Transformer state parameter combination prediction method based on cloud system similarity weight distribution

Abstract

The invention discloses a transformer state parameter combination prediction method based on cloud system similarity weight distribution, and the method comprises the steps of collecting and obtainingthe state parameter time sequence data generated by the operation of a transformer, forming an original time sequence data flow X, and dividing the original time sequence data flow X into a trainingset Xtr, a verification set Xva, and a test set Xte; constructing and obtaining an independent model training tuple omega = {U, V} by using the training set Xtr, training M pre-selected independent prediction models by using the independent model training tuple, obtaining prediction results of the trained M independent prediction models for the verification set Xva, and converting the prediction results for the test set Xte into a normal cloud system; calculating the overlapping area between the cloud models at the same positions of the two cloud systems to obtain the similarity, and further obtaining the overall similarity between the cloud systems; and distributing prediction weights to the independent prediction models according to the overall similarity between the cloud systems, and obtaining a final combined prediction result in combination with a prediction result. Compared with an existing prediction method, the method has thehigher prediction precision and the higher fault-tolerant rate.

Description

technical field [0001] The invention belongs to the technical field of transformer state monitoring, and relates to a transformer state parameter combination prediction method, in particular to a transformer state parameter combination prediction method based on cloud system similarity weight distribution. Background technique [0002] On-line monitoring of state parameters is an important basis for carrying out transformer state monitoring. However, at present, there is still a gap between the online monitoring of transformer state parameters and the realization of comprehensive condition maintenance of transformers. One of the important reasons is that the accurate value of state parameters cannot be obtained at any time. Predicting the changes of state parameters according to the existing state parameters is a very important and necessary supplement to the on-line monitoring of transformer state, and it is also an important means of early warning of transformer faults. I...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to provide a combination prediction method of transformer state parameters based on cloud system similarity weight distribution, to solve the problem that a single optimal adaptation prediction model cannot be obtained due to the high randomness of transformer state parameter time series data

Images