Generation method of wind power and photovoltaic output scenarios based on LHS-GRU

Abstract

With the continuous increase in the installed capacity of new energy sources, the proportion of new energy in the power system is gradually increasing. However, due to the strong stochasticity and volatility of new energy sources, the task of constructing a New Power System has become extremely daunting, posing difficulties for power dispatch. Traditional scenario generation methods cannot accurately describe the characteristics of new energy sources under typical and extreme scenarios. Therefore, this paper proposes a method for generating scenarios based on Latin Hypercube Sampling (LHS) and Gated Recurrent Unit (GRU). The kernel density estimation theory is used to model the probability density of wind and photovoltaic output, and the optimal bandwidth is found by the cross-validation. On this basis, the initial scenario sets are generated using the LHS. The GRU model is used to learn the trend of wind power and photovoltaic output at different times respectively, and generate scenario sets that can accurately reflect the typical and extreme output of wind power and photovoltaic. Three kinds of indicators were used to evaluate the generated scenario sets in this paper. The ADO indicators of the generated wind power and photovoltaic scenario sets are 2415 and 1462, respectively, ranking second and first among the four methods. The ES indicators of the generated wind power and photovoltaic scenario set are 1.3005 and 1.0286, respectively, both ranking first. Compared with the real scenario sets in 2022, the ES indicators of 2023 are 1.5361 and 2.2826, respectively, ranking first, and second. The BS indicator diagram also shows that the generated scenario set can reflect the real scenario set.