Research And Improvement Of Power System Discretization Algorithm Based On Runge-Kutta Method And CAIM Method

With the construction of smart grid and the development of power Internet of things in modern power system,a large number of data from continuous power grid monitoring equipment brings great challenges to the traditional attribute discretization method.Therefore,how to improve the discretization efficiency of massive data from continuous power network monitoring equipment,find the needed data in such huge data,so as to carry out fault prediction,stability analysis and so on,has become a serious problem that needs to be solved.The effectiveness of discrete algorithm determines the accuracy of subsequent machine learning directly while reducing the redundancy of the data as much as possible.And the algorithm is also suitable for real-time big data analysis and prediction of modern power system.The Runge-Kutta discretization method is commonly used in stability analysis and off-line safety analysis of power grid system.This method has high calculation precision but due to the large amount of calculation,the computing efficiency is not satisfied.CAIM discretization algorithm is one of the most common methods for digital-to-analog conversion of data.The algorithm is efficient and suitable for big data analysis.However,the algorithm itself has some defects and errors are easy to appear.In this paper,from the point of improving the discrete algorithm commonly used in power system,firstly,the common Runge-Kutta method is analyzed,and the repeatability calculation process is pointed out.Aiming at this shortcoming,it is simplified and the simplified operation formula is obtained.Combining the PID control strategy commonly used in power system with Runge-Kutta method,the calculation formula is obtained,and an example is given.It is proved that this method has a high accuracy and numerical stability with greatly reducing the performance period of the algorithm.Secondly,the data source is optimized,and the idea of preprocessing the data from UCI machine learning database is used to analyze the operation of CAIM algorithm.It is pointed out that the comparison of attribute weight and data difference is neglected in the process of operation of the algorithm,which is easy to cause large operation error.By means of mathematical definition of attribute weight and quantification of data difference,etc.The CAIM discretization algorithm is improved,and the method is tested by support vector machine(SVM)and C4.5 decision tree.It is proved that the accuracy and effectiveness of the proposed method are much higher than that of the traditional method.