Electronic Theses and Dissertation

Prediksi konsumsi energi listrik merupakan aspek penting dalam manajemen sistem kelistrikan karena berperan dalam mendukung efisiensi perencanaan dan keandalan distribusi daya. pola konsumsi energi listrik di wilayah banda aceh bersifat fluktuatif dan non-linear, sehingga diperlukan metode prediksi yang mampu merepresentasikan karakteristik data tersebut secara akurat. penelitian ini bertujuan untuk mengembangkan model prediksi konsumsi energi listrik berbasis data historis menggunakan algoritma decision tree regression (dtr). data yang digunakan merupakan data konsumsi energi listrik dalam tiga skala waktu, yaitu jam-jaman, harian, dan mingguan, tanpa melibatkan variabel eksternal seperti suhu dan indikator ekonomi. metode penelitian meliputi pengumpulan dan praproses data, pemisahan data menjadi data pelatihan dan pengujian, pelatihan model dtr, serta evaluasi kinerja model menggunakan metrik mean absolute error (mae), mean squared error (mse), root mean squared error (rmse), dan koefisien determinasi (r²). hasil pengujian menunjukkan bahwa model dtr menghasilkan tingkat akurasi yang relatif tinggi pada seluruh titik pengamatan, dengan nilai r² berkisar antara 0,958 hingga 0,999. secara kuantitatif, model menghasilkan mae dan rmse masing-masing sebesar 59,745 dan 77,456 pada td 1 lambaro; 51,840 dan 72,689 pada td 2 lambaro; 279,183 dan 416,694 pada td 3 lambaro; 19,804 dan 26,345 pada gi jantho; serta 167,803 dan 347,474 pada gi ulee kareng. hasil ini menunjukkan bahwa model dtr memiliki performa sangat baik terutama pada unit dengan pola beban yang fluktuatif seperti td 1 dan td 2 lambaro, serta mampu mengikuti pola beban puncak secara adaptif dibandingkan regresi linier. dengan demikian, model dtr berbasis data historis terbukti efektif digunakan sebagai alat bantu perencanaan dan pengambilan keputusan operasional dalam pengelolaan distribusi energi listrik. kata kunci: prediksi konsumsi energi, decision tree regression, data historis, banda aceh, time-series.

Electricity consumption forecasting is a crucial aspect of power system management, as it supports efficient planning and reliable power distribution. The electricity consumption patterns in Banda Aceh exhibit fluctuating and non-linear characteristics, requiring forecasting methods capable of accurately representing such data behavior. This study aims to develop an electricity consumption forecasting model based on historical data using the Decision Tree Regression (DTR) algorithm. The dataset consists of electricity consumption data at three temporal resolutions hourly, daily, and weekly without incorporating external variables such as temperature or economic indicators. The research methodology includes data collection and preprocessing, division of the dataset into training and testing sets, DTR model development, and performance evaluation using Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and the coefficient of determination (R²). The experimental results indicate that the DTR model achieves relatively high accuracy across all observation points, with R² values ranging from 0.958 to 0.999. Quantitatively, the model obtained MAE and RMSE values of 59.745 and 77.456 for Lambaro TD 1; 51.840 and 72.689 for Lambaro TD 2; 279.183 and 416.694 for Lambaro TD 3; 19.804 and 26.345 for Jantho Substation; and 167.803 and 347.474 for Ulee Kareng Substation. These results demonstrate that the DTR model performs particularly well for units with highly fluctuating load patterns, such as Lambaro TD 1 and TD 2, and is able to adaptively capture peak load behavior more effectively than linear regression. Therefore, the historical data–based DTR model is proven to be an effective tool for supporting planning and operational decision-making in electricity distribution management. Keywords: Energy consumption prediction, decision tree regression, historical data, Banda Aceh, time-series.