Pub Date : 2024-06-01DOI: 10.1088/1742-6596/2730/1/012037
Shuping Li, Yan wang
A chemical vapor deposition (CVD) method for preparing carbon nanotube fibers was proposed. This method used ethanol/acetone as the carbon source, ferrocene as the catalyst and thiophene as the promoter. It used a high-temperature gas-phase flow reaction to spin out carbon tube fibers. Subsequently, the mechanical properties of carbon tube fibers prepared under different conditions (carbon source, gas flow and raw material ratio) were characterized and analyzed. The results demonstrated that the carbon fiber prepared with ethanol as the carbon source under specific synthesis conditions has good continuity and can be prepared for 3 hours in succession. In a feasible condition (Fe, S, C ratio, air flow ratio, and growth temperature), using pure acetone as the carbon source can prepare carbon nanotube fibers with fiber-specific strength > 1.5N / tex.
{"title":"Characterization of mechanical properties of carbon nanotube fibers prepared by chemical vapor deposition","authors":"Shuping Li, Yan wang","doi":"10.1088/1742-6596/2730/1/012037","DOIUrl":"https://doi.org/10.1088/1742-6596/2730/1/012037","url":null,"abstract":"A chemical vapor deposition (CVD) method for preparing carbon nanotube fibers was proposed. This method used ethanol/acetone as the carbon source, ferrocene as the catalyst and thiophene as the promoter. It used a high-temperature gas-phase flow reaction to spin out carbon tube fibers. Subsequently, the mechanical properties of carbon tube fibers prepared under different conditions (carbon source, gas flow and raw material ratio) were characterized and analyzed. The results demonstrated that the carbon fiber prepared with ethanol as the carbon source under specific synthesis conditions has good continuity and can be prepared for 3 hours in succession. In a feasible condition (Fe, S, C ratio, air flow ratio, and growth temperature), using pure acetone as the carbon source can prepare carbon nanotube fibers with fiber-specific strength > 1.5N / tex.","PeriodicalId":16821,"journal":{"name":"Journal of Physics: Conference Series","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141407137","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 : 2024-06-01DOI: 10.1088/1742-6596/2767/4/042020
E. Yazici, G. Jacobs, J. Röder, M. Zweiffel
This paper investigates a method for a novel accelerated test procedure for rolling bearings in wind turbine gearboxes. Established test procedures e. g. highly accelerated lifetime tests (HALT) mainly reflect fatigue damage and not slip-induced damage patterns (e.g. smearing). A commonly used criterion to rate the severity of an operating point regarding slip-induced damage is the frictional energy. Frictional energy is introduced into the rolling contact as soon as there is a significant combination of simultaneously occurring slip and pressure. Up to now only critical thresholds for this criterion that must not be exceeded have been identified on small-scale component test rigs. However, the permissible amount and duration of overshoots of the critical threshold that lead to damage in the actual application are not understood yet. Therefore, the aim is to conduct tests on full-size test rigs in which frictional energy is applied until slip-induced damage occurs. In order to perform these tests in a reasonable time, it is essential to accelerate the test procedures. Thus, this paper introduces a method for accelerated test procedures based on frictional energy in the rolling contact. The requirement is that the same cumulated frictional energy as in field operation is applied in a shorter time on the test rig. A further requirement is that the frictional power in the accelerated test procedure never exceeds the maximum frictional power occurring in the field. This paper shows the theoretical background regarding frictional energy and the transfer to the test procedure.
