1,1,1,3,4,4,4-heptafluoro-3-(trifluoromethyl)-2-butanone (C5-PFK) gas has attracted widespread attention at home and abroad due to its excellent electrical insulation properties and good environmental protection properties. Preparing high-precision C5-PFK mixed gas and accurately detecting the mixing ratio is conducive to the scientific demonstration of C5-PFK mixed gas and minimizes potential power hazards. The FTIR experiment, combined with the B3LYP method for spectral theoretical calculation, was used to study the infrared spectral absorption characteristics of C5-PFK gas. For CO2 and micro-water gas that may exist in the test environment, the cross-interference analysis of spectral lines was carried out under the same conditions as temperature, pressure and optical path; The C5-PFK gas mixture ratio sensor was simulated and tested based on NDIR technology, and the overall design of the sensor hardware system was carried out. According to the output characteristics of the sensor, a BP neural network temperature compensation model is established, and the repeatability and indication error of the sensor are tested. The results show that the strong absorption peak positions of C5-PFK gas are 1 200, 1 262 and 1 796 cm-1, respectively, and the molecular theoretical calculation agrees with the gass measured infrared spectrum. Under the background of synthetic air, the absorbance of CO2 gas at 1 262 cm-1 position is 6.04×10-7, the influence factor of micro water peak area in 150 nm filter bandwidth is about 3.15×10-3, and the cross-interference of spectral lines can be ignored. It is feasible to realize mixing ratio detection at a 1 262 cm-1 position. The range tracks the optical path, and the sensor simulation test results show that the 6.5 mm optical path can realize the detection of the C5-PFK mixture ratio of 0~15%. The output characteristics of the sensor show that the value of the absorption variable SA/SB decreases with the increase in temperature, showing a nonlinear relationship. The maximum indication error of 10% C5-PFK/Air mixture before and after temperature compensation by the BP neural network algorithm is 29.23% and 1.29%, respectively, and the output absorption variable SA/SB value remains unchanged after compensation. The sensor repeatability test shows that the RSD is 0.27%, less than 3%. The linear fitting coefficient R2 of the sensor indication corresponding to different concentrations is 0.999, and the maximum indication error is 2.47%. In summary, the advantages of the detection method and its sensor in the detection range of C5-PFK gas mixture ratio, anti-interference ability and reliability are verified, and a feasible solution for detecting the mixture ratio of C5-PFK gas mixture electrical equipment is provided.