Study On The Correlation Between The State Variables Of Substation Equipment In Poweron Test And Power-off Test

The scale of the power grid and the number of electrical equipment reached all-time high.However,current electrical equipment maintenance relies on the regular off-line tests.Although much manpower and resources are invested in the maintenance to ensure the stable operation of electrical equipment,the dilemma that reaching the limitation of maintenance capacity is imminent.The widespread usage of the sensor in power systems provides status data without shutdown of the equipment to be tested.However,the data accuracy has yet to be verified,resulting in the on-line data does not receive sufficient attention.Besides,power companies are actively promoting the state maintenance mode based on the charge test with no comprehensive and objective study of the relevance between the power-off test and the power-on test.Blindly prolonging or canceling a power outage test may prevent equipment defects from being detected opportunely and even lead to serious faults.This paper counted the distribution of anomalous cases detected by power-off test state variables and power-on test state variables according to the actual operation of the main part of the transformer,bushing of transformer,capacitor voltage transformer and current the Transformer.The abnormal case detection effect for each state variable was calculated.State variables with high level of detection are selected.Two correlation analysis model was established for the same and different state variables in the power-on test and power-off test respectively.(1)For each of the four substations equipment mentioned above,the main state variables that can characterize the faults of each equipment were summarized.The distribution of anomalous cases detected by patrol,power-on test and power-off test was statistically obtained.A method for calculating the detection sensitivity of the state variables of these three detection methods was established to obtain the detection effect of the corresponding state variables.A method for calculating the level of state parametric detection in online monitoring of substations is established based on missing data and the anomalies of non-missing data.The detection effect of various state variables for patrol,power-on detection,power-off test and online monitoring was quantitatively studied by case studies.(2)For the same state variables in the four types of substation power-on test and power-off test,a method for calculating numerical errors between power-off and poweron state variables data based on distance metric was established.A method for calculating trend error between power-off and power-on state variables data based on Dynamic Time Warping was established.A correlation coefficient calculation method non-stationary time series of the same state variable in power-off test and power-on detection was established based on the Detrended Cross Correlation Analysis.The correlation index system and analysis model for the same state parameter were established.The correlation between power-off and power-on data of dissolved gas in transformer oil,dielectric loss factor and capacitance of capacitive equipment were calculated by case study.Moreover,the relevance of the correlation results was validated.(3)For the different state variables in the transformer power-off test and power-on test,association rules were mined to establish association relationship between different state variables based on the Apriori Algorithm.A method for calculating the information correlation between different state variables based on the information entropy theory was established.Based on Detrended Partial Cross Correlation Analysis,a method for calculating the detrended partial correlation of non-stationary time series of different state variables is established.Then,the index system and analysis model for the correlation between different state variables were established.The correlations between the dissolved gas in transformer oil and the typical state variables in transformer power-off test were calculated by case study.The validity of the correlation model between the different state variables was verified.