{"title":"Method for accelerated testing of wind turbine gearbox bearings based on frictional energy in the rolling contact","authors":"E. Yazici, G. Jacobs, J. Röder, M. Zweiffel","doi":"10.1088/1742-6596/2767/4/042020","DOIUrl":"https://doi.org/10.1088/1742-6596/2767/4/042020","url":null,"abstract":"This paper investigates a method for a novel accelerated test procedure for rolling bearings in wind turbine gearboxes. Established test procedures e. g. highly accelerated lifetime tests (HALT) mainly reflect fatigue damage and not slip-induced damage patterns (e.g. smearing). A commonly used criterion to rate the severity of an operating point regarding slip-induced damage is the frictional energy. Frictional energy is introduced into the rolling contact as soon as there is a significant combination of simultaneously occurring slip and pressure. Up to now only critical thresholds for this criterion that must not be exceeded have been identified on small-scale component test rigs. However, the permissible amount and duration of overshoots of the critical threshold that lead to damage in the actual application are not understood yet. Therefore, the aim is to conduct tests on full-size test rigs in which frictional energy is applied until slip-induced damage occurs. In order to perform these tests in a reasonable time, it is essential to accelerate the test procedures. Thus, this paper introduces a method for accelerated test procedures based on frictional energy in the rolling contact. The requirement is that the same cumulated frictional energy as in field operation is applied in a shorter time on the test rig. A further requirement is that the frictional power in the accelerated test procedure never exceeds the maximum frictional power occurring in the field. This paper shows the theoretical background regarding frictional energy and the transfer to the test procedure.","PeriodicalId":16821,"journal":{"name":"Journal of Physics: Conference Series","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141409193","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}
With the rapid development of Electric Power Internet of Things (IoT) technology, a large number of devices are being networked and exposed to cyberspace. Due to the disparity in security design levels and lax management during usage, electric power terminals are susceptible targets for network attackers. This not only causes losses to device owners but also poses a threat to the overall cybersecurity of the network, as compromised devices can serve as nodes for botnets. The importance of addressing this issue cannot be underestimated. Asset identification is a prerequisite for the secure management of IoT devices. This paper analyzes and studies existing asset identification technologies. Existing device identification methods based on message content features have problems such as reliance on textual features of message content and difficulties in labeling large-scale data. To overcome these limitations, a machine learning-based classification method for IoT devices is proposed. This method extracts features from the web homepage of devices and generates feature vector fingerprints. By leveraging the random forest algorithm, the accuracy of IoT device classification is improved. This approach is suitable for asset identification of IoT devices and provides support for the precise implementation of vulnerability scanning for IoT devices.
{"title":"Research on machine learning-based device fingerprint recognition technology for power internet of things","authors":"Fangfang Dang, Lijing Yan, Ying Yang, Shuai Li, Dingding Li, Dong Niu","doi":"10.1088/1742-6596/2781/1/012024","DOIUrl":"https://doi.org/10.1088/1742-6596/2781/1/012024","url":null,"abstract":"With the rapid development of Electric Power Internet of Things (IoT) technology, a large number of devices are being networked and exposed to cyberspace. Due to the disparity in security design levels and lax management during usage, electric power terminals are susceptible targets for network attackers. This not only causes losses to device owners but also poses a threat to the overall cybersecurity of the network, as compromised devices can serve as nodes for botnets. The importance of addressing this issue cannot be underestimated. Asset identification is a prerequisite for the secure management of IoT devices. This paper analyzes and studies existing asset identification technologies. Existing device identification methods based on message content features have problems such as reliance on textual features of message content and difficulties in labeling large-scale data. To overcome these limitations, a machine learning-based classification method for IoT devices is proposed. This method extracts features from the web homepage of devices and generates feature vector fingerprints. By leveraging the random forest algorithm, the accuracy of IoT device classification is improved. This approach is suitable for asset identification of IoT devices and provides support for the precise implementation of vulnerability scanning for IoT devices.","PeriodicalId":16821,"journal":{"name":"Journal of Physics: Conference Series","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141392060","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 : 2024-06-01DOI: 10.1088/1742-6596/2767/9/092088
Maria Sarcos, Julian Quick, A. Hahmann, Nicolas G. Alonso-De-Linaje, Neil Davis, M. Friis-Møller
The Wind in my Backyard (WIMBY) project is developing a web interface to aid communities in siting wind energy projects. As part of this siting tool, users will be able to find realistic wind farm layouts for any proposed site in Europe, given certain constraints. When designing this tool, there arises a need for speed: realistic layouts must be designed in computational times that are appropriate for a web interface. In this study, we compare two optimization algorithms: a gradient-based algorithm, referred to as stochastic gradient descent (SGD), and a gradient-free method, referred to as smart-start. The trade-offs between the optimal energy yield and optimization computational time are characterized via a parameter sweep, considering a site in Denmark. This analysis considered farms with 10, 25, and 50 turbines. We find that smart-start yielded the best results for very short computational times, and that SGD yielded layouts with higher energy yields when considering larger computational times.
