Fault diagnosis method and system of npc three-level inverter based on dbn

Abstract

The invention relates to a DBN-based NPC three-level inverter fault diagnosis method, comprising: building an NPC three-level inverter experimental platform; The three-way voltage signal of the bridge arm is sent to the upper computer of the NPC three-level inverter experimental platform; the upper computer will use the upper, middle and lower bridge arm voltage signals after S transform and singular value decomposition SVD as the feature vector of the final classification ; Send the training set to the improved deep belief network I-DBN for training; use the test set to test the trained and improved deep belief network I-DBN, analyze the diagnosis results and display the results in the display interface of the host computer come out. The invention also discloses a DBN-based NPC three-level inverter fault diagnosis system. The invention solves the problem that the traditional feature extraction method is not suitable for the feature extraction of the sudden change fault signal and the non-stationary fault signal in the three-level inverter.

Description

technical field [0001] The invention relates to the technical field of power electronic fault diagnosis, in particular to a DBN-based NPC three-level inverter fault diagnosis method and system. Background technique [0002] Inverters are widely used in power generation systems, and are power electronic devices that undertake power conversion. The failure of the inverter is directly related to whether the power conversion can be carried out normally, and even the safe and reliable operation of the entire power grid. The three-level inverter is the most common inverter in the power system at this stage. Compared with the previous traditional two-level inverter, the three-level inverter has the advantages of small switching loss, more output levels, and output current. Low harmonic content and adaptability to higher voltage levels. However, compared with the two-level inverter, the three-level inverter is applied to more power switching devices, resulting in a higher failure r...

AI Technical Summary

Problems solved by technology

However, the above-mentioned feature extraction method is not effective in extracting the abrupt non-stationary signal that occurs when the NPC three-level inverter is faulty, and the above-mentioned neural network is easy to fall into a local optimal solution during training, which affects the final diagnosis result.

Images