State estimation of a power network
Various embodiments include a computer-aided method for state estimation of a power network using an artificial neural network with one or more inputs technically associated with the power network. The method may include calculating a state estimation using captured temporal measurement signals of the power network as inputs for the neural network. At least an amplitude of a harmonic of one of the measurement signals is used as an input for the neural network.
1 . A method for state estimation of a power network using an artificial neural network with one or more inputs technically associated with the power network, the method comprising:
calculating a state estimation using captured temporal measurement signals of the power network calculated at nodes of the power network as inputs for the neural network, the state estimation includes real-time voltages at the nodes and real-time branch currents between the nodes;
wherein at least an amplitude of a harmonic of one of the measurement signals is used as an input for the neural network;
wherein the neural network has been trained using a training dataset comprising a plurality of amplitudes of harmonics of one or more measurement signals or comprising a plurality of rise times and/or fall times of one or more measurement signals.
2 . The method as claimed in claim 1 , further comprising ascertaining the amplitude of the harmonic using a Fourier transformation of the measurement signal.
3 . The method as claimed in claim 1 , further comprising using one or more ratios of amplitudes of harmonics of the measurement signal as inputs.
4 . The method as claimed in claim 1 , further comprising using one or more rise times and/or one or more fall times within the measurement signal as inputs.
5 . The method as claimed in claim 4 , further comprising ascertaining the rise times and/or fall times using a Padé-Laplace method;
wherein the rise times and/or fall times are ascertained using the poles of the Padé approximation of the Laplace transforms of the measurement signal.
6 . The method as claimed in claim 5 , further comprising using an order [n/(n−1)] for the Padé approximation.
7 . The method as claimed in claim 6 , further comprising increasing n up to a stipulated maximum value until no further rise times and/or fall times are ascertained.
8 . The method as claimed in claim 1 , further comprising using amplitudes of subharmonics of the measurement signal as inputs.
9 . The method as claimed in claim 1 , wherein the measurement signals comprise one or more voltage signals, one or more current signals, one or more active power signals, or one or more reactive power signals of the power network.
10 . The method as claimed in claim 1 , further comprising capturing the measurement signals at nodes of the power network.
11 . The method as claimed in claim 1 , wherein the power network comprises a low-voltage network and/or a medium-voltage network.
12 . A method for training an artificial neural network for the state estimation of a power network with one or more inputs technically associated with the power network, wherein the state estimation is calculated using captured temporal measurement signals of the power network calculated at nodes of the power network, the state estimation includes real-time voltages at the nodes and real-time branch currents between the nodes, the method comprising:
training the network using a training dataset with a plurality of amplitudes of harmonics, a plurality of rise times, and/or a plurality of fall times of one or more measurement signals; and
using the amplitudes of the harmonics, the rise times, and/or the fall times as input to train the neural network.