{"title":"Need For Speed: Fast Wind Farm Optimization","authors":"Maria Sarcos, Julian Quick, A. Hahmann, Nicolas G. Alonso-De-Linaje, Neil Davis, M. Friis-Møller","doi":"10.1088/1742-6596/2767/9/092088","DOIUrl":"https://doi.org/10.1088/1742-6596/2767/9/092088","url":null,"abstract":"The Wind in my Backyard (WIMBY) project is developing a web interface to aid communities in siting wind energy projects. As part of this siting tool, users will be able to find realistic wind farm layouts for any proposed site in Europe, given certain constraints. When designing this tool, there arises a need for speed: realistic layouts must be designed in computational times that are appropriate for a web interface. In this study, we compare two optimization algorithms: a gradient-based algorithm, referred to as stochastic gradient descent (SGD), and a gradient-free method, referred to as smart-start. The trade-offs between the optimal energy yield and optimization computational time are characterized via a parameter sweep, considering a site in Denmark. This analysis considered farms with 10, 25, and 50 turbines. We find that smart-start yielded the best results for very short computational times, and that SGD yielded layouts with higher energy yields when considering larger computational times.","PeriodicalId":16821,"journal":{"name":"Journal of Physics: Conference Series","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141397123","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 : 2024-06-01DOI: 10.1088/1742-6596/2767/7/072014
Filippo Trevisi, Alessandro Croce
Windplanes (i.e. Fly-Gen airborne wind energy systems) harvest wind power via the turbines placed on the tethered wing, which flies crosswind trajectories. In this paper, the optimal design of windplanes is investigated with simplified models, enabling an intuitive understanding of their physical characteristics. The windplane is then idealized as a point mass flying circular fully crosswind trajectories. If the gravity is neglected, the dynamic problem is axial symmetric and the solution is steady. The generated power can be expressed in non-dimensional form by normalizing it with the wind power passing by a disk with radius the wingspan. Since the reference area is taken to be a function of just the wingspan, looking for the design which maximizes this power coefficient addresses the question ”Given a wingspan, which design maximizes power?”. This is different from the literature, where the design problem is formulated per wing area and not per wingspan. The optimal designs have a finite aspect ratio and operate at the maximum lift-to-drag ratio of the airfoil. Airfoils maximizing the lift-to-drag ratio are then optimal for windplanes. If gravity is included in the model, gravitational potential energy is being exchanged over one revolution. Since this exchange comes with an associated efficiency, the plane mass and the related trajectory radius are designed to reduce the potential energy fluctuating over the loop. However, for decreasing turning radii, the available wind power decreases because the windplane sweeps a lower area. For these two conflicting reasons, the optimal mass is finite.
