Abstract
To address the problems of missing measurement data and low efficiency of large-scale measurement data complement in power distribution system, a measurement data complement method based on Clustering-BayesLSTM combined model is proposed. Firstly, K-mediods clustering is utilized to categorize all the quantity measurements, select the main/standby crucial measurements in each category, and calculate their average deviation from other measurements in order to effectively reduce the solution size of the long short-term memory (LSTM) neural network model. Then, through the LSTM prediction model based on Bayesian optimization, the missing main/standby crucial measurements are complemented. In addition, the average deviations from the primary/backup crucial quantity measurements are used to supplement the other missing measurements. Finally, the effectiveness of the proposed method is verified based on the real measurement data of a local power distribution system.
Funding Information
  • State Grid scientific and technological projects of Tianjin Electric Power Company (KJ22-1-39)

This publication has 18 references indexed in Scilit: