Method for predicting transformer oil chromatographic data based on extreme learning machine

Abstract

The invention discloses a method for predicting transformer oil chromatographic data based on an extreme learning machine. The method comprises the following steps: 1) collecting a plurality of gas concentration historical data in the transformer oil chromatographic data to obtain a training sample set; 2) taking the concentration at certain time of a plurality of gas as input, taking the concentration at next time of gas to be predicted in a plurality of gas as output, and using the training sample set to train on the extreme learning machine; wherein the node number of an input layer of the extreme learning machine is the variety number of gas, the node number of an output layer is 1; and a prediction model of the gas to be predicted can be obtained after the training is completed; and 3) using the model to be predicted to predict the concentration of the gas to be predicted. The good nonlinear function approaching capability of the extreme learning machine can be used to predict the gas concentration in the transformer oil, and the precision and the generalization capability of the transformer oil chromatographic data predicted model can be improved.

Description

technical field [0001] The invention relates to a transformer oil chromatographic data prediction method, which predicts the gas concentration in the transformer oil and belongs to the technical field of power system safety. Background technique [0002] In actual operation, transformer insulating oil and organic insulating materials will gradually age and decompose under the action of electric field and magnetic field, producing a small amount of low-molecular hydrocarbons and gases such as carbon dioxide and carbon monoxide, which are dissolved in large quantities in transformer oil. When there is a latent overheating fault or a discharge fault, the generation rate and the amount of these gases dissolved in the oil will also increase, that is, the composition and content of the fault gas are closely related to the severity of the fault type. Therefore, detecting the chromatogram of the insulating oil in the transformer is one of the important means to see the safe operatio...

AI Technical Summary

Problems solved by technology

[0004] At present, the commonly used forecasting methods are the gray forecasting model and its improved model, but the gray forecasting model and its improved forecasting model have a better forecasting effect when the series has a deterministic trend, otherwise the effect is poor, and it cannot be guaranteed to be accurate and satisfactory under any circumstances. the result of

The artificial neural network algorithm is easy to fall into the local minimum problem, and the global optimal solution cannot be obtained, and the convergence speed is slow

Images