{"title":"Given a wingspan, which windplane design maximizes power?","authors":"Filippo Trevisi, Alessandro Croce","doi":"10.1088/1742-6596/2767/7/072014","DOIUrl":"https://doi.org/10.1088/1742-6596/2767/7/072014","url":null,"abstract":"Windplanes (i.e. Fly-Gen airborne wind energy systems) harvest wind power via the turbines placed on the tethered wing, which flies crosswind trajectories. In this paper, the optimal design of windplanes is investigated with simplified models, enabling an intuitive understanding of their physical characteristics. The windplane is then idealized as a point mass flying circular fully crosswind trajectories. If the gravity is neglected, the dynamic problem is axial symmetric and the solution is steady. The generated power can be expressed in non-dimensional form by normalizing it with the wind power passing by a disk with radius the wingspan. Since the reference area is taken to be a function of just the wingspan, looking for the design which maximizes this power coefficient addresses the question ”Given a wingspan, which design maximizes power?”. This is different from the literature, where the design problem is formulated per wing area and not per wingspan. The optimal designs have a finite aspect ratio and operate at the maximum lift-to-drag ratio of the airfoil. Airfoils maximizing the lift-to-drag ratio are then optimal for windplanes. If gravity is included in the model, gravitational potential energy is being exchanged over one revolution. Since this exchange comes with an associated efficiency, the plane mass and the related trajectory radius are designed to reduce the potential energy fluctuating over the loop. However, for decreasing turning radii, the available wind power decreases because the windplane sweeps a lower area. For these two conflicting reasons, the optimal mass is finite.","PeriodicalId":16821,"journal":{"name":"Journal of Physics: Conference Series","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141403666","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 : 2024-06-01DOI: 10.1088/1742-6596/2730/1/012017
Bin Wang, Jinfeng Lv, Yu Ning, Zhongbin Tang, Cunxian Wang
This study initially conducted bird strike tests and numerical simulations on hollow plates with the Warren structure. The plates manifested notable plastic deformation devoid of fracture. The simulated deformations and strain data of the plates closely corresponded with the outcomes of bird strike impact tests, thereby validating the employed bird strike simulation methodology in this study. Subsequently, while maintaining the outer dimensions of the hollow plate constant, an investigation into the structural parameters of the hollow plate was conducted. Correlations between the thickness of the outer panel, thickness of the Warren, angle of the Warren, and the anti-bird-strike performance of the hollow plate structure were established. The research findings offer valuable insights for informing the design of bird strike resistance in engine blades and hollow plates.
{"title":"Research on bird strike resistance performance of hollow plates with warren structure","authors":"Bin Wang, Jinfeng Lv, Yu Ning, Zhongbin Tang, Cunxian Wang","doi":"10.1088/1742-6596/2730/1/012017","DOIUrl":"https://doi.org/10.1088/1742-6596/2730/1/012017","url":null,"abstract":"This study initially conducted bird strike tests and numerical simulations on hollow plates with the Warren structure. The plates manifested notable plastic deformation devoid of fracture. The simulated deformations and strain data of the plates closely corresponded with the outcomes of bird strike impact tests, thereby validating the employed bird strike simulation methodology in this study. Subsequently, while maintaining the outer dimensions of the hollow plate constant, an investigation into the structural parameters of the hollow plate was conducted. Correlations between the thickness of the outer panel, thickness of the Warren, angle of the Warren, and the anti-bird-strike performance of the hollow plate structure were established. The research findings offer valuable insights for informing the design of bird strike resistance in engine blades and hollow plates.","PeriodicalId":16821,"journal":{"name":"Journal of Physics: Conference Series","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141407282","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 : 2024-06-01DOI: 10.1088/1742-6596/2767/9/092021
P. Maheshwari, Julien Haize, M. Pallud
We apply engineering approaches for estimating the induction zone in front of a wind turbine and escalating to the global blockage at wind farm and cluster of farms scale as well accounting for wind farm wake. The methods include the vortex cylinder model and the self-similar model for blockage, and Jensen and Eddy-viscosity models for the wake estimation. To calibrate each model, Reynold Average Navier-Stokes with Actuator disc simulations are employed. Every model’s efficacy is assessed in comparison to the validation data of an operated wind farm in Scotland with more than 100 turbines of 10 MW each. We also examine the widely used process within the wind industry for transferring the application of calibrated wake and blockage models from one wind farm. According to our simulation analysis, blockage causes sideways acceleration and upstream slowdown. This also result in a gradient in the power produced by each turbine on the farm, which, if not properly accounted for, appears to be a wake model constraint. Furthermore, the response of different turbine type in another environment is not effectively predicted by the blockage and wake model, which is based on a particular turbine type and environment.
