A Decision-making Model Of The Power Grid Fault Evolution Based On The Alarm Data

With the construction of smart grids,data centers of power grids accumulate a large amount of data,including alarm signals produced by grid equipment automatically.Stud-ies of grid data traditionally focused on numerical data.There are not many studies of grid alarm data because of the text data in them,which are difficult to process.Grid alarm data contains the grid equipment state transition information,which can be used in grid faults location and diagnostic.It is of great significance to the automation of power grid operation and maintenance that mine the information in the grid alarm data and establish a grid fault evolution model.This article analyzes the characteristics of grid alarms from a grid alarm database.The modeling base characteristics of grid alarms are as follows: the text pattern of the text data of grid alarms can be enumerated? the grid alarm sequence can be partitioned into grid events? the alarm event can be represented by the key information in its grid alarm sequence.Based on the above characteristics,this paper proposes text pattern matching,density clustering,sequence partitioning,and natural language processing to process grid alarms.With these methods,grid alarms are transformed into vectorized grid alarms,vectorized grid alarm sequences and vectorized grid faults.These vectorized data are modelable and raised the scale of individual data from grid alarms to grid faults.This paper analyzes the correlation and evolution direction of grid alarms and grid faults using these vectorized data based on statistical analysis.A sequence prediction model called the “sequence prediction tree” is proposed to predict grid alarms,which is based on the variable-order Markov model.And a model called the “grid fault evolution model” is proposed to analyze the grid fault evolution direction and distribution,which is based on the Markov chain.Combining the above two models and the above data process-ing methods,this paper proposes a complete chain from grid alarms to grid fault evolutions called the “grid fault evolution decision-making model” and designs a software system for this model.Grid alarm data verification in this paper indicates that the grid fault evolution decision-making model is able to complete the analysis process from grid alarms to grid fault evo-lutions.The analysis results comply with the law of power grid alarms and the law of statistics,and can effectively represent the evolution direction and distribution of grid alarms and grid faults.The software system of this model can be used to support power grid operation and maintenance and assist grid decision-making.