Prediksi Umur Pakai Baterai dengan Model Random Forest Regressor pada Penggunaan Smart Meter Gas

Abstract

Gas bumi merupakan sumber daya alam strategis yang berperan penting dalam memenuhi kebutuhan energi di sektor industri, komersial, dan rumah tangga. Namun, distribusinya belum sepenuhnya merata, terutama bagi konsumen skala kecil. Untuk mengatasi permasalahan tersebut, digunakan teknologi smart meter berbasis Internet of Things (IoT) yang mampu mengukur dan mengirimkan data konsumsi energi secara berkala. Pada generasi sebelumnya perangkat belum memiliki kemampuan untuk memantau umur pakai baterai secara otomatis, sehingga menyulitkan proses pemeliharaan. Kemudian dilakukan penelitian untuk memprediksi umur pakai baterai lithium-ion 18650 dengan pendekatan machine learning. Dua algoritma regresi diterapkan, yaitu Random Forest Regressor dan Regresi Linier, yang dibandingkan berdasarkan tingkat akurasi prediksi. Hasil evaluasi model Random Forest Regressor menunujukkan bahwa nilai Mean Absolute Error (MAE) sebesar 0,6535, Mean Squared Error (MSE) sebesar 0,9887, dan Root Mean Squared Error (RMSE) sebesar 0,9943 yang kemudian diimplementasikan pada sistem smart meter gas.

Natural gas is a strategic natural resource that plays an important role in meeting energy needs in the industrial, commercial, and household sectors. However, its distribution is not yet fully equitable, especially for small-scale consumers. To address this issue, Internet of Things (IoT)-based smart meter technology is used to measure and transmit energy consumption data on a regular basis. In previous generations, devices lacked the ability to automatically monitor battery lifespan, complicating maintenance processes. A study was conducted to predict the lifespan of 18650 lithium-ion batteries using a machine learning approach. Two regression algorithms were applied: Random Forest Regressor and Linear Regression, which were compared based on prediction accuracy. The evaluation results of the Random Forest Regressor model showed a Mean Absolute Error (MAE) of 0.6535, a Mean Squared Error (MSE) of 0.9887, and a Root Mean Squared Error (RMSE) of 0.9943, which were then implemented in the gas smart meter system.