{"title":"Modeling of Blockage and Wake Effect: Comparison with Field data","authors":"P. Maheshwari, Julien Haize, M. Pallud","doi":"10.1088/1742-6596/2767/9/092021","DOIUrl":"https://doi.org/10.1088/1742-6596/2767/9/092021","url":null,"abstract":"We apply engineering approaches for estimating the induction zone in front of a wind turbine and escalating to the global blockage at wind farm and cluster of farms scale as well accounting for wind farm wake. The methods include the vortex cylinder model and the self-similar model for blockage, and Jensen and Eddy-viscosity models for the wake estimation. To calibrate each model, Reynold Average Navier-Stokes with Actuator disc simulations are employed. Every model’s efficacy is assessed in comparison to the validation data of an operated wind farm in Scotland with more than 100 turbines of 10 MW each. We also examine the widely used process within the wind industry for transferring the application of calibrated wake and blockage models from one wind farm. According to our simulation analysis, blockage causes sideways acceleration and upstream slowdown. This also result in a gradient in the power produced by each turbine on the farm, which, if not properly accounted for, appears to be a wake model constraint. Furthermore, the response of different turbine type in another environment is not effectively predicted by the blockage and wake model, which is based on a particular turbine type and environment.","PeriodicalId":16821,"journal":{"name":"Journal of Physics: Conference Series","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141399701","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 : 2024-06-01DOI: 10.1088/1742-6596/2767/6/062040
J. E. M. Da Silva, G. C. Fraga, A. P. Petry
In the present work, the CFD modeling of neutral and stable atmospheric boundary layers regarding the horizontal homogeneity condition is discussed to assess the impact of local influences in wind measurements made by a Light Detection and Ranging (LiDAR) remote sensor installed at an offshore site. The LiDAR will be mounted on a real floating, production, storage, and offloading (FPSO) vessel, 200 km far from the coast, and the simulations serve as preliminary study for the impact of the geometry presence, in the air flow and in the measurements taken by the LiDAR. The reproduction of the homogeneity condition is discussed and the impacts of the presence of the ship structure are quantified considering the scanning scheme of the LiDAR. Turbulence is modeled with the k-ε and k-ω models for the neutral case, and k-ε with a modification following the Monin-Obukhov Similarity Theory for the stable boundary laye. By carrying out simulations with empty and blocked (i.e., with the FPSO installed) domains, we show that the effect of the platform is local and only significant at small heights. The neutral and stable cases show similar deviations (≤ 2 %) between the velocity fields in the blocked and empty domains. The neutral case shows locally a little more impact than the stable case, and the reason for that is discussed. We also found that the LiDAR scanning scheme could attenuate the impacts of the flow distortion in comparison with direct punctual measurements.
