Distributed Pv Power Output Forecasting Considering the Impact of Degradation Rate

The degradation rate of photovoltaic (PV) panels affects the accuracy of PV power output forecasting. To address this problem, this paper proposes a distributed PV power output forecasting method that considers the impact of degradation. First, an annual degradation rate histogram is constructed based on measured data to calculate the annual degradation rate of the PV system. Next, the Fuzzy C-Means (FCM) clustering algorithm is applied to classify data into four weather types: sunny, cloudy, overcast/rainy, and extreme weather. Furthermore, a forecasting model combining a Deep Neural Network (DNN) and LightGBM is proposed, where DNN is used to predict PV power output on sunny days, while LightGBM handles the remaining weather conditions. Case study analysis demonstrates that the proposed model achieves higher forecasting accuracy compared to other models.