Arc fault diagnosis method based on time domain feature parameter fusion

Abstract

The invention discloses an arc fault diagnosis method based on time domain feature parameter fusion. The arc fault diagnosis method comprises the steps of: step 1, designing a BP neural network used for arc fault diagnosis, wherein the design is implemented by (1) determining input and output modes of the neural network, (2) acquiring training set and test set samples of the neural network, (3) determining a number of layers of the BP neural network, (4) determining a number of nerve cells of a network hidden layer, and (5) selecting training parameters of the BP neural network; step 2, optimizing the BP neural network by adopting a genetic algorithm in a process shown in the chart 7, wherein the genetic algorithm optimization process is mainly implemented by (1) coding individuals and initializing a population, (2) calculating fitness, (3) and generating a new population. The arc fault diagnosis method has the beneficial effects that: the genetic algorithm optimized BP neural networkhas better performance compared with a BP neural network trained by using random weights and threshold values, and has relatively higher arc fault identification accuracy rate.

Description

technical field [0001] The invention relates to the field of arc fault detection, in particular to an arc fault diagnosis method based on fusion of time domain characteristic parameters. Background technique [0002] The time-domain fault characteristics of arc current are mainly reflected in the following three aspects: (1) The current waveform produces a "flat shoulder" that is approximately zero; (2) The high-frequency components of arc current are rich, and the "flat shoulder" of the current signal At the moment of arc reignition corresponding to the end, the current waveform undergoes periodic mutations; (3) when an unstable arc fault occurs, the symmetry of the current waveform decreases and random half-wave loss is likely to occur, and the average value of the current signal increases significantly. According to the above fault characteristics, the following three time-domain characteristic parameters are selected to extract the time-domain fault characteristics of th...

AI Technical Summary

Problems solved by technology

In addition, the individual current signal samples shown in the figure cannot be identified simply by a plane in three-dimensional space. Multiple test analyzes have found that the following two situations may make it difficult to identify current signal samples: (1) In the initial stage of arc combustion , the distance between the electrodes is small, the voltage between the electrodes is also relatively small, the arc combustion is not sufficient, the time-domain fault characteristics of the arc current are not obvious, and it is easier to identify the faults of individual current signal samples.

(2) Affected by factors such as electrode distance, electrode material, ambient temperature, and line load during the arc combustion process, the arc current waveform changes randomly, and the time domain fault characteristics of the arc current signal in individual cycles are weak. A plane of which does not recognize the failure sample

Images