Effect of temperature measurement error on parameters estimation accuracy for thermal response tests.

Obtaining soil thermophysical parameters is the premise for design ground heat exchanger in ground source heat pump system, but it may not be accurately determined due to the limitations of the analytical models. In this paper, artificial neural network (ANN) is used to directly establish the mapping relationship between temperature response and soil thermophysical parameters, and the identification accuracy of traditional method and ANN under different measurement errors is compared. In addition, Kalman filter and fitting regression are used to remove the interference noise. The results show that the identification accuracy and stability of the traditional method are relatively weak affected by temperature measurement error, but the identification accuracy is limited. The maximum deviation errors of thermal conductivity and volumetric heat capacity are 10.68% and 18.42%, respectively, and no matter which kind of noise reduction method cannot improve the identification accuracy. The identification stability of ANN is relatively greatly affected by temperature measurement error, but the identification accuracy is high. The maximum deviation errors of the two parameters are 10.05% and 5.4%, respectively. Through the logarithmic function fitting of noise date can further improve the identification accuracy and stability, the maximum deviation errors are only 2.12% and 3.65%. • A way to determine soil thermal parameters based on neural network is proposed. • This way can establish link between temperature response and unknow parameters. • Different temperature measurement errors on identification accuracy are analyzed. • The identification accuracy of volume heat capacity by the new way is improved. • Logarithmic fitting of the noise data can further improve identification accuracy. [ABSTRACT FROM AUTHOR]