Low-voltage Distributed Photovoltaic Power Forecasting System Based on Decision Tree and Short-duration Memory Neural Network
Addressing the issue of weak"observability,measurability,adjustability,and controllability"in low-voltage distributed photovoltaics,this paper proposes a cross-supervised training model based on a lightweight gradient boosting decision tree and a long short-term memory neural network method from the perspective of low-voltage distributed photovoltaic power generation forecasting.The model establishes a forecasting algorithm for the total power generation of regional low-voltage distributed photovoltaics,achieving or exceeding the current quality of power generation forecasting for centralized photovoltaic power stations and medium-voltage distributed photovoltaics in both short-term and medium-term predictions.The proposed low-voltage distributed photovoltaic power generation forecasting system can provide an important reference for power grid scheduling,contributing to enhancing the consumption level of renewable energy and ensuring the reliability and safe and stable operation of the power grid.
photovoltaic power generationlow-voltage distributedlong short-term memory neural network
NETL
NSTL
万方数据
