Online Tuning Method of Main Grid Relay Protection Based on Convolutional Neural Network
[Purposes]Due to the diversity of fault states of the main grid and the adaptability of the set-ting rules,the closed-loop step response amplitude of the relay protection is difficult to be effectively controlled.Therefore,an online setting method for the relay protection of the main grid based on convolu-tional neural network is proposed.[Methods]The same protection function in the main grid is taken as a set of fixed value variables,and the online setting mathematical model of the main grid relay protection is constructed.In the specific setting stage of the main grid relay protection,convolutional neural network is introduced,and the loss function is used to calculate the optimal regularization relay protection setting under the stochastic gradient descent mechanism.[Findings]In the test results,the amplitude of the closed-loop step response of the test grid relay protection under the design method not only realized rapid convergence within 50s,but also was more stable than that of the control group.[Conclusions]It is shown that the design method has a significant advantage in the closed-loop step response,which not only improves the convergence speed of the system,but also improves the amplitude stability of the sys-tem,which is helpful to improve the operation efficiency and reliability of power grid relay protection sys-tem,and has the potential of practical application and popularization.
NETL
NSTL
万方数据
