Addressing the issues of low processing efficiency and subjective diagnosis results in traditional transformer fault diagnosis methods, this paper proposes a transformer fault diagnosis method based on Deep Belief Neural Networks (DBN) for intelligent transformer fault diagnosis. The method conducts research through the following steps: Ratio Method without Encoding: Analyzing fault types by monitoring the ratio of specific dissolved gases in transformer oil. Deep Belief Neural Network Architecture: Utilizing Restricted Boltzmann Machines (RBM) to optimize the initial weights and thresholds of BP neural networks to enhance network performance. Model Construction: Building a fault diagnosis model based on DBN to improve diagnosis speed and accuracy. Training the diagnostic network with information collected from different dissolved gases in transformer oil, compared with traditional BP neural network method and Genetic Algorithm optimized BP neural network method (GA-BP), experimental results demonstrate that the proposed method can diagnose transformer faults more accurately, significantly improving diagnosis accuracy and efficiency. This research provides a new method and perspective for intelligent transformer fault diagnosis, with significant engineering application value and scientific research significance.