-time data of a FMSP confirm the effectiveness of the proposed algorithm, achieving up to 52% error reduction in 5-step ahead demand forecasting when compared with other approaches.
{"title":"Deep Learning Assisted Online Multi-Step Demand Forecasting of Fused Magnesia Smelting Processes","authors":"Mingyu Li, Jingwen Zhang, Tianyou Chai","doi":"10.1109/IAI55780.2022.9976577","DOIUrl":"https://doi.org/10.1109/IAI55780.2022.9976577","url":null,"abstract":"This paper proposes a multi-step ahead power demand model for fused magnesia smelting processes (FMSP) which combines a linear model and an unknown nonlinear term to predict the electricity demand and its variation tendency for the next 5 steps. The linear model is identified by the multi-output fast recursive algorithm (MFRA) while the unknown nonlinear term is fitted with a long-short term memory (LSTM) model. The hyperparameters in the LSTM are estimated by the Bayesian optimization (BO) algorithm. Since the sampling period of the power is only 7 seconds, and we have to predict the next 5 steps electricity demand and its tendency within one sampling period, we therefore update parameters of the linear model by the MFRA while parameters of the dense layer of the LSTM are updated by the gradient descent algorithm within the online multi-step demand forecasting framework. The experimental results using the real-time data of a FMSP confirm the effectiveness of the proposed algorithm, achieving up to 52% error reduction in 5-step ahead demand forecasting when compared with other approaches.","PeriodicalId":138951,"journal":{"name":"2022 4th International Conference on Industrial Artificial Intelligence (IAI)","volume":"357 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115863075","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：