Regional Distributed Photovoltaic Short Term Power Prediction Method Based on Cluster Analysis and Stacking Ensemble Learning

Accurate and reliable photovoltaic short term power prediction is of great significance to the improvement of photovoltaic consumption capacity, the day ahead scheduling and the safe and stable operation of the power grid. To ensure an accurate prediction, in this paper, a regional distributed photovoltaic short term power prediction method based on stacking ensemble learning and cluster analysis is proposed. In the proposed method, several weather patterns are firstly identified by using the K-means++ algorithm based on the historical data. Then, for each weather pattern, all the photovoltaic panels are clustered into several groups based on the k-Shape algorithm, where a prediction model for each photovoltaic cluster is established by employing the Stacking ensemble learning algorithm. Finally, based on the numerical weather forecast (NWP) of the forecast day, we select the best suited weather pattern and obtain the final prediction result by employing the trained prediction model under this weather pattern. The effectiveness of the proposed strategy is verified by the actual dataset.