In order to accurately predict the trend of dissolved gas concentration

a prediction method of dissolved gases concentration in transformer oil based on ensemble empirical mode decomposition（EEMD） and extreme learning machine（ELM） was proposed in this paper. Firstly

the original data set of dissolved gases was analyzed by EEMD to obtain several sub-sequences. Then

different prediction models were

respectively

established for different components based on the extreme learning machine algorithm

moreover

an improved particle swarm optimization（IPSO） algorithm was introduced to optimize parameters of prediction model to improve accuracy of the prediction. Finally

prediction results of each component were superimposed and summed to obtain the overall prediction results. The results show that the root mean square error of the IPSO ELM mode based on EEMD is 0.13

the average relative error is 0.502%

and the maximum relative error is 1.914%. Compared with the prediction results of other models

the improved particle swarm optimization algorithm can obtain better extreme learning machine network input weights and hidden layer threshold

so as to improve the prediction accuracy. The validity of the method was verified by the calculation results of a case.