Research on harmonic source location method based improved GROMP

In the case of insufficient measuring equipment and less harmonic measurement data, harmonic state estimation based on compressed sensing can effectively locate harmonic sources. Firstly, considering the general fact that harmonic sources in power systems are spatially sparse, improved Gradient Descent Regularized Orthogonal Matching Pursuit algorithm (GROMP) is proposed to estimate harmonic states by using limited measurement data. It solves the problem of divergence and leverage in harmonic state estimation based on least square method under limited measurement data. Then, a neural network combining temporal convolutional network, Long Short-Term Memory network and attention mechanism is proposed to generate pseudo-measurement data based on different frequency harmonics to expand the measurement range. Finally, an example of IEEE33 nodes is given to verify the effectiveness of the proposed method.