Abstract: To fully utilize a power cable, temperature sensors or probes are commonly used in engineering practice to monitor the instant temperature to increase the ampacity while ensuring that the maximum temperature of the conductor is under the permissible limit. Due to the engineering restriction, directly measuring the conductor temperature of a constructed cable line is not feasible. Instead, a commonly used engineering method involves measuring the temperature at some accessible points and then inverting the conductor temperature and ampacity based on the measured temperature from the sample point. However, due to the inevitable installing error (at least ±1 centimeter), the actual position of the temperature probe in engineering application is not exactly the “designed” position, i.e., the there is an unavoidable uncertainty in the probe position. In such cases, if the temperature in the “designed” position of the temperature probe varies sharply, the measured temperature will deviate significantly, and the subsequent computational results of the temperature fields and ampacity will also incur significantly errors. To address the aforementioned issues on the uncertainty of the installing positions of temperature probes, this paper firstly explores the robust optimization model of the temperature probe location, then proposes a population based incremental learning method to solve the robust optimization problem of the temperature probe location, and finally presents a numerical example to validate the proposed model and method. Also, to increase the solution speed, the paper introduces a polynomial chaos based response surface model.