The oil-immersed transformer represents a pivotal core component within the newly devised power system. It is of paramount importance to guarantee the secure and reliable operation of this apparatus, as it serves as the foundation upon which the overall stability and integrity of the power system is contingent. The accurate detection of transformer fault characteristics and the timely detection of potential faults can effectively prevent major safety accidents. During overheating, discharge faults and catalytic/aging processes (under the action of copper, iron and other metals) in the operation of oil-immersed power transformers, the insulating oil and solid insulating materials, including insulating paper, laminates or wood blocks, etc., decompose, producing a variety of gases which are dissolved in the transformer oil. By measuring the concentration of these gases in the oil or gas phase after pumping, it is possible to detect early faults in the transformer and to carry out the necessary maintenance in time.At present, commonly used multi-component gas detection methods mainly include gas chromatography, electrochemical sensing, semiconductor sensing, infrared absorption spectroscopy/photo-acoustic spectroscopy and so on, but all these methods have certain limitations. Raman spectroscopy offers several advantages over traditional gas detection methods, including high selectivity, non-destructive detection, and real-time detection. However, the Raman scattering cross-section of the gas is relatively small, and the Raman scattering intensity of the trace gas is weak, which results in a low sensitivity of Raman spectroscopy for the detection of gases. Therefore, how to enhance the intensity of the gas Raman scattered light signal is a core problem that urgently needs to be solved for Raman spectroscopic gas detection. Based on the cavity enhanced Raman detection technology, the multiple inverse cavities can extend the path length of the laser and the gas, which is one of the commonly used methods for gas Raman signal enhancement. Currently used Herriot multi-pass cavity or near-concentric multi-pass cavity can achieve tens to more than one hundred times the number of reflections, but the number of laser reflections in the cavity is still small, resulting in Raman signal enhancement of low amplitude, the gas detection limit is difficult to meet the requirements of the transformer fault characteristics of gas detection.This paper proposes a Z-shaped folded multi-pass cavity enhanced Raman spectroscopy detection technique. The optimal number of beams in the multi-inverse cavity is determined by establishing a mathematical relationship between the Raman signal intensity and the number of laser reflections in the multi-pass cavity. The design of intracavity parameters enables the realization of 144 intracavity beams, which complete two cycles within the cavity. This greatly extends the interaction length between the laser and the gas, thereby improving the Raman signal intensity of the gas. The constructed Z-shaped folded multi-pass cavity enhanced Raman spectroscopy detection platform was utilized to achieve highly sensitive detection of the primary fault characteristic gases of transformers. Furthermore, a high-fit quantitative analysis model between the gas concentration and the intensity of Raman spectral peaks was established based on the identified characteristic peaks. Notably, the lowest detection limit of each gas surpassed the IEC standard. This method represents a novel and efficacious approach to the highly sensitive detection of transformer fault characteristic gases.