Harmonic Source Identification Based on RF-GSO-LSTM Model

With the development of ultra-high voltage and smart grids, the problem of grid harmonic pollution caused by nonlinear loads and equipment has become increasingly serious. In the comprehensive management process of grid harmonics, harmonic source identification is the primary problem to be solved. This paper proposed a harmonic source identification model that utilizes Random Forest (RF) to select important features, Geese Swarm Optimization (GSO) algorithm to optimize the hyperparameters of the Long Short-Term Memory (LSTM) model. And then employed the RF-GSO-LSTM model for harmonic source identification. The feature subset selected by RF can simplify model input and reduce information redundancy, while GSO algorithm can select the optimal parameter combination of LSTM model, improving the identification efficiency. The experimental results demonstrate that the proposed RF-GSO-LSTM model has better performance.