{"title":"CFD analysis of local influences in offshore wind measurements employing a floating LiDAR system","authors":"J. E. M. Da Silva, G. C. Fraga, A. P. Petry","doi":"10.1088/1742-6596/2767/6/062040","DOIUrl":"https://doi.org/10.1088/1742-6596/2767/6/062040","url":null,"abstract":"In the present work, the CFD modeling of neutral and stable atmospheric boundary layers regarding the horizontal homogeneity condition is discussed to assess the impact of local influences in wind measurements made by a Light Detection and Ranging (LiDAR) remote sensor installed at an offshore site. The LiDAR will be mounted on a real floating, production, storage, and offloading (FPSO) vessel, 200 km far from the coast, and the simulations serve as preliminary study for the impact of the geometry presence, in the air flow and in the measurements taken by the LiDAR. The reproduction of the homogeneity condition is discussed and the impacts of the presence of the ship structure are quantified considering the scanning scheme of the LiDAR. Turbulence is modeled with the k-ε and k-ω models for the neutral case, and k-ε with a modification following the Monin-Obukhov Similarity Theory for the stable boundary laye. By carrying out simulations with empty and blocked (i.e., with the FPSO installed) domains, we show that the effect of the platform is local and only significant at small heights. The neutral and stable cases show similar deviations (≤ 2 %) between the velocity fields in the blocked and empty domains. The neutral case shows locally a little more impact than the stable case, and the reason for that is discussed. We also found that the LiDAR scanning scheme could attenuate the impacts of the flow distortion in comparison with direct punctual measurements.","PeriodicalId":16821,"journal":{"name":"Journal of Physics: Conference Series","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141401225","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 : 2024-06-01DOI: 10.1088/1742-6596/2767/8/082018
M. Souza De Alencar, T. Göçmen, N. Cutululis
There are no known polynomial-time algorithms to perform the cable layout optimization (CLO) for the electrical network within a wind power plant (WPP). This means the computational cost for solving the CLO problem grows exponentially with plant size, which often postpones its solution to until after most design decisions are made, thus forgoing some trade-offs that would be beneficial to the plant’s goal. This work presents a method to obtain a fast estimate of the CLO that can be performed at each iteration of the broader optimization framework. The proposed surrogate model comprises of a graph neural network (GNN) regression model, as this architecture resembles the graph nature of the CLO problem. The GNN is trained with a dataset of procedurally generated site instances that are optimized by costly solving an integer linear programming model. While the inference time for GNNs is constant, the features calculation proposed has time complexity O(N log N) (where N is the number of wind turbines). This still means a major speed up for problems of non-trivial size when compared to exact CLO. A simpler feed-forward neural network (FNN) was trained on the same dataset and used as baseline. Both GNN and FNN achieved r 2 scores of 0.997 for the regression on unseen data of actual WPP, with standard deviations of the relative errors of 1.59% for the FNN and 1.66% for the GNN. Although the GNN did not improve on the performance of the FNN, the latter is an original contribution to the state of the art and an useful tool for the integrated optimization of WPP. This work looked only into a sliver of what is possible with GNNs, leaving ample space for improvements in applying that architecture to the CLO problem.
{"title":"GNN-based surrogate modeling for collection systems costs","authors":"M. Souza De Alencar, T. Göçmen, N. Cutululis","doi":"10.1088/1742-6596/2767/8/082018","DOIUrl":"https://doi.org/10.1088/1742-6596/2767/8/082018","url":null,"abstract":"There are no known polynomial-time algorithms to perform the cable layout optimization (CLO) for the electrical network within a wind power plant (WPP). This means the computational cost for solving the CLO problem grows exponentially with plant size, which often postpones its solution to until after most design decisions are made, thus forgoing some trade-offs that would be beneficial to the plant’s goal. This work presents a method to obtain a fast estimate of the CLO that can be performed at each iteration of the broader optimization framework. The proposed surrogate model comprises of a graph neural network (GNN) regression model, as this architecture resembles the graph nature of the CLO problem. The GNN is trained with a dataset of procedurally generated site instances that are optimized by costly solving an integer linear programming model. While the inference time for GNNs is constant, the features calculation proposed has time complexity O(N log N) (where N is the number of wind turbines). This still means a major speed up for problems of non-trivial size when compared to exact CLO. A simpler feed-forward neural network (FNN) was trained on the same dataset and used as baseline. Both GNN and FNN achieved r 2 scores of 0.997 for the regression on unseen data of actual WPP, with standard deviations of the relative errors of 1.59% for the FNN and 1.66% for the GNN. Although the GNN did not improve on the performance of the FNN, the latter is an original contribution to the state of the art and an useful tool for the integrated optimization of WPP. This work looked only into a sliver of what is possible with GNNs, leaving ample space for improvements in applying that architecture to the CLO problem.","PeriodicalId":16821,"journal":{"name":"Journal of Physics: Conference Series","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141412563","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}