Research On Temperature Prediction Method And Its Application In Thermal Management Of High-temperature Logging Tool

In the field of petroleum exploration,the high-temperature environment causes the deterioration of the heat dissipation conditions of electronic instruments,which puts forward an urgent demand for the thermal management of logging instruments.This thesis is oriented to the thermal management requirements of high-temperature logging tools,and researches on the temperature measurement and thermal prediction issues among them.Based on the Si P system-in-package technology,a temperature measurement circuit that can work at an ambient temperature of 200°C is designed.The FPGA,DSP,and SRAM related to temperature data acquisition and processing are packaged into a Si P device,and the temperature measurement circuit is designed using a proportional measurement method.The requirements for temperature measurement in high-temperature logging environments are met.The error analysis of the temperature measurement circuit shows that the temperature characteristic of the device is the main factor causing the measurement error.At high temperatures,the temperature measurement error increases significantly.The logging tool with thermos flask studied in this thesis has complicated internal circuits during long downhole operation and it is difficult to establish an accurate heat transfer model.Taking into account the continuous thermal inertia characteristics inside the tool,based on the measured temperature time series data,establish a data-driven temperature prediction method.The calculation of the existing temperature prediction algorithm based on the adaptive extended Kalman filter is relatively complicated for the embedded equipment in the logging instrument,and cannot be applied in the logging instrument in real-time.This thesis designs and implements an improved adaptive Kalman filter temperature prediction algorithm(IAKF),which uses the temperature state transition matrix as the system adaptive identification parameter to adapt to the influence of the internal thermodynamic circulation process of the instrument on the system state transition matrix,not only obtains better prediction accuracy but also simplifies the calculation.Analysis and test results show that the algorithm has stable performance and better solves the problem of real-time temperature prediction in high-temperature logging environments.In order to further improve the prediction accuracy and reduce the impact of nonGaussian noise in the time series,a temperature prediction method based on the XGBoost regression model is designed,which uses a combination of K-fold cross-validation and grid search for model training and hyperparameter tuning,simulation test,high-temperature experiment show that this method effectively improves the accuracy of temperature prediction.