Thermocouple Nonlinear Compensation Method Based On Whale Algorithms For Optimizing Extreme Learning Machine

Sensor is the core component of the measurement and control system,and its measurement accuracy seriously affects the overall performance of the measurement and control system.With the development of sensor technology,the market has put forward higher and higher requirements for sensors.therefore,how to compensate sensor effectively has important theoretical research and application value.In practical applications,the measurement errors of various sensors are caused to varying degrees by external interference factors or their own non-linear factors.With the development of artificial intelligence technology,the error compensation method based on neural network has been widely used in modern sensor signal processing.In this paper,thermocouple is selected as the compensation object of sensor.Considering the influence of environmental temperature and its non-linear characteristics on measurement accuracy,advanced artificial intelligence technology is used to compensate sensor.The main research work is as follows:(1)Improvement of whale optimization algorithmAiming at the shortcomings of traditional whale optimization algorithm(WOA),such as slow convergence speed and low convergence accuracy,an improved whale optimization algorithm(AWOA)is proposed.AWOA algorithm adaptively guides the whale population to search in the right direction by elite individuals' evolution information through elite individuals' guidance mechanism,avoiding the algorithm falling into local optimum and improving the efficiency of the algorithm.In the later stage of the algorithm,chaotic dynamic weight factor is added to increase the search intensity near the optimal solution and improve the local search ability of the algorithm,Thus,the convergence speed of the algorithm is accelerated..The simulation results of 23 benchmark functions show that,compared with the basic WOA algorithm,the WOAWC algorithm proposed in reference [38] and the basic particle swarm optimization algorithm,the proposed AWOA algorithm has higher convergence speed,solution accuracy and stability.(2)Compensation of Nonlinear Error of ThermocoupleIn order to solve the problems of cold-end compensation and non-linear correction in thermocouple application.A non-linear processing method of AWOA-KELM thermocouple is presented in this paper.The input-output model is established by using the S-type thermocouple scale data.The parameters of the model are trained and optimized by the improved whale optimization algorithm proposed in this paper.Compared with the basic WOA algorithm optimization model,WOAWC optimization model and PSO optimization model,the advantages of AWOA algorithm are further proved.In addition,the training results of AWOA-KELM and the traditional least squares support vector are compared.Compared with RBF neural network,the simulation results show that AWOA-KELM model achieves higher compensation accuracy,faster training speed and better meet the real-time requirements.The method proposed in this paper can effectively eliminate theinfluence of temperature change at the cold end of thermocouple and non-linear thermoelectric characteristics.It can be used as software compensation for high-precision thermocouple sensors and compensation for similar non-linear systems.It has high practical value.