Electric load forecasting, a crucial support function for power system planning, faces increasing complexity with the deepening of electricity market reforms and the emergence of new technologies and roles like the sharing economy, load aggregators, and virtual power plants. These developments introduce new characteristics to electric loads, posing significant challenges to predictive modeling. Traditional load forecasting methods, primarily reliant on statistical analysis or expert models, struggle to accurately construct load models in this evolving power system environment, owing to the diverse and complex factors influencing loads. To address this challenge, in the paper, we proposed a novel electric load forecasting model based on Fast Fourier Transform-optimized (FFT) Transformer model. This model redefines electric load forecasting as a time series problem, empowering it to extract periodic features of electricity usage from data. The model has demonstrated significant effectiveness on an electric load dataset from a specific region, covering over 2,000 power users in three years.