PENGGUNAAN MACHINE LEARNING DALAM

PROYEKSI KONSUMSI BBM BIO SOLAR SUBSIDI DI

INDONESIA

Atiq Mujtaba
2206013003
PROPOSAL PENELITIAN

FAKULTAS TEKNIK
PROGRAM MAGISTER TEKNIK
DEPARTEMEN TEKNIK INDUSTRI
2023
 Latar Belakang Masalah
Konsumsi Energi di Indonesia Tahun 2022 (Million BOE)

550.00 42.50%
40.30%
37.50%
450.00 32.50%
350.00 27.50%
25.24%
22.50%
250.00 17.50%
15.47% 6.16%
150.00 12.50%
6.29%
6.06% 0.38% 0.02% 0.00% 7.50%
50.00 0.08%
2.50%
Oil Fuel Coal Electricity Gas LPG Traditional Industrial Solar water BioGas Direct Use Oil Fuel Coal Electricity Gas LPG Traditional Industrial Solar water BioGas Direct Use
Biomass Biomass Heater of Biomass Biomass Heater of
Geotherm Geotherm

C 477.82 299.19 183.41 74.58 72.99 71.9 4.52 0.94 0.21 2E-31
o
n
s
u
m
p
t
i
o
n

(
M
i
l
l Consumption (Million BOE) Percentage
i
o
n

B
O
E
)
P 40.30% 25.24% 15.47% 6.29% 6.16% 6.06% 0.38% 0.08% 0.02% 0.00%
e
r
c
e
n
t
a
g
e

Konsumsi Energi di Indonesia tertinggi didominasi oleh energi jenis bahan

bakar Minyak sebesar 40,30%. Disusul oleh batu bara sebesar 25,24%.
(HEESI KESDM, 2022)
 Latar Belakang Masalah
Konsumsi Energi Per Sektor Tahun 2022
45.11%
550
36.15% 42.50%
450
32.50%
350
250 22.50%
13.62%
150 12.50%
4.19% 0.93%
50 2.50%
Industry Transportation Household Commercial Other Sector Industry Transportation Household Commercial Other Sector

Con- 534.75 428.61 161.48 49.65 11.07

sump-
tion
(Mil-
lion
BOE)
Per- 45.11% 36.15% 13.62% 4.19% 0.93%
centage

Consumption (Million BOE) Percentage

Konsumsi energi digunakan di berbagai sektor, antara lain industri, rumah tangga,
komersial, dan transportasi. Sektor industri merupakan penyumbang terbesar yaitu
sebesar 45,11% dari total konsumsi energi. Disusul transportasi sebesar 36,15% (HEESI
KESDM, 2022)
KEBIJAKAN JENIS-JENIS BBM
Berdasarkan Perpres 191/2014, PMK 16/2021, PMK 193/2017 dan
Perubahannya
SUBSIDI
Diberikan Subsidi dan Alokasi anggaran yang
kompensasi
ENERGI
diberikan kepada
perusahaan atau
 Minyak Tanah (Kerosene) lembaga yang
Jenis BBM Tertentu menyediakan dan
 Minyak Bio Solar (Gas Oil
(JBT) mendistribusikan BBM
CN 48) JBT, Liquefied Gas for
Vehicle (LGV), LPG,
(subsidi dan kompensasi)
untuk konsumsi rumah
Diberikan tangga dan usaha mikro
Kompensasi serta tenaga listrik
Jenis BBM Khusus sehingga harga jualnya
Bensin (Gasoline) terjangkau oleh
Penugasan (JBKP) RON 90 masyarakat yang
JENIS membutuhkan.
BBM
TMT 1 Januari 2022 KOMPENS
ASI
Dana yang dibayar oleh
Tidak Diberikan Subsidi dan Pemerintah ke Badan
Usaha (BU) atas
Kompensasi kekurangan penerimaan
Seluruh jenis BU akibat selisih atas
Jenis BBM Umum BBM diluar JBT Harga Jual Eceran (HJE)
(JBU) dan JBKP* BBM/Tarif Tenaga Listrik
(TTL) Non Subsidi yang
ditetapkan Pemerintah,
* a.l Avtur, Avgas, Bensin (RON 92, 95,98), Minyak Solar dengan HJE/TTL
(CN 48, CN 51, CN 53), Minyak Diesel, Minyak Bakar dihitung sesuai UU.
 Kuota VS Realisasi Bio Solar Subsidi (Juta KL) • Penyaluran Bio Solar tahun 2020 turun karena
20
17.83
covid19. tahun 2021 dan 2022 realisasi vs kuota
18

