Machine Learning in Electric Power Systems Adequacy Assessment Using Monte-Carlo Method

Abstract

The article considers the question of increasing the computational efficiency of 
the procedure for electric power systems adequacy assessment using the Monte Carlo method. 
In the framework of using this method, it is necessary to randomly generate a certain number 
of system states. As it is known the speed and accuracy of the calculation depends on the number 
of such states to be analyzed, so one of the ways to solve this problem is to reduce the this 
number while observing the required accuracy of the estimate. For this purpose it is proposed 
to use machine learning methods, whose task is to classify the calculated states of the electric 
power system. During the experiment, the support vector machines method and the random forest 
method were applied. The results of the calculations showed that these methods using allowed to 
reduce the number of random states of the system to be analyzed, thereby reducing the total time 
spent on calculations in general and proving the effectiveness of the proposed approach. Wherein 
the best results were obtained while using the random forest method.