基于PSO-LSSVM的热电偶非线性校正方法研究

Research on nonlinear correction method of thermocouple based on PSO-LSSVM

为改善热电偶温度传感器的非线性特性,构建基于粒子群优化算法(particle swarm optimization,PSO)和最小二乘支持向量机(least squares support vector machine,LSSVM)的热电偶非线性校正模型。针对LSSVM算法参数难确定的问题,选用PSO算法搜索LSSVM算法中惩罚系数和核函数参数的最优组合,用优化后的PSO-LSSVM校正模型逼近热电偶的非线性函数关系。为验证该模型的有效性,分别采用BP网络模型、RBF网络模型、LSSVM模型和PSO-LSSVM模型进行热电偶非线性校正,结果表明:PSO-LSSVM模型在热电偶非线性校正应用中表现出最优的稳定性和准确性,其最大拟合误差仅为0.12℃,均方误差为0.0033,准确率达到99.82%。将该模型应用于有限空间爆炸温度的非线性校正中,可取得较好的实际应用效果。

In order to improve the nonlinear characteristics of thermocouple temperature sensor,a nonlinear correction model of thermocouple based on particle swarm optimization(PSO)and least squares support vector machine(LSSVM)was constructed.Given the high difficulty of determining the parameters of LSSVM algorithm,PSO algorithm was used to search the optimal combination of penalty factor and kernel function parameter in LSSVM algorithm.Subsequently,the optimized PSO-LSSVM correction model was used to approximate the thermocouple nonlinear function.Furthermore,to detect the validity of the model,BP network model,RBF network model,LSSVM model and PSO-LSSVM model were used to correct the thermocouple nonlinearity.The results showed that the PSO-LSSVM model has the best stability and accuracy in the application of thermocouple nonlinear correction.Its maximum fitting error is only 0.12℃;the mean square error is 0.0033;and the accuracy rate is 99.82%.In summary,this model has been applied to nonlinear correction of explosion temperature in finite space,and has achieved good practical application results.