We analyze time series of potential energy fluctuations and principal components at several temperatures for two kinds of off-lattice 46-bead models that have two distinctive energy landscapes. The less-frustrated "funnel" energy landscape brings about stronger nonstationary behavior of the potential energy fluctuations at the folding temperature than the other, rather frustrated energy landscape at the collapse temperature with a significant 1/f(alpha)(alpha approximate to 1.5) noise structure and a significant deviation of the Allan variance from the law of large numbers. The principal components are analyzed by an embedding nonlinear time-series analysis. The fast fluctuations with small amplitudes of similar to70-80% of the principal components cause the time series to become almost "random" in only 100 simulation steps. However, the stochastic feature of the principal components tends to be suppressed through a wide range of degrees of freedom at the transition temperature.