Prediction of Output Power of PV Power Stations Based on Particle Swarm Optimization-Random Forest

The photovoltaic (PV) power output exists uncertainty andvolatility due to external factors, so that the accurate prediction of its output power plays an important role in power scheduling and balance between supply and demand. To solve this problem, our prediction model was established based on the power characteristics of two PV power stations in a place for analysis. Firstly, the data set was pretreated and visualized in order to observe the data characteristics. Secondly, in order to optimize the performance of the model, the prediction model of Random Forest optimized by Particle Swarm Optimization (PSO-RF) was established, and the two parameters of decision tree number and split feature number of Random Forest were optimized. Finally, the cross-validation method was used to check and avoid the overfitting of the model, so that the generalization ability was also improved. The experimental results show that the average prediction accuracy of the model is higher than 94%, which has high prediction accuracy and practicability for the general PV power forecasting.