Research On Classification Of Power Quality Disturbances Based On Random Forest

In modern society,electricity as a widely used high-efficiency renewable energy,with the characteristics of clean and convenient,economical and practical has become the most widely used energy production and people's life.Its application has become an important symbol of the comprehensive national strength reflects the state and level of economic development.With the new power electronic equipment the emergence of a large number of and put into use,the electricity sector has begun to pay attention to power quality problems,and the majority of users are also putting forward higher requirements on the quality of power supply.Therefore,it is necessary to study the classification of power quality disturbances.The main work of this paper are as follows:first,use S transform to extract the feature.According to the IEEE's power quality standard,the normal waveform and 16 common power quality disturbances' waveforms are modeled mathematically,and the power quality disturbance signals are analyzed by S transformation.By comparing the time domain characteristic curve and frequency domain characteristic curve of the power quality disturbance signal after the S transform,11 kinds of time domain and frequency domain feature are extracted.Second,this paper proposes a method of identification of power quality disturbance based on random forest.The 11 kinds of power quality disturbance signals,which are extracted from the S transform,are used as input to train the random forest model,and the accuracy of the method is verified by test data.Compared with decision tree and support vector machine,the power quality disturbance recognition algorithm based on random forest has higher accuracy of recognition and has good anti-noise ability.Third,the power quality disturbance recognition algorithm based on random forest is optimized.A feature extraction method combining wavelet packet entropy and S transform,a total of 19 kinds of extraction feature as the random forest model input.And the random forest model was optimized.The comparison experiment results prove that compared with the original method,the optimized method with disturbance higher recognition accuracy and better anti-noise ability.The power quality disturbance identification method proposed in this paper is very important for monitoring the power quality of the power grid,ensuring the safety and stability of the power grid,and improving the economic benefits.