SMARTHYDRO AI: DEVELOPMENT OF AN INTELLIGENT HYDROPONICS SYSTEM UTILIZING IOT AND ARTIFICIAL INTELLIGENCE
Abstract
Abstract: Modern agriculture faces resource constraints and efficiency demands, driving the adoption of Precision Agriculture. This study aims to develop SmartHydro AI to overcome the limitations of conventional hydroponic systems, which typically focus on partial parameter monitoring and lack integrated data analytics. The novelty lies in the hybrid integration of the Internet of Things (IoT), machine learning (K-Means), rule based logic, and Large Language Models (LLM) into a smart, communicative decision support system architecture. Developed using the Verification and Validation Model (V-Model), the system utilizes six sensors pH, TDS/EC, DS18B20, DHT11, BH1750, and water level to monitor environmental and nutritional parameters. A continuous 30 day experiment was conducted, targeting 518.400 data logs stored in a Supabase database via 5 second MQTT transmission intervals. Testing results demonstrated high sensor accuracy with percentage errors ranging from 0.00% to 7.35% compared to standard reference instruments. MQTT communication exhibited stable performance with actuator response times within second. K-Means effectively categorized hydroponic conditions into three distinct clusters. Furthermore, the Gemini API integration generated natural-language corrective recommendations with strong contextual appropriateness and an average processing latency of 12.10 seconds. Ultimately, SmartHydro AI successfully unifies monitoring, analysis, and recommendations within a technically valid, intelligent hydroponic architecture. Keywords: smarthydro AI; IoT; machine learning; LLM; intelligent hydroponics Abstrak: Pertanian modern menghadapi keterbatasan sumber daya dan tuntutan efisiensi yang mendorong penerapan Precision Agriculture. Penelitian ini bertujuan mengembangkan SmartHydro AI untuk mengatasi keterbatasan sistem hidroponik konvensional yang berfokus pada pemantauan parsial dan belum menyediakan analisis data terintegrasi. Kebaruan penelitian ini terletak pada integrasi hibrida Internet of Things (IoT), machine learning (K-Means), logika rule based, dan Large Language Model (LLM) dalam satu arsitektur sistem pendukung keputusan yang cerdas dan komunikatif. Dikembangkan menggunakan Verification and Validation Model (V-Model), sistem ini memanfaatkan enam sensor (pH, TDS/EC, DS18B20, DHT11, BH1750, dan water level) untuk memantau parameter nutrisi dan lingkungan. Eksperimen kontinu selama 30 hari berhasil mengumpulkan 518.400 log data ke basis data Supabase via interval MQTT setiap 5 detik. Hasil pengujian menunjukkan akurasi sensor yang tinggi dengan persentase kesalahan 0,00% sampai 7,35% dibandingkan instrumen standar. Komunikasi MQTT berkinerja stabil dengan waktu respons aktuator detik. K-Means efektif mengelompokkan kondisi hidroponik ke dalam tiga kluster. Selain itu, integrasi Gemini API menghasilkan rekomendasi tindakan korektif berbahasa alami yang sesuai konteks dengan rata-rata latensi pemrosesan 12,10 detik. SmartHydro AI berhasil menyatukan pemantauan, analisis, dan rekomendasi dalam arsitektur hidroponik cerdas yang valid secara teknis. Kata kunci: smarthydro AI; IoT; machine learning;LLM; hidroponik cerdas