Search

Deep Neural Network (DNN)-based point cloud semantic segmentation has presented significant breakthrough using large-scale labeled aerial laser point cloud datasets. However, annotating such large-scaled point clouds is time-consuming. Self-Supervised Learning (SSL) is a promising approach to this problem by pre-training a DNN model utilizing unlabeled samples followed by a fine-tuned downstream task involving very limited labels. The traditional contrastive learning for point clouds selects the hardest negative samples by solely relying on the distance between the embedded features derived from the learning process, potentially evolving some negative samples from the same classes to reduce the contrastive learning effectiveness. This work proposes a hard-negative sample-aware self-supervised contrastive learning algorithm to pre-train the model for semantic segmentation. We designed a k-means clustering-based Absolute Positive And Negative samples (AbsPAN) strategy to filter the possible false-negative samples. Experiments on two typical ALS benchmark datasets demonstrate that the proposed method is more appealing than supervised training schemes without pre-training. Especially when the labels are severely inadequate (10% of the ISPRS training set), the results obtained by the proposed HAVANA method still exceed 94% of the supervised paradigm performance with full training set. [ABSTRACT FROM AUTHOR]

The electricity consumption behavior of the inhabitants is a major contributor to the uncertainty of the residential load system. Human-caused uncertainty may have a distributional component, but it is not well understood, which limits further understanding the stochastic component of load forecasting. This study proposes a short-term load-interval forecasting method considering the stochastic features caused by users' electricity consumption behavior. The proposed method is composed of two parts: load-point forecasting using singular spectrum analysis and long short-term memory (SSA-LSTM), and load boundaries forecasting using statistical analysis. Firstly, the load sequence is decomposed and recombined using SSA to obtain regular and stochastic subsequences. Then, the load-point forecasting LSTM network model is trained from the regular subsequence. Subsequently, the load boundaries related to load consumption consistency are forecasted by statistical analysis. Finally, the forecasting results are combined to obtain the load-interval forecasting result. The case study reveals that compared with other common methods, the proposed method can forecast the load interval more accurately and stably based on the load time series. By using the proposed method, the evaluation index coverage rates (CRs) are (17.50%, 1.95%, 1.05%, 0.97%, 7.80%, 4.55%, 9.52%, 1.11%), (17.95%, 3.02%, 1.49%, 5.49%, 5.03%, 1.66%, 1.49%), (19.79%, 2.79%, 1.43%, 1.18%, 3.37%, 1.42%) higher than the compared methods, and the interval average convergences (IACs) are (−18.19%, −8.15%, 3.97%), (36.97%, 21.92%, 22.59%), (12.31%, 21.59%, 7.22%) compared to the existing methods in three different counties, respectively, which shows that the proposed method has better overall performance and applicability through our discussion. [ABSTRACT FROM AUTHOR]

The segmentation of airborne laser scanning (ALS) point clouds remains a challenge in remote sensing and photogrammetry. Deep learning methods, such as KPCONV, have proven effective on various datasets. However, the rigid convolutional kernel strategy of KPCONV limits its potential use for 3D object segmentation due to its uniform approach. To address this issue, we propose an Integrated Point Convolution (IPCONV) based on KPCONV, which utilizes two different convolution kernel point generation strategies, one cylindrical and one a spherical cone, for more efficient learning of point cloud data features. We propose a customizable Multi-Shape Neighborhood System (MSNS) to balance the relationship between these convolution kernel point generations. Experiments on the ISPRS benchmark dataset, LASDU dataset, and DFC2019 dataset demonstrate the validity of our method. [ABSTRACT FROM AUTHOR]

