基于优化VMD-TCN-LSTM的变压器油中溶解气体预测OA

As for the such complex characteristics as both long-term trends and short-term subtle fluctuations of non-stationary dissolved gas sequences in transformer oi,the GSSA-VMD model is formed by integrating the golden sine algorithm(GSA)-optimized sparrow search algorithm(SSA)with variational mode decomposition(VMD).The origi-nal dissolved gas sequences in transformer oil are decomposed by GSSA-VMD and a set of stationary modal compo-nents is finally obtained.Then,for accurately predicting the long-term trends and short-term fluctuations of gas se-quences of transformer,in this paper the temporal convolutional network(TCN)and long short-term memory(LSTM)are combined,which is further combined with GSSA-VMD to form a hybrid prediction model for dissolved gas con-tent in transformer oil.Finally,in this paper the dissolved gas CO2 in teranformer oil is selected for experimental verification.It is concluded by the comparative experiments with VMD-TCN-LSTM,EMD-TCN-LSTM,and GSSA-VMD-LSTM models that the hybrid prediction model proposed in the paper achieves the best performance,with a mean absolute percentage error(MAPE)of 0.71%and a root mean square error(RMSE)of 9.04 μL/L.