关键词: 电力系统, 惯量估计, 特征提取, XGBoost, 超参数调优

Abstract: Power system inertia estimation is crucial to ensure the stability of the power grid and to cope with unexpected events. A Bird-mating Convolutional XGBoost Network(BCXNet) algorithm is proposed to accurately estimate the inertia value of the power system. Firstly, a convolutional neural network(CNN) is used to extract features of system inertia and construct high-level time series data features containing multidimensional features. Secondly, the eXtreme gradient boosting(XGBoost) is used for inertia regression analysis, and the XGBoost hyperparameters are optimized by combining the bird-mating optimizer(BMO), thereby further improving the accuracy of inertia estimation. Finally, the accuracy of inertia estimation by the proposed method is verified through experiments, and the coefficient of determination(R²) reaches 0.987 3, the root mean square error(RMSE) is 0.079 7, and the mean absolute percentage error(MPAE) is 2.6%, which indicates that the estimated model output is in high agreement with the real value.

Key words: power system, inertia estimation, feature extraction, eXtreme gradient boosting(XGBoost), hyperparameter tuning