Grid Sense AI AI-Based Fault Detection in Electrical Systems

Abstract

Power distribution systems should also have the ability to detect electrical faults so that they may keep their systems running everlastingly. Modern current protection devices such as fuses and circuit breakers do not have remote monitoring . This paper is a design and development project of a smart fault detector called Grid Sense AI, which is designed around the Internet of Things (IoT). It measures electrical parameters in real-time and categorizes faults with the help of an Artifcial Neural Network (ANN), realised directly on the ESP32 microcontroller. Hardwarewise, this system consists of a main isolation transformer (220 V or 110 V), three secondary step-down transformers that are used to represent distribution branches, a PZEM-004T power sensor module, a voltage sensor module (ZMPT101B) , a current sensor module, a relay module with four relays, a 5V AC-DC power supply module, and an ESP The system measures the voltage and current on-the-fly and classifes the electrical condition with an ANN as one of fve statewide, operating normally, overvoltage, undervoltage, over-current or short circuit fault. A fault condition is transmitted to the Blynk IoT cloud service via the ESP32 microcontroller and Wi-Fi in the event of fault detection and the faulty branch is automatically de-energized by the relay whilst at the same time the operator is informed by a push notifcation in a mobile application. ANN model was developed with TensorFlow/Keras in Python and was trained on the data gathered in the prototype, exported in the TensorFlow Lite format [6], and executed the on-device inference in the ESP32 without cloud-dependency. The outcomes of the experiment confrm that the prototype is capable of processing all the four kinds of faults and provides real-time messages, thus becoming a promising option instead of the conventional method.