16
17.05
15.5 15.5 15.62
16.23
15.31
15.8
17
hampir mencapai 100%, sedangkan tahun 2019
mengalami kelebihan kuota.
17.59

14 15.58 15.6
14.51 14.5
12
14.16
13.75
14
9.91
• Dampak over kuota adalah badan usaha
10 penugasan tidak dibayarkan subsidi.
8
• Pada tahun 2022, sebagian besar subsidi energi
6 Over
dialokasikan untuk subsidi dan kompensasi BBM.
4
Kuota
2 subsidi dan kompensasi BBM (53%), Gas LPG (27%)
0
2015 2016 2017 2018 2019 2020 2021 2022 Juli 2023
dan listrik (20%) (HEESI KESDM, 2022).
Kuota Realisasi • Proyeksi yang akurat diharapkan dapat
meningkatkan efisiensi belanja subsidi bahan
bakar, dan meminimalisir kerugian Badan Usaha
penugasan untuk tidak dibayarkan subsidi karena
over kuota.
 Literature
Review What method
What is the have been used to
What are significant
topic of study the
issues and debates about
No Tittle interest in questions and independent variables
the topic in this paper?
this paper? problems in this
paper?

Modelling industry energy forecasting energy the accuracy of these forecasts can The method used in this 1. Population
1. demand using multiple demand in the be influenced by various factors study is the multiple linear
2. Price Index
linear regression industrial sector such as economic growth, regression model.
using multiple linear population growth, technological 3. Purchasing Power Parity
analysis based on consumed
regression models changes, and energy prices
quantity of goods 4. Gross Domestic Product,
(Maaouane, et.Al, 2021)
5. Household Final Consumption
Expenditure

Energy demand forecasting Analysis of the The paper identifies several drivers Methode: Variable independent
2 in China: A support vector drivers of energy of energy demand, including
1. (Lasso-RF) 1. Economic Development
demand and the economic development level,
regression-compositional model (Regression ) Level;
prediction of energy population, science and technology,
data second exponential 2. (SVR-CDSES) 2. Population;
consumption in energy policies, related products,
smoothing model (Congjun model, 3. Science and Technology;
China and resources.
Rao, et Al, 2023) 4. Energy Policies;
5. Related products
 Literature
Review
What is
What method have been
the topic What are significant
used to study the
of interest issues and debates about
No Tittle questions and problems in independent variables
in this the topic in this paper?
this paper?
paper?

Energy demand forecasting The challenge of accurately The machine learning algorithms Independent variable:
3. forecasting in seven energy demand predicting energy demand used include ANN, AR, ARIMA,
1. population growth,
sectors by an in different SARIMA, SARIMAX, and LSTM.
optimization model sectors in Iran up 2. economic development,
to 2040. The difficulty of addressing the lack
based on machine 3. energy rates
of specific data types, such as
learning algorithms
population and GDP statistics, in
(Majid et. Al,2023)
energy demand prediction
Gasoline and diesel Forcasting two key issues and debates: Estimating the elasticities of The independent variables in
4 demand in the EU: gasoline and gasoline and diesel demand using this paper include price, income,
1.variability of gasoline and diesel
Implications for the diesel demand in the Autoregressive Distributed Lag vehicle stock, and trend
demand elasticities across studies
2030 the EU-28 (ARDL) (Regression) consumption gasoline and diesel.
and nations.
countries.
emission goal 2.the study explores the
(Abenezer, 2020) effectiveness of fuel tax policies in
reaching emission reduction goals
 Literature
Review
What is the topic What are significant issues What method have been used to
of interest in this and debates about the topic study the questions and
No Tittle independent variables
paper? in this paper? problems in this paper?

Supply and demand The paper primarily Understanding the major methodologies such as two-stage independent variable include
5. socio-economic and
of biofuels in the fuel focuses on examining demographic factors that affect least square (2SLS), three-stage socio-economic factors (e.g., real
market of Thailand: the factors affecting biofuel production and least square (3SLS), and Artificial GDP, employment, number of
Two the supply and consumption Neural Networks (ANNs) to vehicles, number of biofuel
demand of biofuels in The need to develop accurate analyze the long-run biofuel stations), cost and substitution
stage least square long-run supply and demand
the fuel market of models for biodiesel and demand and supply. factors (e.g., price of oil product
and three least square
Thailand bioethanol in Thailand, and to and price of biofuel)
approaches identify the most suitable
(Anuman, et. Al, factors affecting these [7].
2016)

