Line loss is directly responsible for the management profitability of the grid company. The traditional method of calculating the theoretical line loss for Low Voltage Distribution Networks (LVDN) necessitates more el...
详细信息
Line loss is directly responsible for the management profitability of the grid company. The traditional method of calculating the theoretical line loss for Low Voltage Distribution Networks (LVDN) necessitates more electrical parameters. which cannot be obtained easily. Besides, due to the backward communication conditions of LVDN, the problem of smart meter data missing is significant, which poses a challenge to an exact theoretical line loss calculation. In an attempt to solve the issues above, a theoretical line loss computation approach via matrix completion and relieff-convolutional neural network (CNN) for LVDN is proposed. Firstly, a feature weighting algorithm based on relieff is presented to analyze the relevance of the electrical parameters, which can be obtained easily. Secondly, a theoretical line loss calculation method is proposed for CNN-based. In the view of the data missing problem, a matrix completion method based on singular value thresholding (SVT) is introduced to obtain the high-precision data, in order to enhance the calculation accuracy of the theoretical line loss calculation. Finally, the proposed method is tested on the data sample of 789 LVDNs. The results show that comparing with CNN, back-propagation and other methods, the mean absolute percentage error (MAPE) of the presented method can reduce by more than 90%. When data missing, the MAPE of the proposed method can reduce by more than 95% compared with the method without considering the data completion. (c) 2023 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://***/licenses/by-nc-nd/4.0/).
With(1) the rapid development of the Internet, Internet credit business has emerged and the process is now booming. As a result, there is a problem of predicting credit demand of users. Therefore, we propose a method ...
详细信息
ISBN:
(纸本)9781450365123
With(1) the rapid development of the Internet, Internet credit business has emerged and the process is now booming. As a result, there is a problem of predicting credit demand of users. Therefore, we propose a method of using big data analysis to forecast the credit demand of users in this paper, which is used to reduce the risk of credit business and improve the utilization of funds.
暂无评论