Kalman filtering transformer fault prediction method and system based on neural network

Abstract

The invention discloses a Kalman filtering transformer fault prediction method based on a neural network. The Kalman filtering transformer fault prediction method comprises the following steps that monitoring data of a transformer are obtained; the monitoring data are subjected to segmentation processing according to time periods, and a data set of each time period is obtained; outlier points of the data sets are judged, the number of the outlier points is counted, the outlier points are converted into normal data values, and null points and ultra-range points in the data sets are removed to form new data sets; the new data sets are subjected to regression analysis, a model of performing regression analysis is established, and the monitoring data are predicted to obtain prediction trend results; and a fault diagnosis model is established, the prediction trend results are input to the fault diagnosis model to conduct fault diagnosis, and fault diagnosis prediction results are obtained.Possible future trends of the monitoring data can be predicted. Through establishment of the fault diagnosis model and a neural network model and through model prediction, diagnosis results of possible future faults of a high voltage transformer can be obtained.

Description

technical field [0001] The invention relates to a transformer fault prediction method, in particular to a neural network-based Kalman filter transformer fault prediction method and system. Background technique [0002] High-voltage transformers are key equipment in power systems, and the reliable operation of high-voltage transformers is crucial to the stable operation of power systems. Improving the reliability of transformers can significantly improve power supply reliability. In order to improve the reliability of the transformer, the current mainstream method is to conduct online monitoring of some operating parameters of the transformer according to national standards, monitor its key operating parameters, and generate an alarm when the key operating parameters exceed a certain threshold. This method is an online diagnosis of transformer faults. When the transformer online monitoring system sends out an alarm, the transformer often has a significant abnormal state or s...

AI Technical Summary

Problems solved by technology

This method is an online diagnosis of transformer faults. When the transformer online monitoring system sends out an alarm, the transformer often has a significant abnormal state or some faults have occurred, which has affected the normal and reliable operation of the power grid.

Another common practice is to carry out regular patrol inspection, live detection and condition maintenance for each transformer according to national standards. This method is an offline detection method. Compared with online monitoring, this method greatly improves the accuracy rate. Slight abnormalities or early failures of the transformer can be found in advance at any time, but this method requires a lot of manpower and material resources, and both live detection and condition maintenance will cause the transformer to be unavailable or limited during detection, reducing the effective work of the transformer. time

There is a long time interval between live inspection and condition inspection of the transformer, and the inspection is performed every six months to a year on average, depending on the operation of the transformer. The method of offline detection can know or avoid these abnormalities and failures in advance

Images