Prediction of predicting The need for accurate Multivariate Adaptive Regression GDP, oil price, population,
6. transportation energy transportation energy prediction of transportation Splines (MARS). This method is a vehicle-km, ton-km, and
demand: Multivariate demand in Turkey energy demand form of non-parametric regression passenger-km, the number of
Adaptive using the Multivariate that is flexible and fast, and can vehicles, import, export, and
Adaptive Regression handle both linear and non-linear income are also mentioned
Regression Splines
Splines (MARS) modeling
(Sahraei, et. Al,
method
2021)
 Literature
Review What method have
What are significant issues and been used to study
What is the topic of
debates about the topic in this the questions and
No Tittle interest in this paper? independent variables
paper? problems in this
paper?

Forecasting forecasting automobile The paper emphasizes the need to machine learning (ML) The independent variables
7. automobile gasoline demand in Australia incorporate economic and behavioral applied to forecast
1. Gross domestic product
gasoline demand using machine learning-based theory into machine learning models. automobile gasoline (GDP)
in Australia using regression. While machine learning techniques can demand in Australia
machine provide accurate predictions, they with support vector 2. Population
should be guided by economic intuition regression (SVR) and 3. Oil price
learning-based
and implications LASSO regression
regression (Zheng 4. Number of vehicles
Li, et. Al,2022)
5. Passenger transport
volumes
Forecasting forecasting of automobile One significant issue and debate is the Different Time Series The independent variables
8. automobile petrol petrol demand in Australia. selection of the most appropriate Model, Autoregressive
Gross Domestic Product
demand in The paper evaluates different forecasting model for automobile petrol integrated moving
(GDP)
Australia: An models and techniques for demand. There is a debate about average (ARIMA)
evaluation predicting petrol consumption whether simpler statistical models or model Real Petrol Price (RPP)
and examines the factors more complex models are more
of empirical Total Petrol Consumption
influencing petrol demand, effective in accurately predicting petrol
models (Zheng Li, (TPC)
such as price and income demand
et. Al,2010)
elasticities.
 Literature
Review What method
have been used
What is the What are significant issues and
to study the
topic of interest debates about the topic in this
No Tittle questions and independent variables
in this paper? paper?
problems in this
paper?

Modeling and Forecasting forecasting The paper debates the difficulty in Multiple Linear Annual gasoline
9 consumption, prices,
Gasoline Consumption in gasoline demand developing reliable forecasting regression
Cameroon using Linear in cameroon models due to the challenge of vehicle fleet, GDP, and GDP
per capita
Regression Models deciding which variables to use as
(Sapnken, et. Al, 2021) independent variables. If the input
data is insufficient or redundant,
the outcome will be poor
Modeling of the energy Forecasting focused on the development and multiple linear population, GDP,
10 demand of the residential demand energy comparison of energy-demand regression and household size, median
sector in the US models, specifically multiple linear artificial neural household income, and
in the United States using
regression models and regression (MLR) and artificial network energy costs
artificial neural neural network (ANN) models, for
Networks (Kialashaki, et. predicting future energy demand
Al, 2013) in the residential sector of the
United States until 2030
Perumusan Masalah
• Penelitian ini membandingkan 3 metode yang paling sering digunakan
dalam proyeksi demand bbm yaitu regressi model, Model Lasso, dan
ARIMA model (time series).
• Penelitian ini akan menganalisis pengaruh faktor ekonomi dan sosial
terhadap realisasi JBT Bio Solar Subsidi. Faktor ekonomi yang dipilih
didasarkan pada tinjauan literatur.
Tujuan Penelitian
1. Penelitian ini bertujuan untuk mengukur pengaruh dari
berbagai indikator ekonomi terhadap realisasi BBM JBT
Bio Solar Subsidi.
2. Meramalkan konsumsi BBM JBT Bio Solar subsidi

Batasan Penelitian
1. Data konsumsi BBM Bio Solar Subsidi Nasional dari tahun
2014-2022 dengan data triwulanan/bulanan.
2. Data Variabel Independen data triwulanan/ bulanan dari
tahun 2014-2022
3. Data pada tahun terdampak covid pada tahun 2020 tidak
dipakai karena penurunan konsumsi bbm yang signifikan
 Tahapan Penelitian
Menentukan variabel dependen dan independen

Mengumpulkan data

Melakukan uji asumsi klasik

Beberapa tahapan
yang perlu dilakukan Melakukan estimasi model
yaitu:
Mengevaluasi model

