In a long running oil-immersed transformer, the inner oil would inevitably be decomposed to many kinds of gases including C2H2, C2H4, CH4, H-2, by partial discharge resulting from certain faults. The detections of these gases have been extensively accepted as a good method to estimate the operation state of transformers. In this paper, Pd-4 cluster CNT is employed to simulate its adsorption performances and response behaviors to these typical gases based on density functional theory. The adsorption energy, DOS as well as orbital theory are combined analyzed towards these adsorption systems, which reaches high consensus. The adsorption ability for these analytes are in order as C2H2 > C2H4 > H-2 > CH4. Specifically, such Pd modified complex has a good sensitivity to C2H2, C2H4 and H-2, with increasing conductivity in C2H4 system while decreasing conductivity in C2H2 and H-2 systems, but presents insensitivity to CH4 owing to their weak sorption. The conductivity change of Pd doped adsorbent upon C2H2 is larger than that upon H-2. Our calculations suggest that Pd-CNT is a promising material to be prepared as sensors and employed in transformers to estimate their operation status as its good sensitivity and selectivity to oil decompositions. (C) 2017 Elsevier B.V. All rights reserved.