Adaptability verification and application of the t-distribution in short-term load forecasting error analysis

Probabilistic short-term load forecasting plays an important role in the risk assessment and stochastic scheduling of power systems. Most related studies assumed that short-term load forecasting errors follow the normal distribution However, this assumption lacks substantial supports of real data, and is more like an empirical judgment. This paper firstly analyzes the characteristics of forecasting errors, by comparison with a variety of other probabilistic distribution functions. Then, the adaptability of the t-distribution is discussed and identified. An empirical analysis based on solid, massive, extensive and credible real data in China is carried out to verify the effectiveness of the t-distribution, from the perspectives of different months, periods as well as locations. Finally, the t-distribution is implemented in probabilistic short-term load forecasting based on case studies.