In order to eliminate temperature measurement errors caused by temperature variation at the cold side and nonlinear thermoelectric characteristics in the thermocouple temperature measurement process

a nonlinear compensation method for thermocouple sensor based on whale optimizing algorithm kernel extreme learning machine（ WOA-KELM） is proposed. This method uses whale algorithm to search and optimize on parameters of the kernel extreme learning machine to construct input-output model for thermocouple

at the same time

the non-linear correction and cold-junction-compensation of the thermocouple are completed. The training results of the proposed model based on the thermoelectric potential-temperature scale data of S-type thermocouple are compared with the traditional LSSVM method and RBF neural network

which verify the faster training speed and higher fitting precision of the proposed method.