Research On Power Quality Prediction Of Low-voltage Substation Based On Situation Awareness

As the proportion of new energy in the power system continues to increase,the number of low-voltage substations including photovoltaic grid connected systems will increase.PV grid connection will bring voltage fluctuation,flicker,harmonic distortion and other power quality problems to low-voltage substations,which may seriously damage the electrical equipment on the user side of the substation.In order to deal with the power quality problem in low-voltage substation area,this paper uses situation awareness technology to realize the power quality prediction in low-voltage substation area.The situation display link in situation awareness technology is achieved by identifying the topology of low-voltage substation area.On the basis of situation display,power quality analysis and prediction are conducted on the nodes in the substation area.The situation prediction link of situation awareness technology is realized by predicting the power quality trend at the nodes in advance.Combining the situation display and situation prediction,it can fully realize the situation awareness and fine control of the low pressure platform area,and try to avoid losses,which has important engineering application value.Firstly,this paper establishes a low-voltage substation area model with photovoltaic grid connection through Matlab/Simulink,and simulates the actual electrical data of the photovoltaic grid connection system by bringing in the actual environmental data,including the voltage data,frequency data and harmonic distortion data at each node.After that,the relevant data obtained from the simulation are processed with missing values and outliers to provide data support for subsequent topology identification and power quality prediction.Aiming at the problem of low-voltage substation topology identification,this paper introduces a substation topology identification method based on improved principal component analysis and hierarchical clustering algorithm through the node voltage data obtained from the simulation model.This method first reduces the dimension through improved principal component analysis,then realizes the phase relationship of substation users through reduced dimension clustering,and finally realizes the identification of substation electrical boxes by clustering each phase.Finally,the topological relationship among substation,box and a home are acquired.Through simulation trials,the viability of this approach is confirmed,and the case analysis accurately identifies the topology of the station area.For the prediction of power quality at the nodes of low-voltage substations,this paper predicts the relevant indicators that affect the stability of power quality to reflect the quality of power quality in the future.First,the voltage data,frequency data and harmonic data collected from the photovoltaic grid connection model established by Matlab/Simulink are calculated to obtain the steady-state index data affecting power quality,such as voltage deviation,frequency deviation and harmonic distortion rate.The simulation experiment uses BP neural network,GRU cyclic neural network and Bayesian optimized GRU cyclic neural network prediction model to predict and analyze three typical indicators that affect power quality factors.The results show that the GRU prediction method using Bayesian optimized super parameters has higher accuracy.Finally,the Bayesian optimized GRU prediction model shows the best prediction performance by predicting the power quality of the relevant nodes in the actual photovoltaic grid connected low-voltage substation area.