The renewable energy sector is becoming more important to cover the current demand of electricity. Solar energy is a promising source of renewable energy, especially in the last few years. However, like any other industrial processes, a photovoltaic (PV) system is susceptible to diverse defaults and anomalies that might occur during its operational functioning. The system’s performance is affected by the interaction between the climate and material properties of photovoltaic panels, leading to a decrease in performance. Therefore, fault detection and identification in PV systems are essential tasks to achieve optimal operating performance. They provide operators with relevant informations about system operation. In this paper, smart detection algorithms of anomalies in solar plants are proposed based on Artificial intelligence techniques. We have proposed two approaches: The first one is based on Machine Learning (ML) methods for multi-class classification and the second one is based on Deep Learning (DL) methods for anomaly detection in the dataset. By comparing the performance of various anomaly detection models in two solar power plants, the experimental results showed that the AE-LSTM method outperforms XGBOOST in detecting anomalies in data. Concerning the machine learning methods KNeighborsClassifier, Support Vector Classifier (SVC) and ExtraTreesClassifier are more efficient, they achieve an accuracy more than 97% compared to the AdaBoostClassifier method which provides a low accuracy less than 47%.