Prediction of Dissolved Gases Concentration in Transformer Oil Based on VMD and ELM

Effective prediction of dissolved gases in transformer oil serves as a crucial foundation for diagnosing faults and issuing early warnings for power transformers. In this paper, we propose a dissolved gas concentration prediction model in oil based on variational modal decomposition (VMD) and extreme learning machine (ELM). Firstly, the original oil dissolved gas concentration sequence is decomposed by variational modal decomposition to obtain a number of relatively smooth sub-sequences; secondly, the extreme learning machine model is used to establish different prediction models for each sub-sequence; in conclusion, the prediction outcomes of each sub-sequence are combined to derive the prediction results for the original gas concentration sequence. The exemplar outcomes demonstrate that the model, employing variational modal decomposition and the extreme learning algorithm, yields a root mean square error of 0.327 and an average absolute percentage error of 2.71. In comparison with the prediction outcomes of other models, it is evident that the method introduced in this paper exhibits higher prediction accuracy and better capability to forecast the concentration of dissolved gas in the oil.