PV power forecasting method based on IFCM

Abstract

To address the issues of abnormal data interference, insufficient utilization of meteorological similarity, and low temporal resolution in photovoltaic (PV) power forecasting, this paper proposes a multi-dimensional fusion model architecture—the IFCM prediction model—which integrates Isolation Forest (IForest), Fuzzy C-means (FCM), and Support Vector Machine (SVM). This architecture organically integrates three modules—anomaly detection, meteorological classification, and regression prediction—into a complete forecasting framework. Furthermore, to overcome the limitation of the traditional one-hour prediction interval, a refined temporal resolution is designed with a 5-minute sampling interval and a 24-hour forecast horizon (288 sampling points), meeting the stringent requirements of power grid dispatch for high temporal resolution. Experimental results demonstrate that the proposed method significantly improves prediction accuracy under complex weather conditions, such as rainy days, while exhibiting strong robustness and adaptive parameter capability.