Research On Power Quality Prediction And Early-Warning

The diversification of power equipment types,the complexity of line structure and the difference of power users’ requirements for power quality have led to the increasing attention to power quality problems.At present,the monitoring sites of power quality are increasing and the monitoring technology is becoming mature,leading to the expansion of the scale of power quality data.Effective analysis of power quality and acquisition of valuable information are the basis for solving power quality problems.Scientific prediction of power quality and timely warning can quickly master the steady-state indexes of power quality.Effective improvement and governance measures can be taken to ensure the safe and economic operation of the power grid.Therefore,the research on power quality prediction and early-warning is proposed.Firstly,the SOM-K-means combined clustering algorithm is designed to effectively solve the problems of the defect of slow convergence of SOM algorithm and initial clustering center of K-means.The optimal clustering number is determined by introducing contour coefficient.Secondly,an anomaly analysis and correction method of power quality based on SOM-Kmeans algorithm is proposed.The purpose is to detect outlier points and correct outlier data before establishing model.Reliable historical data can be provided for power quality prediction.Then,the prediction model of power quality based on BiLSTM optimized by Bayesian is constructed to predict the future state of steady-state indexes with high accuracy.The actual data of power quality are used to test the performance of the model.The results show that the proposed prediction model is effective.Finally,the traditional early-warning system based on national restriction and the earlywarning system based on SOM-K-means algorithm are established.The traditional early-warning system takes the national standard as the benchmark and divides the early-warning level into five levels from low to high.In view of the shortcomings of the traditional threshold setting method of power quality,SOM-K-means algorithm is proposed to divide the early-warning objects.Set the threshold and reset the early-warning level interval according to the clustering results.The analysis results of 18 monitoring points verify that the early-warning system based on SOM-K-means clustering can flexibly adjust the threshold and early-warning interval according to different objects,which is more in line with the requirements of the actual situation.