Dynamic safety assessment method based on random forest and extreme learning regression

Abstract

The invention discloses an online dynamic safety assessment scheme based on random forest and extreme learning regression. The method specifically comprises the following steps: (1) obtaining an electric power system operation data sample by utilizing historical operation data of an electric power system and fault simulation based on an anticipated accident set, constructing a dynamic safety indexand forming an original sample set; (2) obtaining a key variable by utilizing a Gini index and a variable importance score by adopting a feature selection method based on a random forest; (3) training an extreme learning regression machine by using the key variables to obtain a mapping relationship; (4) updating the model by receiving the real-time operation data of the power system from the wide-area measurement system server, thereby completing real-time dynamic safety evaluation of the power system. According to the scheme, rapid and efficient real-time safety assessment is carried out onthe power system, and stable operation of the power system can be maintained.

Description

technical field [0001] The invention belongs to the field of power system dynamic security assessment, and specifically relates to random forest, Gini index, extreme learning regression fitting, and single hidden layer feedforward neural network algorithm. Background technique [0002] The modern power system is one of the most complex artificial systems in the world. Its operation is to maintain continuous power generation, transmission and distribution. However, the operation of the power system will inevitably suffer from various disturbances and failures. According to the disturbance and failure In severe cases, the safety of the power system may be lost, which may lead to catastrophic consequences, such as widespread blackouts. With the continuous increase of the scale of the power system, the safe operation of the power system is facing unprecedented challenges. In order to avoid the emergence of catastrophic situations and adapt to the expansion of the power system s...

AI Technical Summary

Problems solved by technology

[0006] (1) Some commonly used mutual information and relationship exploration tools, when performing correlation detection, more or less have the problem of overestimation or underestimation, which makes the feature selection not accurate enough, and the selected key variables are not enough for the following One-step classification or regression training

At the same time, it is difficult to find a good balance between computational efficiency and key feature subset quality

[0007] (2) Some traditional data-driven methods have some limitations when applied to the online dynamic security assessment of power systems, such as low computing efficiency, easy over-fitting, and difficulty in the analysis of large amounts of data, etc.

Images