Prediction of dissolved gas in power transformer oil based on LSTM-GA

Abstract

Because the change process of the content of dissolved gas in transformer oil is fluctuating and is affected by many factors such as oil temperature and external environment, the gas content does not increase exponentially and is nonlinear and non-stationary. This feature determines the current Prediction technology cannot accurately predict the concentration of dissolved gas in transformer oil. To improve the prediction accuracy of the dissolved gas content in the transformer oil, more accurately evaluate the transformer status and customize the entire preventive plan under the alarming development trend of the transformer, to minimize the prediction error, this paper proposes a transformer combining genetic algorithm and long short-term memory (LSTM)neural network Prediction model of dissolved gas content in oil. The genetic algorithm (GA) is used to optimize look back(lb), lstm nets(ls), epochs(ep), and the dropout(dp), and then the genetic algorithm is combined with the long short-term memory neural network. The gas content is predicted. This model overcomes the problem of low prediction accuracy caused by selecting parameters based on experience. The analysis result of the calculation example shows that compared with the traditional prediction algorithm, the proposed method can better track the change law of the dissolved gas concentration in the oil, improve the prediction accuracy, and provide a strong guarantee for the safe and stable operation of power transformers.