Multi-Dimensional Clustering-Based Innovative Machine Learning Algorithm for Short-Term Load Forecasting

This paper presents a novel machine learning algorithm based on multi-dimensional clustering for short-term load forecasting. The proposed approach employs a combination of Catboost and LSTM techniques using the recursive least square algorithm (RLS). The primary objective of this paper is to clearly identify the Electric Peak demand as well as their occurring times so that energy control algorithms can optimize and set appropriate resources to reduce the peak load demand, which can result in improved efficiency of local distribution networks and reduced environmental impact of energy generation and distribution. Simulation experiments demonstrate that the proposed algorithm outperforms several benchmark methods in terms of key metrics such as mean absolute percentage error (MAPE), root mean square error (RMSE), and bias.