Electric power system state estimation method based on convolutional neural network

Abstract

The invention provides an electric power system state estimation method based on a convolutional neural network. According to the method, offline power system power flow section data is taken as a sample for training; the quantity measurement is used as input data; The convolutional neural network state estimation method is characterized in that a state quantity is taken as an expected output, training parameters are reduced by means of local connection, weight sharing, downsampling and the like of a convolutional neural network, the network complexity is reduced, overfitting is prevented by means of back propagation of errors between input and output, and a convolutional neural network state estimation model based on quantity measurement is trained. The method estimates the new quantity measurement data through the trained convolutional neural network to obtain the state quantity of the system at the moment. According to the method, the convolutional neural network is adopted for state estimation, the defects that a traditional iterative least squares state estimation method is too long in calculation time and gradient transmission of a full-connection neural network state estimation method is difficult are overcome, and the training difficulty of the network is reduced while the calculation time is reduced.

Description

technical field [0001] The invention belongs to the technical field of power system operation and control, and in particular relates to a method for estimating the state of a power system based on a convolutional neural network. Background technique [0002] With the development of the smart grid, the scale of the power grid increases, and the structure and operation mode become increasingly complex. In order to more accurately grasp the operating state of the power grid, it is necessary to perform state estimation on the operating state of the power system. [0003] At present, the least squares estimation is mostly used in the state estimation of the power system, but due to the nonlinear constraint relationship between the quantity measurement and the state quantity, the least squares state estimation needs repeated Gauss-Newton iterations to perform nonlinear equations. The calculation time is long and the computer hardware requirements are high, which cannot meet the re...

AI Technical Summary

Problems solved by technology

[0003] At present, the least squares estimation is mostly used in the state estimation of the power system, but due to the nonlinear constraint relationship between the quantity measurement and the state quantity, the least squares state estimation needs repeated Gauss-Newton iterations to perform nonlinear equations. Solving, the calculation time is long and the computer hardware requirements are high, which cannot meet the real-time state estimation needs of large power grids

In recent years, with the development of artificial intelligence and data mining technology, the fully connected BP neural network has also been used in the field of state estimation, but the fully connected neural network will increase exponentially with the increase in input measurement. , and the gradient transfer of the fully connected neural network is difficult to exceed three layers, which makes the model training of the fully connected neural network more difficult

[0004] The present invention proposes a power system state estimation method based on convolutional neural network, which is used to solve the problems of traditional least squares estimation with long calculation time, high hardware requirements and difficult training of fully connected neural network

Images