Kajian Strategi Pemodelan Time Series untuk Peramalan Penjualan Tenaga Listrik : Studi Kasus PT PLN (Persero

Dinaputri, Ryanti Setyoningtyas (2026) Kajian Strategi Pemodelan Time Series untuk Peramalan Penjualan Tenaga Listrik : Studi Kasus PT PLN (Persero. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6010241101-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Penelitian ini membahas pengembangan peramalan penjualan tenaga listrik menggunakan time series analysis di PT PLN (Persero) dengan mempertimbangkan variabel baru yang relevan yaitu iklim melalui suhu udara dan electric lifestyle melalui adaptasi penggunaan alat rumah tangga berbasis tenaga listrik. Latar belakang penelitian berawal dari ditemukannya selisih antara rencana penjualan tenaga listrik pada dokumen Rencana Usaha Penyediaan Tenaga Listrik (RUPTL) dengan realisasi penjualan listrik yang berpotensi menimbulkan risiko strategis berupa oversupply, tingginya reserve margin dan kerugian finansial bagi perusahaan. Untuk menjawab permasalahan tersebut, penelitian ini mengembangkan model analisis berbasis time series. Analisis time series digunakan untuk memprediksi penjualan tahun mendatang berdasarkan pola data historis yang turut mempertimbangkan exogenous variable. Pengolahan data dilakukan menggunakan pendekatan agregasi nasional, agregasi regional dan agregasi per UID dan UIW melalui beberapa metode yang relevan meliputi Panel Linear Model (PLM), Linear Model (LM), Holt’s Double Exponential Smoothing dan algoritma berbasis machine learning yaitu XGBoost dan Random Forest. Selanjutnya akan dilakukan uji performansi model menggunakan metrik Mean Absolute Percentage Error (MAPE) dan Root Mean Square Error (RMSE). Secara keseluruhan, penelitian ini membuktikan bahwa pendekatan lingkup agregasi per UID dan UIW menunjukkan performansi peramalan yang lebih baik dibandingkan lingkup nasional dan regional, secara spesifik metode Holt's Double Exponential Smoothing mampu menekan tingkat kesalahan hingga senilai 2,63% disusul oleh model LM lag 1 senilai 3,11% dan LM senilai 3,25%. Selain itu, metode LM secara empiris menemukan bahwa fluktuasi suhu udara rata-rata dan transisi gaya hidup masyarakat menuju electric lifestyle kini menjadi variabel krusial di beberapa unit yang diteliti, mendampingi variabel makroekonomi tradisional.
=====================================================================================================================================
This study discusses the development of electricity sales forecasting using time series analysis at PT PLN (Persero) by considering relevant new variables, namely climate that represented by air temperature and electric lifestyle that represented by the adaptation of electric-based household appliances. The research was initiated by the discovery of discrepancy between electricity sales plan in the RUPTL document and the actual electricity sales, which potentially poses strategic risks such as oversupply, high reserve margins, and financial losses for the company. To address these problems, this study develops time series-based analytical models. Time series analysis is utilized to predict future sales based on historical data patterns while also incorporating exogenous variables. Data processing is conducted using national aggregation, regional aggregation, and aggregation per UID and UIW approaches through several relevant methods, including the Panel Linear Model (PLM), Linear Model (LM), Holt's Double Exponential Smoothing, and machine learning-based algorithms, namely XGBoost and Random Forest. Furthermore, model performance will be evaluated using the Mean Absolute Percentage Error (MAPE) and Root Mean Square Error (RMSE) metrics. The findings clearly show that forecasting at the localized UID and UIW levels significantly outperforms broader national and regional approaches. Holt's Double Exponential Smoothing stood out in particular, successfully driving the error rate down to just 2.63%. The localized linear models also performed exceptionally well, with LM Lag 1 and the standard LM recording error rates of 3.11% and 3.25%, respectively. Additionally, the linear modeling empirically confirmed an important shift: daily temperature fluctuations and the transition toward an electric lifestyle are now crucial drivers of electricity demand in several areas, working right alongside traditional macroeconomic indicators.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Peramalan, electric lifestyle, iklim, time series, Forecasting, climate, electric lifestyle, time series.
Subjects:	H Social Sciences > HA Statistics > HA30.3 Time-series analysis H Social Sciences > HD Industries. Land use. Labor > HD30.27 Business forecasting H Social Sciences > HD Industries. Land use. Labor > HD30.28 Planning. Business planning. Strategic planning.
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Industrial Engineering > 26101-(S2) Master Thesis
Depositing User:	Ryanti Setyoningtyas Dinaputri
Date Deposited:	27 Jul 2026 07:49
Last Modified:	27 Jul 2026 07:49
URI:	http://repository.its.ac.id/id/eprint/137887

Actions (login required)

View Item