Medium-long-term electricity load forecasting based on NSNP systems and attention mechanism

Accurate load forecasting can provide important information support for intelligent operation of power systems, it can assist the power grid to deploy production plans in advance to uphold the equilibrium between the supply and demand for electrical power, or plan investment strategies based on the results of the forecast. Nonlinear Spiking Neural P (NSNP) system [1] belongs to a category of computational systems with distributed, parallel, and non-deterministic characteristics that have the analytical skill to solve nonlinear problems. Aiming at the temporal characteristics and complex nonlinear characteristics of electrical load data, this paper proposes a new Medium-Long-Term Load Forecast model LF-ASNP based on NSNP system and attention mechanism, which can accurately analyze the characteristics of historical load data and forecast the electrical load. In this paper, the LF-ASNP model is validated in several benchmark datasets, and the analysis of the experimental results fully demonstrates that the model can forecast the power load effectively and reliably.