This paper presents an innovative system for smart agriculture, combining the Internet of Things with Long Range technology (IoT LoRa) and Machine Learning (ML) to respond to two major challenges: 1) Real-time wheat disease detection and intelligent irrigation management. The system is designed to identify and classify three common wheat diseases in Algeria (yellow rust, brown rust, and powdery mildew) using the MobileNetV2 Deep Learning (DL) model implemented on a Raspberry Pi 4 with a camera. The model achieves a significant accuracy, precision, recall, and F1-score of 98%, which enables quick and accurate disease identification. 2) For better optimization of irrigation time, environmental parameters such as humidity, soil moisture, temperature, and light intensity are monitored using IoT sensors. The data of these parameters are transmitted via LoRa protocol for long-range and low-power communication. Then, the data are analyzed in real-time using ML algorithms such as k-Nearest Neighbors (k-NN), Support Vector Machine (SVM) and Random Forest (RF) which, have the best accuracy, precision, recall, and F1-score of 99.8%. Fuzzy logic is also used to facilitate calculations, ensuring efficient use of water in irrigation. Farmers can access this system remotely via a platform and receive real-time alerts on potential diseases and irrigation needs. This integrated approach improves crop health and yield while meeting the challenges of climate change and water scarcity, especially in semi-arid regions like Algeria.