Abstract: Due to equipment failures, operational errors, external interference, and other reasons, there are often a large number of outliers in power big data. This paper studies an intelligent detection method for outliers in power big data based on the Isolation Forest algorithm. Collect and preprocess power big data as a foundation. Select key features from the fused power big data using principal component analysis. Build an isolated forest algorithm model, classify and identify selected features, and obtain intelligent detection results for abnormal values in power big data. The experimental results show that the design method performs well in accuracy, ROC curve, and acceleration ratio, and can achieve good intelligent detection of power big data outliers.