Power transformers are the main equipment of transmission and transformation systems

and is of great significance to ensure its healthy and stable operation.The collected data from transformers can reflect the operating status from various aspects.Using this information to predict the running state of the transformer has important research significance.The state monitoring data is a complex data sequence.The prediction method is used to predict the future trend of the parameters

so that the operating state of the transformer can be better grasped.The input information of the traditional method is usually a single or a few dimensions to fit the future trend

and there is a phenomenon that the input feature information is insufficient and there is coupling between different information dimensions.In order to improve the prediction accuracy of transformer state parameters

this paper proposes a short-term prediction of gas concentration in transformer oil based on a combined model of GRNN and KPCA.The gas concentration ratios are also used as characteristic parameters

which expands the dimension of the characteristic parameters.Secondly

the kernel principal component analysis method is used to select the main parameters from the feature parameters

which reduces the redundancy of information.Finally

the generalized regression neural network is used to predict the gas concentration in the transformer oil.