<![CDATA[Many countries have adopted hybrid renewable power generation systems, i.e. combination of Solar Power Plants (SPP) and Wind Power Plants (WPP), due to the increasing global demand for electricity. The inverter which converts direct current (DC) into alternating current (AC) is the most important part of these systems. Inverter faults such as overvoltage, overcurrent, and abnormal switching, if not detected, can seriously degrade or even shut down system performance. Existing studies on inverter fault detection are predominantly based on large-scale PV-only systems using complex methods such as LSTM or power curve analysis, leaving a gap for lightweight, laboratory-scale hybrid SPP-WPP implementations. The novelty of this work lies in three aspects not addressed simultaneously in prior literature at laboratory scale: (1) the integration of MQTT-based IoT communication with an industrial Modbus RTU protocol in a single unified architecture, (2) the application of a duration-gated threshold detection method validated on a real hybrid SPP-WPP object rather than a PV-only system, and (3) end-to-end experimental validation covering sensor accuracy, communication latency, and fault detection rate across 40 structured injection trials. This paper presents the design, implementation, and experimental validation of a real-time SCADA-based supervision framework integrating MQTT-based IoT communication and an industrial Modbus RTU protocol for early inverter fault detection in a laboratory-scale SPP-WPP hybrid power plant at Politeknik Negeri Manado. The system consists of a microcontroller ESP32, voltage sensor ZMPT101B, current sensor ACS712 and power module PZEM-004T communicating over Modbus RTU and MQTT protocol. The data is processed on a Python FastAPI backend and shown in real-time on a React.js web dashboard. The fault detection deliberately employs a duration-gated threshold monitoring approach on voltage (220V +/-10%), current (>10A sustained for >=500 ms) and power fluctuation (>+/-15% for >=1 s), chosen specifically for its computational lightness on the ESP32 microcontroller, deterministic response time, and pedagogical transparency in an academic laboratory context where operators and students must be able to trace the exact cause of every alarm. Experimental results across 40 structured fault injection trials show sensor measurement deviation below 2%, dashboard communication latency averaging 0.87 seconds, and a fault detection rate of 95% (38 out of 40 injected fault events correctly detected, covering overvoltage, undervoltage, overcurrent, and power fluctuation). This result is competitive with LSTM-based approaches (92%) and power curve analysis methods (93%) reported in recent literature, while delivering faster response and simpler implementation. The current architecture is designed as a foundation for future enhancement toward Edge AI and Machine Learning-based predictive maintenance, where the stored historical data from this system can serve directly as a training dataset. This framework provides a practical, replicable, and academically transparent baseline for SCADA integration in hybrid renewable energy laboratory environments.]]>
