Renewable Energy in Smart Grid: Photovoltaic Power Monitoring System Based on Machine Learning Using an Open-Source IoT Platform

Abstract

This paper proposes a PV monitoring system based on the Internet of Things, which is the best solution to monitor our system. In this system, we have utilized the Raspberry Pi card as a server that communicates with a Node Mcu (ESP32) as a client using the MQTT and HTTP protocols. Our solution is composed of two steps; the first step consists of power measurement. The voltage constant and current were measured by the ACS712 sensor, and we measured the power and energy of the solar panels every 5 min. A DHT sensor was utilized to measure the temperature of the solar panels. These measures will be shown on the Node-red platform and stored as a database in the SQLite programming language; SQLite is introduced to reduce the database complexity. A Raspberry Pi card recorded this database in real time using WiFi or cable Ethernet; with this database, we can make accurate projections about the efficiency of solar panels and command our system. However, in the second step, based on this database, we involve the command in our system by the best method (algorithm) of prediction for our case. The power delivered by solar panels was predicted with the use of machine learning (a model decision tree), which enabled us to generate forecasts. In practice, we use an electronic card that can support this type of machine learning algorithm; for this, we used the raspberry pi card. Node- red is the most suitable interface to apply this algorithm, and it allows us to monitor all measurements by the dashboard in real time with a laptop (WiFi) and with a smartphone (4G). As a result of this work, we have made a smart system based on machine learning that allows us to integrate PV into the smart grid. This approach allows us to manage our system in a more efficient, automated, and intelligent manner.