Microalgae play a vital role in both environmental and industrial sectors. However, large-scale cultivation faces critical challenges, particularly microbial contamination, which drastically reduces biomass yield up to 70 %, often leading to sudden pond crashes. Preventing such biomass loss requires early and accurate identification of contaminants is crucial. Identification of contaminants based on microscopic and molecular methods remains labor-intensive, time-consuming, and require expertise in interpretation. To address these limitations, this study proposes a Deep Learning (DL)-based approach for automated classification of microalgae species using image data. A comprehensive dataset comprising 37 species, including green algae, diatoms, and dinoflagellates, was developed and 4 types of models (ResNet, MobileNet, DenseNet, EfficientNet) with a total of 12 variants were evaluated based on classification accuracy and F1-score. Among them, EfficientNet-B0 demonstrated the best performance, achieving an accuracy of 92.2 % while maintaining low computational overhead. The model uses AI to identify microalgae based on their shape and texture, enabling cost-effective and reliable monitoring.