The Intelligent kWh Export-Import Utilizing Classification Models for Efficiency in Hybrid PLTS

Abstract

Electricity demand is integral to the stability of the community's economic condition, where currently electricity is predominantly sourced from fossil fuels, posing limitations. One effort to maintain this stability is through the utilization of renewable energy, particularly solar energy. The abundance of solar energy in Indonesia presents an opportunity to maximize its potential. This study develops an intelligent kWh export-import system based on the Internet of Things (IoT) and integrated with machine learning. This integrated system enables smooth data flow and communication among various components. Sensors collect information from the solar panel, which is then transmitted to the ESP-32 microcontroller for preprocessing. The ESP-32 facilitates wireless connectivity by transmitting data to the Firebase data cloud using the MQTT protocol. Once stored in the cloud, the data undergoes further analysis and modeling, providing users with valuable insights into the performance of the solar panel system. Overall, this integrated approach allows for efficient monitoring and assessment of the system's efficiency and performance in real-time. Users can access real-time conditions via mobile based on three parameters: "current," "power," and "voltage." Machine learning is employed to classify conditions as "efficient" or "less efficient" by analyzing and comparing five different models: AdaBoost Classifier, DecisionTree Classifier, support vector machine (SVM), naïve Bayes classifier, and extra tree classifier. Model evaluation using accuracy percentage and F1-score indicates that the AdaBoost classifier exhibits high accuracy and F1-score values of 94.5% and 0.937, respectively.