Membuat proyeksi

Melakukan validasi model

 Tahapan Penelitian
Menentukan variabel independen
Mengumpulkan data
• Penelitian ini fokus pada hubungan realisasi BBM JBT
Bio Solar Subsidi dengan variable independen yang
berpengaruh: • Studi ini menggunakan data primer
1.Pengeluaran Rumah Tangga maupun data sekunder. Data primer
2.Belanja Pemerintah
3. PKP (Pertanian, Kehutanan, Perikanan)
adalah data yang dimiliki BPH Migas
4.TP (transportasi dan Pergudangan) dan bersumber dari laman SILVIA
5.JKS (Jasa Kesehatan dan kegiatan sosial BPH Migas serta Kementerian
6.KRT (Konsumsi rumah tangga) ESDM. Data sekunder juga
7.IHK (Index Harga Konsumen)
8.Jumlah Mobil Penumpang digunakan sesuai kebutuhan studi
9.Jumlah Bus yang berasal dari berbagai instansi
10.Jumlah Truk Pemerintah antara lain Badan Pusat
11.PDB
12.Harga solar vs pertamina dex
Statistik (BPS), Pertamina, dan BI.
 Pengolahan Data
Pengolahan data menggunakan python untuk model Multiple regresi linear dan Model Lasso.
Uji normalitas untuk variable
independent
Berdasaran tabel nilai statistik dari uji Kolmogorov-Smirnov
hasil statistic dibanding dengan nilai tabel KS N=35, α = 0,05
adalah 0,224, dan P-Value > 0,05 terdistribusi normal.

Sehingga semua data variable independen terdistribusi

normal.
 Multiple Regressi Linear
dropping variables untuk variable independen sehingga
Uji multikolinearitas untuk variable independen dan didapatkan variable yang akan digunakan dalam model
dependen dengan correlation matrix Regresi linear sebagai berikut: Belanja Pemerintah, PKP
(Pertanian, Kehutanan, Perikanan), JKS (Jasa Kesehatan dan
kegiatan sosial, Index Harga Konsumen, Penjualan Solar
Subsidi, dan Harga Solar vs Pertamina Dex
 Multiple Regressi Linear
uji homoskedastisitas
uji autokorelasi

autocorrelation: 2.2177934467777924

 residual tidak memiliki korelasi

Dari grafik diatas bahwa tidak ada pola dari residual tersebut, yang
berarti data residual ini bersifat baik
 Multiple Regressi Linear

Hasil Prediksi Regresi Linier

r2 score metode linear regresi =

0.9498284989571328

mape score metode linear regresi =

2.484387449127996%
 9.31708066e+00, 2.70210710e+01, 4.81691540e+00, -3.59073281e+03, -4.80093130e+01, -2.73089939e+06])

Model Regressi Lasso

Variabel independen Nilai Koefisien Variabel independen Nilai Koefisien
9.68632798e-
Pengeluaran RT
03 9.31708066e
TP (transportasi dan Pergudangan)
+00
3.28419695e-
Belanja Pemerintah JKS (Jasa Kesehatan dan kegiatan 2.70210710e
03
DROP sosial +01
PKP (Pertanian, Kehutanan, 4.97152034e- Variable
Perikanan) 01 4.81691540e
KRT (Konsumsi rumah tangga)
+00
2.27618385e
TP (transportasi dan Pergudangan)
+01 -
IHK (Index Harga Konsumen untuk
3.59073281e
JKS (Jasa Kesehatan dan kegiatan 1.23750668e inflasi)
+03
sosial +01
-
6.78859227e
KRT (Konsumsi rumah tangga) Jumlah Bus 4.80093130e
+00
+01
-
IHK (Index Harga Konsumen untuk
1.41846739e -
inflasi) solar vs pertadex 2.73089939e
+04
+06
-
 Model Regressi Lasso
uji homoskedastisitas
uji autokorelasi

autocorrelation: 1.049582074433898

 residual tidak memiliki korelasi

Dari grafik diatas bahwa tidak ada pola dari residual tersebut, yang
berarti data residual ini bersifat baik
 Model Regressi Lasso

Hasil Prediksi Regresi Lasso

r2 score metode regresi lasso=

0.9534346494919743

mape score metode regresi lasso=

0.7429182256821113%
 Model ARIMA
Pengolahan data menggunakan python untuk model ARIMA
 Model ARIMA
ACF & PACF Plot

Uji ACF dan PACF yaitu untuk menguji korelasi data dari
suatu periode dengan periode sebelumnya
 Model ARIMA
Hasil Prediksi ARIMA model

