Joint communication and sensing for fallen tree localization on overhead lines
In sharp contrast to the prior art, a fallen tree detection and localization method based on distributed fiber optical sensing (DFOS) technique and physics informed machine learning is described in which DFOS leverages existing fiber cables that are conventionally installed on the bottom layer of distribution lines and used to provide high-speed communications. The DFOS collects and transmits fallen tree induced vibration data along the length of the entire overhead lines, including distribution lines and transmission lines, where there is a fiber cable deployed. The developed physics-informed neural network model processes the data and localizes the fallen tree location along the lines. The location is interpreted in at least two aspects: the fallen tree location in terms of the fiber cable length; and the exact cable location (power cable or fiber cable) that the fallen tree mechanically impacts.
1 . A fallen tree localization method comprising:
operating a distributed fiber optic sensing (DFOS) system configured to receive vibration data of overhead transmission and distribution lines;
analyzing the vibration data using frequency domain decomposition (FDD) techniques, producing identification of natural frequencies of the overhead transmission and distribution lines;
performing a Singular Value Decomposition (SVD) on a Power Spectral Density (PSD) matrix obtained from the FDD analysis, generating a singular sequence containing discriminative features;
performing a feature extraction and fallen tree detection by using a residual convolution block to process the singular sequence;
determining the fallen tree location and overhead line onto which the fallen tree has fallen; and
reporting the fallen tree occurrence, location, and overhead line affected.
2 . The method of claim 1 wherein the residual convolution block is included in a ResNet network.
3 . The method of claim 2 wherein the natural frequencies are determined by FDD techniques by determining a Singular Value Decomposition (SVD) of a Power Spectral Density (PSD) matrix, plotting singular values against corresponding frequency lines, and estimating the natural frequencies from the plotted singular values.
4 . The method of claim 3 wherein the ResNet network includes a convolutional layer, two max-pooling layers, twelve 1-D residual blocks and two fully connected layers.
5 . The method of claim 4 wherein the ResNet network residual blocks include a set of learnable filters configured to perform a convolutional operation based on the filter size.