This paper deals with the estimation of the actuator temperature in an electropneumatic system. First, an appropriate nonlinear system that accounts for the actuator temperature dynamics is introduced. Then in order to overcome the difficulty of installing temperatures sensors in each chamber of the actuator, two nonlinear high-gain observers are proposed to provide online estimate of these temperatures. The gain of both observers can be tuned by the choice of a scalar design parameter. However, the design parameter is constant for the first observer and its choice has to satisfy a compromise between an accurate estimation of the state estimation and a satisfactory sensitivity of the observer with respect to the unavoidable output noise measurements. This difficulty is overcome in the second observer since the scalar design parameter is time varying and is governed by a Riccatti differential equation. The involved adaptation process of the design parameter is mainly driven by the power of the output observation error norm computed on a moving horizon window. Simulation results are given to show the effectiveness of the proposed observers and in particular to compare the performance of both observers, namely, the accuracy of the respective estimates and their sensitivity with respect to noise measurements.