ARIMA Model MAPE: 3.7181745701391926%

 Kesimpulan

Bio Solar Subsidi

Regresi Linier Lasso ARIMA
R Square MAPE R Square MAPE MAPE
0,9498 2,484% 0,9534 0,743% 3,718%

Berdasarkan hasil MAPE dari model Regressi linear, Regressi Lasso

dan Model Arima, dapat disimpulkan bahwa model Lasso memiliki
nilai MAPE terbaik dengan nilai 0,743%, disusul Regressi linear
dengan nilai 2,48% dan ARIMA dengan nilai 3,718%.
 Timeline penelitian

Kegiatan November Desember

2023 2023
Januari
2024
Februari
2024
Maret
2024
April
2024

Studi Literatur
Pengumpulan
Data
Pengolahan Data
Analisis Data dan
estimasi model
Validasi Model
Penyusunan
naskah jurnal
 daftar Pustaka.
• Kementerian ESDM Indonesia. (2022). Handbook of Energy & Economic Statistics of Indonesia 2022. Jakarta.
• BPH Migas.(2022). Laporan Kinerja BPH Migas
• Aklilu, A. Z. (2020). Gasoline and diesel demand in the EU: Implications for the 2030 emission goal. Renewable and Sustainable
Energy Reviews, 118. https://doi.org/10.1016/j.rser.2019.109530
• Chanthawong, A., Dhakal, S., & Jongwanich, J. (2016). Supply and demand of biofuels in the fuel market of Thailand: Two stage
least square and three least square approaches. Energy, 114, 431–443. https://doi.org/10.1016/j.energy.2016.08.006
• Handra, N., & Hafni, H. (2017). Effect of Binder on Combustion Quality on EFB Bio-briquettes. IOP Conference Series: Earth and
Environmental Science, 97(1). https://doi.org/10.1088/1755-1315/97/1/012031
• Kim, Y. (2013). Some recent empirical developments regarding the bio-industry: Using the Granger-causality test with Indonesia’s
palm oil data. International Journal of Bio-Science and Bio-Technology, 5(5), 13–24. https://doi.org/10.14257/ijbsbt.2013.5.5.02
• Kristyadi, T., Permana, D. I., Sirodz, M. P. N., Saefudin, E., & Farkas, I. (2022). Performance and Emission of Diesel Engine Fuelled
by Commercial Bio-Diesel Fuels in Indonesia. Acta Technologica Agriculturae, 25(4), 221–228. https://doi.org/10.2478/ata-2022-
0032
• Li, Z., Zhou, B., & Hensher, D. A. (2022). Forecasting automobile gasoline demand in Australia using machine learning-based
regression. Energy, 239. https://doi.org/10.1016/j.energy.2021.122312
• Maaouane, M., Zouggar, S., Krajačić, G., & Zahboune, H. (2021a). Modelling industry energy demand using multiple linear
regression analysis based on consumed quantity of goods. Energy, 225. https://doi.org/10.1016/j.energy.2021.120270
 daftar Pustaka (lanjutan)
• Rao, C., Zhang, Y., Wen, J., Xiao, X., & Goh, M. (2023). Energy demand forecasting in China: A support vector
regression-compositional data second exponential smoothing model. Energy, 263.
https://doi.org/10.1016/j.energy.2022.125955

• Sahraei, M. A., Duman, H., Çodur, M. Y., & Eyduran, E. (2021). Prediction of transportation energy demand:
Multivariate Adaptive Regression Splines. Energy, 224. https://doi.org/10.1016/j.energy.2021.120090

• Setiyawan, A., Novianto, A., Afkar, N. B. A., Chabib, F., Amelia, F. R., & Pratiwi, I. (2022a). Diesel engine
performance test using solar-dex and biodiesel (B30) on power and torque. IOP Conference Series: Earth and
Environmental Science, 969(1). https://doi.org/10.1088/1755-1315/969/1/012034

• Setiyawan, A., Novianto, A., Afkar, N. B. A., Chabib, F., Amelia, F. R., & Pratiwi, I. (2022b). Diesel engine
performance test using solar-dex and biodiesel (B30) on power and torque. IOP Conference Series: Earth and
Environmental Science, 969(1). https://doi.org/10.1088/1755-1315/969/1/012034

• Sugiyono, A. (2008). Pengembangan Bahan Bakar Nabati untuk Mengurangi Dampak Pemanasan Global
Utilization of Biomass for Energy Sources View project Perencanaan energi nasional dan daerah View project.
https://www.researchgate.net/publication/275652084