Transient stability evaluation method and system suitable for large power grid operation

Abstract

The invention discloses a transient stability evaluation method and system suitable for large power grid operation, and the method comprises the steps: forming different power system operation data through the dynamic adjustment of the size of a load and generated output based on the operation data of a power grid; forming an N-1 fault set based on the power system operation data; calling time domain simulation software to perform time domain simulation on the N-1 fault set to obtain a time domain simulation result; discretization processing is carried out on a state quantity used as a representation power system in the time domain simulation result; performing supervised learning training on the state quantity subjected to discretization processing and the state quantity which is not usedas a representation power system in the time domain simulation result by adopting a depth tree model to obtain a transient stability judgment model; and testing the transient stability judgment model, and performing hyper-parameter adjustment according to a test result to obtain an optimal model. According to the method of the invention, the transient stability judgment method based on the depthtree method provides support for the rapid decision of the operation of the large power grid.

Description

technical field [0001] The invention relates to the technical field of power systems, in particular to a transient stability evaluation method and system suitable for large power grid operation. Background technique [0002] Power system transient stability assessment has long been based on expert experience, combined with a large number of time-domain simulation stability calculations. The calculation is large and time-consuming. It is not suitable for large-scale power systems and needs to make quick decisions. It is difficult to quickly and comprehensively Analyze system characteristics to make decisions. [0003] In the prior art, there are methods based on time-domain simulation and deep learning methods based on neural networks. The method based on time-domain simulation can adapt to various complex models, and the accuracy of calculation results is high, but the disadvantage is that the calculation amount is large and the speed is slow. For predictive accident analy...

AI Technical Summary

Problems solved by technology

The method based on time domain simulation can adapt to various complex models, and the accuracy of the calculation results is high, but the disadvantage is that the calculation amount is large and the speed is slow

For predictive accident analysis based on time-domain simulation, one of the schemes is to use time-domain simulation to scan the critical cut-off time of all predictable accidents. Although this scheme can guarantee the correctness and completeness of the analysis results, it cannot be online due to the long calculation time. application

Using the method of time domain simulation to obtain the critical cut-off time of the accident usually adopts the half-search method, but it still needs 5 to 6 times or more simulation times, and the time cost is relatively large

The deep learning method based on neural network is not widely used at present. The main disadvantage is that it needs a lot of samples for training to get good results, so the training cost is relatively high; and the neural network often lacks suitable explanatory tools. The external performance is a black box, and it is difficult to explain its output

Images