Rancang Bangun Sistem Monitoring Nutrisi Dan Pencahayaan Pada Tanaman Hidroponik Berbasis IOT Dengan Integrasi Api Cuaca Untuk Penghematan Konsumsi Energi

Saputra, Aji Wahyu Iva (2026) Rancang Bangun Sistem Monitoring Nutrisi Dan Pencahayaan Pada Tanaman Hidroponik Berbasis IOT Dengan Integrasi Api Cuaca Untuk Penghematan Konsumsi Energi. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5009221086-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (5MB) | Request a copy

Abstract

Budidaya hidroponik sangat sensitif terhadap fluktuasi parameter lingkungan seperti suhu, kelembapan, intensitas cahaya, dan nutrisi (TDS/EC), sehingga pengelolaan aktuator secara manual sering tidak efisien dan boros energi, terutama pada pengaturan pencahayaan yang dipengaruhi variabilitas cuaca harian. Penelitian ini merancang dan membangun sistem monitoring terpadu berbasis Internet of Things (IoT) untuk instalasi hidroponik Nutrient Film Technique (NFT) dan aeroponik, yang mengintegrasikan tujuh node mikrokontroler ESP32 dengan protokol MQTT dan dashboard Node-RED, ditambah integrasi data eksternal dari API cuaca BMKG, serta didukung sumber energi hybrid photovoltaik (PV) dan turbin angin. Metode penelitian meliputi perancangan hardware sensor (DHT22, BH1750), sensor kualitas air) dan aktuator (lampu LED, paranet otomatis, kipas DC) berbasis relay, serta pemodelan Extreme Learning Machine (ELM) dengan konfigurasi 800 hidden neuron, fungsi aktivasi sigmoid, dan regularisasi Ridge Regression untuk forecasting suhu dan intensitas cahaya sebagai dasar pengambilan keputusan smart switching menggunakan logika kontrol hysteresis dual-threshold guna mencegah relay chattering. Pada tahap pengujian model, ELM mampu memprediksi suhu dengan akurasi tinggi (akurasi 99,07%; MAPE = 0,93%; R² = 0,9942) dan intensitas cahaya dengan akurasi baik (akurasi 92,48%; MAPE = 7,52% pada data siang; R² = 0,9744). Pengujian forecasting autoregressive 24 jam ke depan menunjukkan performa yang menurun dibandingkan tahap pengujian model, dengan suhu memperoleh MAPE = 4,40% dan R² = 0,83, sedangkan intensitas cahaya memperoleh MAPE = 37,84% dan R² = 0,96, akibat akumulasi error khas pendekatan recursive forecasting. Data cuaca eksternal BMKG dievaluasi memiliki tingkat kesesuaian sedang terhadap data sensor lokal (R² = 0,523 untuk suhu; R² = 0,489 untuk kelembapan) sehingga tidak digunakan sebagai fitur masukan model ELM, melainkan hanya sebagai informasi pendukung dashboard. Performansi pengiriman data IoT melalui protokol MQTT menunjukkan rata-rata delay sebesar 22,31 ms dan packet loss sebesar 1,90%, keduanya tergolong kategori Sangat Bagus. Penerapan smart switching berbasis hasil forecasting berhasil menurunkan konsumsi energi aktuator dibandingkan skenario operasi terus-menerus (always-ON), dengan penghematan sebesar 29,5% pada kipas DC, 17,4% pada lampu LED, dan 83,3% pada paranet otomatis, serta menurunkan konsumsi energi harian aktual sistem sebesar 16,4% (dari rata-rata 1,25 kWh/hari menjadi 1,045 kWh/hari) pada hari penerapan, tanpa mengganggu kestabilan parameter lingkungan tanaman. Penelitian ini membuktikan bahwa integrasi sensor lokal, model prediksi machine learning, dan logika kontrol hysteresis dapat meningkatkan Penghematan energi sekaligus mendukung implementasi pertanian cerdas (smart farming) yang berkelanjutan
===================================================================================================================================
Hydroponic cultivation is highly sensitive to fluctuations in environmental parameters such as temperature, humidity, light intensity, and nutrient concentration (TDS/EC), making manual actuator management inefficient and energy-intensive, particularly for lighting control affected by daily weather variability. This study designs and develops an integrated Internet of Things (IoT)-based monitoring system for Nutrient Film Technique (NFT) and aeroponic hydroponic installations, integrating seven ESP32 microcontroller nodes through the MQTT protocol and a Node-RED dashboard, combined with external weather data from the BMKG API, and supported by a hybrid photovoltaic (PV) and wind turbine energy source. The research method includes hardware design for sensors (DHT22, BH1750, water quality sensors) and relay-based actuators (LED lighting, automatic shading net, DC fan), as well as Extreme Learning Machine (ELM) modeling with a configuration of 800 hidden neurons, sigmoid activation function, and Ridge Regression regularization for temperature and light intensity forecasting as the basis for smart switching decisions using a dual-threshold hysteresis control logic to prevent relay chattering. During the model testing phase, the ELM model achieved high accuracy in temperature prediction (accuracy 99.07%; MAPE = 0.93%; R² = 0.9942) and good accuracy in light intensity prediction (accuracy 92.48%; MAPE = 7.52% on daytime data; R² = 0.9744). A 24-hour-ahead autoregressive forecasting test showed reduced performance compared to the model testing phase, with temperature achieving MAPE = 4.40% and R² = 0.83, while light intensity achieved MAPE = 37.84% and R² = 0.96, due to error accumulation typical of recursive forecasting approaches. External BMKG weather data was evaluated to have only moderate agreement with local sensor data (R² = 0.523 for temperature; R² = 0.489 for humidity), and was therefore not used as an input feature for ELM model training, serving instead only as supporting information on the dashboard. IoT data transmission performance via the MQTT protocol showed an average delay of 22.31 ms and a packet loss ratio of 1.90%, both categorized as Very Good. The implementation of forecasting-based smart switching successfully reduced actuator energy consumption compared to an always-ON scenario, achieving savings of 29.5% for the DC fan, 17.4% for LED lighting, and 83.3% for the automatic shading net, and reduced the system's actual daily energy consumption by 16.4% (from an average of 1.25 kWh/day to 1.04 kWh/day) on the day of implementation, without compromising the stability of plant environmental parameters. This research demonstrates that integrating local sensors, machine learning prediction models, and hysteresis control logic can improve energy efficiency while supporting the implementation of sustainable smart farming systems

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Extreme Learning Machine, Hidroponik, Internet of Things, Smart switching ============================================================ Extreme Learning Machine, hydroponics, Internet of Things, smart switching
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK1087 Photovoltaic power generation T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5103.8 Switching systems
Divisions:	Faculty of Industrial Technology > Physics Engineering > 30201-(S1) Undergraduate Thesis
Depositing User:	Aji Wahyu Iva Saputra
Date Deposited:	03 Aug 2026 03:34
Last Modified:	03 Aug 2026 03:34
URI:	http://repository.its.ac.id/id/eprint/142083

Actions (login required)

View Item