Metal laminated composites are widely used in industrial and commercial applications due to their excellent overall performance. In this study, the copper/graphene-aluminum-copper/graphene (Cu/Gr-Al-Cu/Gr) laminated composites were prepared by ingenious hot pressing design. Raman, optical microscope (OM), scanning electron microscope (SEM), van der Pauw (vdP), and X-Ray Diffractometer (XRD) were used to investigate the graphene status, interface bonding, diffusion layer thickness, electrical conductivity, Miller indices and secondary phases, respectively. The results demonstrate that the Cu-Al interfaces in the Cu/Gr-Al-Cu/Gr composites were free of pores, cracks and other defects and bonded well. The number of graphene layers was varied by regulating the thickness of the Cu/Gr layer, with the Cu/Gr foils fabricated by chemical vapor deposition (CVD). The electrical conductivity of the composite was significantly improved by the induced high-quality interfaces Cu/Gr structure. The increased number of graphene layers is beneficial for enhancing the electrical conductivity of the Cu/Gr-Al-Cu/Gr composite, and the highest conductivity improved by 20.5% compared to that of raw Al. [ABSTRACT FROM AUTHOR]

With the digital development of the construction industry, a new competence model of CPM (construction project manager) to meet the new requirements is highly needed. However, there is still a paucity of research on CPM competences in the new era, and no new CPM competence model has been constructed. To address this problem, this study uses data mining method to collect 2387 big data of recruitment advertisement in the Chinese construction market between August 2020 and October 2021 to construct a competence model of CPM in the digital era, which is called the Diamond model. The Diamond model consists of nine key competences, among which digital capability is a newly emerging one. The digital capability for CPMs is identified and classified into three levels, i.e., technology, knowledge and management. The requirements from the industry for the digital capability of CPM are mostly at the technology and knowledge levels. The results of the study provide a clear reference for both the demand and supply sides of CPM talents. Based on this, construction enterprises can formulate more accurate recruitment strategies, practitioners can dynamically benchmark themselves to the ever-changing industry needs and education, and training parties can better bridge the talent gap. Furthermore, the introduction of digital capability into the CPM competence system also provides a solid departure point for further follow-up research. [ABSTRACT FROM AUTHOR]

The reliable operation of photovoltaic (PV) power generation systems is related to the security and stability of the power grid and is the focus of current research. At present, the reliability evaluation of PV power generation systems is mostly calculated by applying the standard failure rate of each component, ignoring the impact of thermal environment changes on the failure rate. This paper will use the fault tree theory to establish the reliability assessment method of PV power plants, model the PV power plants working in the variable environment through the hardware-in-the-loop simulation system, and analyze the influence of the thermal characteristics of the inverter's key components on the reliability of the PV power plant. Studies have shown that the overall reliability of bus capacitors, inverters, and PV power plants is reduced by 18.4%, 30%, and 18.7%, respectively, compared to when the thermal characteristics of bus capacitors are not considered. It can be seen that thermal attenuation has a great influence on the reliability of the PV power generation system. [ABSTRACT FROM AUTHOR]

Power systems optimization is generally subject to the compromise between performance and cost. The 2021 Texas grid outage illustrates the worldwide dangers for the regional-centralized power grid, with comparable advantages to safety and flexibility for the distributed energy system. The storage of household batteries helps balance grid load and increase system stability and flexibility. However, household storage battery is still not widely used today because of its high costs. Currently, research on increasing household battery storage applicability is focused largely on optimizing economic strategies, such as configuration, dispatching and subsidy policies, which rely substantially more on technologies and financial perspectives. Consumers are not 'rational' individuals, and non-economic incentives can affect their decisions without raising prices. This paper consequently proposes to encourage users to acquire household battery storage to increase efficiency of power dispatching and economic advantages based on behavioral economics. In this paper, an empirical research builds upon the utility model of behavioral economics incentives and purchase willingness. Moreover, the multi-objective genetic algorithm is utilized to optimize the dispatching of household battery storage by using grid variance and user revenues as optimizing goals. The results of this paper show that the behavioral economics incentive improves intention to buy the household battery energy storage by 10.7% without raising subsidies. By improving the energy dispatching strategy, peak-load shifting performance and user revenues are improved by 4.2% and 10.6%, respectively. [ABSTRACT FROM AUTHOR]