Distribution network historical measurement data correction method based on data mining and support vector machine

Abstract

The invention relates to a distribution network historical measurement data correction method based on data mining and a support vector machine. The method is technically characterized by comprising the following steps: step 1, reading a measurement data record in a distribution network master station historical database within a period of time, and performing discretization processing on measurement data such as voltage, current, active power and reactive power in the historical database; step 2, mining a strong association rule of a frequent item set in the historical measurement data through an Apriori algorithm, and identifying suspicious bad data; and step 3, constructing a data set with suspicious bad measurement data removed, training a regression model of the support vector machine, substituting the time corresponding to the suspicious bad data into the regression model, calculating a fitting value, replacing the suspicious bad data, and finishing data correction. According tothe invention, distribution network scheduling control decision is facilitated, the influence of bad data on distribution network operation is reduced, and efficient, stable, safe and intelligent operation of the distribution automation system is facilitated.

Description

technical field [0001] The invention belongs to the technical field of data mining, and relates to a method for correcting historical measurement data of a distribution network, in particular to a method for correcting historical measurement data of a distribution network based on data mining and a support vector machine. Background technique [0002] At present, the historical data saved by the master station of the distribution network may have bad data due to various reasons. Bad data will affect the accuracy of the distribution network scheduling control, which is not conducive to the safe and stable operation of the distribution automation system. [0003] The existing least square method and neural network are also commonly used for data fitting, but the deviation of the least square method fitting is sometimes too large, and the generalization ability is not strong; the fitting effect of the neural network is very poor when there is not much data , when there is too m...

AI Technical Summary

Problems solved by technology

[0002] At present, the historical data saved by the distribution network master station may have bad data due to various reasons. Bad data will affect the accuracy of distribution network scheduling control, which is not conducive to the safe and stable operation of the distribution automation system.

[0003] The existing least square method and neural network are also commonly used for data fitting, but the deviation of the least square method fitting is sometimes too large, and the generalization ability is not strong; the fitting effect of the neural network is very poor when there is not much data , when there is too much data, there may be a problem of "over-learning"

Images