Dissolved gas analysis (DGA) based in insulating oil has become a more mature method in the field of transformer fault diagnosis. However, due to the complexity and diversity of fault types, the traditional modeling method based on oil sample analysis is struggling to meet the industrial demand for diagnostic accuracy. In order to solve this problem, this paper proposes a probabilistic neural network (PNN)-based fault diagnosis model for power transformers and optimizes the smoothing factor of the pattern layer of PNN by the improved gray wolf optimizer (IGWO) to improve the classification accuracy and robustness of PNN. The standard GWO easily falls into the local optimum because the update mechanism is too single. The update strategy proposed in this paper enhances the convergence ability and exploration ability of the algorithm, which greatly alleviates the dilemma that GWO is prone to fall into local optimum when dealing with complex data. In this paper, a reliability analysis of thirteen diagnostic methods is conducted using 555 transformer fault samples collected from Jiangxi Power Supply Company, China. The results show that the diagnostic accuracy of the IGWO-PNN model reaches 99.71%, which is much higher than that of the traditional IEC (International Electrotechnical Commission) three-ratio method. Compared with other neural network models, IGWO-PNN also has higher diagnostic accuracy and stability, and is more applicable to the field of transformer fault diagnosis.

Fault diagnosis technology of power transformers is essential for the stable operation of power systems. Fault diagnosis technology based on dissolved gas analysis (DGA) is one of the most commonly used methods. However, due to the lack of fault information, traditional DGA fault diagnosis techniques are difficult to meet increasing power demand in terms of accuracy and efficiency. To address this problem, this paper proposes a novel fault diagnosis model for oil-immersed transformers based on International Electrotechnical Commission (IEC) ratio methods and probabilistic neural network (PNN) optimized with the modified moth flame optimization algorithm (MMFO). PNN as a radial neural network has good utility and is often used in classification models, but its classification performance is easily affected by the smoothing factor (σ) of the hidden layer and is not stable. This paper addresses this issue using the MMFO to optimize the smoothing factor, which effectively improves the classification accuracy and robustness of PNN. The proposed method was validated by conducting the experiments with the real data collected from transformers. Experimental results show that the MMFO-PNN model improves the fault diagnosis accuracy rate from 70.65 to 99.04%, which is higher than other power transformer fault diagnosis models.

Since it is difficult for the traditional fault diagnosis method based on dissolved gas analysis (DGA) to meet today’s engineering needs in terms of diagnostic accuracy and stability, this paper proposes an artificial intelligence fault diagnosis method based on a probabilistic neural network (PNN) and bio-inspired optimizer. The PNN is used as the basic classifier of the fault diagnosis model, and the bio-inspired optimizer, improved salp swarm algorithm (ISSA), is used to optimize the hidden layer smoothing factor of PNN, which stably improves the classification performance of PNN. Compared with the traditional SSA, the sine cosine algorithm (SCA) and disruption operator are introduced in ISSA, which effectively improves the exploration capability and convergence speed. To verify the engineering applicability of the proposed method, the ISSA-PNN model was developed and tested using sensor data provided by Jiangxi Province Power Supply Company. In addition, the method is compared with machine learning methods such as support vector machine (SVM), back propagation neural network (BPNN), multi-layer perceptron (MLP), and traditional fault diagnosis methods such as the international electrotechnical commission (IEC) ratio method. The results show that the proposed method has a strong learning ability for complex fault data and has advantages in accuracy and robustness compared to other methods.

In the original publication [...]

No abstract