Research On Data Perception And Intelligent Maintenance Decision Of Distribution Transformer

The development of smart grid puts forward higher requirements for the intellectualization and reliability of distribution transformers.Effective condition assessment and health management of distribution transformers are of great significance for building a strong smart grid and ensuring the safe and stable operation of the power system.At present,the research results of distribution transformers health management are scattered and not integrated systematically,as well as the research at all levels is not perfect and needs to be improved.In this paper,the key technologies of data perception and intelligent maintenance decision-making for distribution transformers are studied,aiming at forming an integrated technology system of intelligent health management for distribution transformers,which can improve the accuracy and effectiveness of transformer health management and intelligent maintenance,as well as provid a technical theory support and a relatively complete implementation process guidance for health management of distribution transformers.Data perception refers to the effective collection of equipment data,the establishment of a unified data environment,and extracting current equipment status data information through further analysis and perception of collected data,as well as completing the diagnosis and analysis of equipment health status.According to the requirements of distribution transformer intelligent health management technology system,the key technologies of distribution transformer data perception and intelligent maintenance decision-making studied in this paper mainly include four aspects:intelligent data acquisition,information feature extraction,health status assessment and prediction,as well as intelligent maintenance decision-making.The existing fixed and undifferentiated data acquisition strategy is prone to generate a large number of redundant data and leads to a low acquisition efficiency.Aiming at the above problem,an intelligent data acquisition method based on intelligent sensing strategy is proposed.For all kinds of data,according to their importance in condition assessment of distribution transformers and the real-time requirement,the priority of data acquisition is determined by using four quadrant diagram combined with fuzzy comprehensive evaluation method,corresponding to different data acquisition strategies.An adaptive sensing acquisition strategy is proposed for real-time data acquisition.According to the influence of equipment status saltation on the data change degree,the adaptive adjustment scheme for acquisition interval is formulated based on the degree of data fluctuation.Then the change scale of acquisition interval is adjusted on the basis of the dynamic proportional relationship between acquisition interval and data change,in order to achieve a high-precision adaptive adjustment of data sampling interval.The simulation results show that this method can reduce the amount of data acquisition while ensuring the data quality,and improve the operational efficiency.Moreover,it can guarantee and improve the reliabilty of distribution transformers health management from the data source.Aiming at the effective feature extraction for distribution transformer operation status data,a data feature extraction method based on Empirical Wavelet Transform(EWT)and Multiscale Entropy(MSE)is proposed.The EWT is adopted to adaptively decompose the operation state data of distribution transformer into empirical wavelet function(EWF)components.Hilbert transform is used to extract the characteristic information of different frequency and establish the time-frequency representation of signals.By calculating the correlation coefficients betweent the EWFs and the original signal,the effective components with high correlation are selected through threshold setting,and their multiscale entropy is calculated.With the combination of multiscale entropies of the effective EWFs,the signals eigenvector is constructed to represent the operation state information of the distribution transformer.The simulation results show that this method can effectively extract the features of the distribution transformers operation status data and provide high value density information for the health status evaluation of distribution transformers.In view of the current problems such as lack of research on the performance degradation modeling of distribution transformers,less consideration on the impact of mechanical faults in health status assessment,and one-sided evaluation method,this paper presents an online performance degradation evaluation method for distribution transformers based on EWT-MSE and k-medoids clustering under multi-index,as well as a prediction method of distribution transformers remaining benign operation time based on online performance evaluation and offline status evaluation.As the vibration signals can reflect the mechanical properties of distribution transformers and dissolved gases in oil can reflect the chemical and electrical faults,these on-line monitoring data are taken as reference indexes for on-line performance evaluation of distribution transformers.The EWT-MSE method is used to extract the characteristics of distribution transformers index data in different operation states.With the extracted eignvectors,the transformer performance degradation evaluation model is established by k-medoids clustering.Through weighted calculation of transformer health confidence under different indicators by analytic hierarchy processs,the final online performance degradation evaluation results of distribution transformers are obtained.Based on the historical data of transformer service life,test information and fault defect information,the offline health status assessment of distribution transformer is realized with the health index formula.Finally,the online health confidence and offline health index are synthesized to calculate the remaining benign operation time of distribution transformers.The simulation results verify the validity and accuracy of this method in the performance evaluation and residual benign operation time prediction of distribution transformers.It provides a theoretical basis for subsequent transformer intelligent maintenance decision-making.Aiming at the lack of research on maintenance cycle decision-making of distribution transformers,an optimization method for transformer maintenance cycle decision-making based on fault rate model is proposed to remedy the deficiency of existing condition-based maintenance and improve the accuracy of maintenance cycle of transformers with deteriorated health performance but no faults.With Levenberg-Marquardt(L-M)parameter estimation method,a transformer basic failure rate model based on Weibull distribution is established.Then the model is modified according to maintenance records and performance degradation assessment results,and the maintenance period of distribution transformer is calculated with reliability function as threshold constraint condition.The simulation results show that,compared with the existing planned maintenance and condition-based maintenance,this method can effectively improve the accuracy of maintenance cycle of distribution transformer under normal working condition but with degraded performance,thus avoiding over-maintenance or under-maintenance problems.To realize intelligent self-maintenance of distribution transformer to a certain extent,a cyber physical system(CPS)model framework for intelligent maintenance of distribution transformer is proposed.The physical system model and information system model of transformer are established separately.The self-maintenance of distribution transformer is realized by visiting and feedback control of physical system with information quantity.An intelligent maintenance CPS model for regulating the distribution transformer output voltage is taken as the simulation example to illustrate the feasibility and validity of CPS technology in the application of distribution transformers intelligent self-maintenance.