Identifying topology of distribution substation in power Internet of Things using dynamic voltage load fluctuation flow analysis
Yongjin Xu,Jifan Lv,Jiaying Wang,Fangbin Ye,Shen Ye,Jianfeng Ji
DOI: https://doi.org/10.7717/peerj-cs.1688
2024-02-09
PeerJ Computer Science
Abstract:At present, the reconfiguration, maintenance, and review of power lines play a pivotal role in maintaining the stability of electrical grid operations and ensuring the accuracy of electrical energy measurements. These essential tasks not only guarantee the uninterrupted functioning of the power system, thereby improving the reliability of the electricity supply but also facilitate precise electricity consumption measurement. In view of these considerations, this article endeavors to address the challenges posed by power line restructuring, maintenance, and inspections on the stability of power grid operations and the accuracy of energy metering. To accomplish this goal, this article introduces an enhanced methodology based on the hidden Markov model (HMM) for identifying the topology of distribution substations. This approach involves a thorough analysis of the characteristic topology structures found in low-voltage distribution network (LVDN) substations. A topology identification model is also developed for LVDN substations by leveraging time series data of electricity consumption measurements and adhering to the principles of energy conservation. The HMM is employed to streamline the dimensionality of the electricity consumption data matrix, thereby transforming the topology identification challenge of LVDN substations into a solvable convex optimization problem. Experimental results substantiate the effectiveness of the proposed model, with convergence to minimal error achieved after a mere 50 iterations for long time series data. Notably, the method attains an impressive discriminative accuracy of 0.9 while incurring only a modest increase in computational time, requiring a mere 35.1 milliseconds. By comparison, the full-day data analysis method exhibits the shortest computational time at 16.1 milliseconds but falls short of achieving the desired accuracy level of 0.9. Meanwhile, the sliding time window analysis method achieves the highest accuracy of 0.95 but at the cost of a 50-fold increase in computational time compared to the proposed method. Furthermore, the algorithm reported here excels in terms of energy efficiency (0.89) and load balancing (0.85). In summary, the proposed methodology outperforms alternative approaches across a spectrum of performance metrics. This article delivers valuable insights to the industry by fortifying the stability of power grid operations and elevating the precision of energy metering. The proposed approach serves as an effective solution to the challenges entailed by power line restructuring, maintenance, and inspections.
computer science, information systems, artificial intelligence, theory & methods