Energy Storage Planning in Distribution Networks Using Prophet Load Decomposition

To address the challenges of photovoltaic (PV) curtailment and bidirectional power flow overloads in distribution networks with high PV penetration, this paper proposes a planning method for energy storage siting and sizing based on load trend extraction and forecasting. First, the Prophet algorithm is used to decompose the net load of each node into trend and seasonal components, and the residual load is obtained. Second, a Generative Adversarial Network (GAN) is employed to model the relatively random residual component, generating new load forecast data for the next year. This generated residual data is then combined with the trend and seasonal components to obtain a new annual net load that conforms to the original statistical patterns. Third, based on the forecasted load data, an improved Bayesian optimization algorithm and non-linear programming are used for energy storage siting and sizing to obtain the optimal storage configuration scheme. Finally, simulations are conducted using real data from the following year to verify the foresight and economic efficiency of the proposed model.