Anomaly detection in environmental data is essential for ensuring the integrity and reliability of weather forecasts and climate models. This study leverages advanced machine learning techniques to detect anomalies in solar power generation data, focusing on key meteorological variables such as temperature, humidity, pressure, precipitation, wind speed, and radiation. Three main algorithms were employed: Isolation Forest, One-Class SVM, and Variational Autoencoder (VAE). The results indicate that VAE and Isolation Forest detected a similar number of anomalies, with significant overlap, while One-Class SVM identified slightly more anomalies. The study found a core set of anomalies consistently identified by all three methods, suggesting these are likely true positives. Additionally, unique anomalies detected by each method highlight their individual strengths in capturing different aspects of the data. By comparing the results of these algorithms, the study provides a robust framework for anomaly detection in solar power generation data, which is critical for improving the quality and reliability of environmental data.