Online Static Voltage Stability Evaluation Based On Data Mining

With the development of social economy and restrictions of environment,power grids are operating closer to their limits than ever before.Integration of renewable energy adds extra complexity and uncertainties into power system operations,which has brought forward new requirements on voltage stability assessment.Traditional static voltage stability assessment methods are time consuming and difficulty to meet the requirements of real-time operation requirements.Implementation of phasor measurement unit(PMU)enable system operators to use real-time measurement data for on-line voltage stability monitoring and control.This paper presents a new method for on-line static voltage stability assessment based on data mining techniques.The core idea is to apply data mining techniques to extract valuable information from a large amount of data obtained by offline voltage stability analysis results.Once voltage stability judgement rules are determined,PMUs are used to obtain real-time measurement data of those key variables for online voltage stability monitoring purpose.Firstly,static voltage stability methods are used to evaluate the power system.P-V curve analysis is performed to obtain the voltage stability critical point and voltage stability reserve factor under different network structures.On the basis of above work,modal analysis and sensitivity analysis are applied at the limit point of the PV curve.And the method of fuzzy cluster analysis is introduced to identify the weak voltage area.Data mining is applied based on the static voltage stability assessment of the power grid.The large number of input feature variables may cause the time longer and the accuracy lower in the process of model classifying.To solve this problem,this paper proposes a feature selection method from the essence of voltage instability.Firstly,the preliminary subjective screening will be done relying on modal analysis.Secondly,optimize the selection according to the Relief Feature Selecting Algorithm.Then the accuracy gets higher and the time modeling gets shorter when the simplest combination in the prediction of static voltage stability is obtained.At last,decision tree is selected as the classifier,and the cost sensitive learning mechanism is introduced into it.This paper presents a cost sensitive decision tree algorithm.The algorithm targets the minimum of the misclassification cost,which can avoid diagnosing voltage instability as stability to a certain extent,thus reducing the false dismissal probability.Then dispatchers can extract voltage stability operating guidelines from decision tree and the operating guidelines can help dispatchers assess system voltage stability with PMU data in real-time.Finally,a simulation example of a province power grid illustrates the effectiveness and feasibility of the proposed procedure.