User Adjustable Load Potential Analysis Method Based on Two-Level Clustering and SA-TCN Prediction

Under the background of the steady development of the new power system and the continuous construction of the "double carbon" target, the application scenarios of power load forecasting are also showing an increasingly complex and diversified trend. Accurate power load forecasting response plays an important role in the safety, stability and economy of power system operation. Therefore, this paper proposes an adjustable potential analysis method based on SA-TCN (Self-Attention Temporal Convolutional Network). Firstly, based on the performance requirements of adjustable load proposed by power system supply and demand balance, the index system of adjustable load characteristics is constructed. Secondly, based on the Canopy and Kmeans two-level clustering method, the user data are classified according to the daily load rate and daily peak-valley difference rate, and the SA-TCN prediction model is used to predict the user's load. Finally, based on the historical load data of industrial users in a typical area and the index system of adjustable load characteristics, the advantages of the proposed model (SA-TCN) in the stability of prediction accuracy are verified by an example analysis, and the adjustable potential of load is analyzed from multiple perspectives.