Peramalan Jangka Pendek Output Daya Pembangkit Listrik Tenaga Surya Menggunakan Machine Learning

Ardianto, Ardianto (2022) Peramalan Jangka Pendek Output Daya Pembangkit Listrik Tenaga Surya Menggunakan Machine Learning. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6032201124-Master_Thesis.pdf Restricted to Repository staff only Download (6MB)

Abstract

Pembangkit listrik tenaga surya (PLTS) menjadi solusi yang paling popular dan diterapkan dibanyak negara di seluruh dunia dibandingkan dengan sumber energi terbarukan lainnya. Namun interkoneksi PLTS ke sistem jaringan transmisi listrik menghadirkan permasalahan kepada operator jaringan dikarenakan memiliki sifat intermiten/fluktuasi dalam menghasilkan energi listrik. Fluktuasi tersebut dipengaruhi oleh beberapa faktor meteorologi dan parameter cuaca seperti intensitas radiasi matahari, cuaca, suhu, pergerakan awan, kecepatan angin, dan lain sebagainya. Salah satu langkah mitigasi mengatasi kondisi tersebut yaitu melalui penerapan sistem peralamalan (forecasting) dengan akurasi tinggi dalam memprediksi produksi energi listrik. Peramalan yang akurat akan membantu operator jaringan dalam mengatur keseimbangan antara produksi energi dengan permintaan beban di seluruh jaringan (supply and demand), menjaga keandalan, mutu dan efisiensi pengelolaan jaringan. Pada penelitian ini, peramalan dilakukan melalui pengolahan data histori dari pembacaan sensor yang dimiliki oleh operator pada sistem SCADA (Supervisory Control And Data Acquisition) berupa nilai produksi daya (MW), radiasi (W/m2 ), suhu lingkungan ( oC) suhu peralatan ( oC), dan kecepatan angin (m/s) selama periode 1 Januari s.d. desember 2021. Dataset tersebut dimodelkan menggunakan masing-masing algoritma yaitu Linear Regression (LR), Decision Tree Regression (DTR) dan Random Forest Regression (RFR). Dalam proses pemodelan dilakukan skenario pengaturan beberapa parameter seperti proses perbaikan hilang rekam (data lompat), normalisasi data dan filter produksi. Evaluasi dilakukan dengan menganalisis perbandingan kinerja setiap algoritma beserta kombinasi skenarionya. Pada dataset yang diuji, RFR mempunyai kinerja terbaik dengan nilai R 2=0.9679 RMSE=0.0438, dan MAPE=0.1461. Dari perbandingan nilai R 2 tertingginya, menunjukan kinerja RFR lebih baik 4,91% dari DTR dan 5,32% dari LR. Pemilihan skenario yang tepat terbukti memberi peningkatan kinerja akurasi sebesar RFR 2,81%, DTR 7,64% dan LR 5,47% didapatkan dari selisih antara kinerja terendah dan tertinggi tiap kombinasi algoritma dan skenario pengaturan parameter. Model dengan kinerja akurasi terbaik akan digunakan sebagai panduan operator jaringan dalam proses pengusahaan listrik secara realtime.
==============================================================================================================================
Solar power plants (PLTS) are the most popular solutions and are applied in many countries around the world compared to other renewable energy sources. However, the interconnection of PV mini-grid to the electricity transmission network system presents problems to network operators due to its intermittent/fluctuating nature in the production of electrical energy. These fluctuations are influenced by several meteorological factors such as solar radiation, weather, temperature, cloud movement, wind speed, and others. One of the steps to overcome this condition is the implementation of a forecasting system with high accuracy in predicting the production of electrical energy. Accurate forecasting will help network operators strike a balance between energy production and load demand across the network, reliability, quality and efficiency of network management. In this study, forecasting is done through historical data processing from sensor readings on the SCADA system in the form of power production value (MW), radiation (W/m2), ambient temperature ( oC), equipment temperature ( oC), and wind speed (m/s). during the period of January 1 s.d. december 2021. The dataset is modeled using algorithms, namely Linear Regression (LR), Decision Tree Regression (DTR) and Random Forest Regression (RFR). In the modeling process scenarios are carried out for setting several parameters such as the process of repairing lost records (jumping data), data normalization and production filters. Evaluation is done by analyzing the performance comparison of each algorithm and the combination of scenarios. In the tested dataset, RFR has the best performance with values of R 2=0.9679 RMSE= 0.0438, and MAPE= 0.1461. From the comparison of the highest R2 value, it shows that the RFR performance is 4.91% better than DTR and 5.32% than LR. Scenario selection is proven to increase the accuracy of R 2=0.9679 RMSE=0.0438, and MAPE=0.1461 obtained from the difference between the lowest and highest performance for each combination of algorithms and parameter setting scenarios. The model with the best accuracy performance will be used as a guide for network operators in the process of real- time electrical operation.

Item Type:	Thesis (Masters)
Additional Information:	RTMT 658.403 55 Ard p-1 2022
Uncontrolled Keywords:	PLTS, Pembelajarn mesin, Jangka pendek, Peramalan, Random Forest, PV. PLTS, Machine learning, Short-term, forecasting, Random Forest, PV.
Subjects:	T Technology > T Technology (General)
Divisions:	Interdisciplinary School of Management and Technology (SIMT) > 61101-Master of Technology Management (MMT)
Depositing User:	Mr. Marsudiyana -
Date Deposited:	08 Jul 2026 03:29
Last Modified:	08 Jul 2026 03:29
URI:	http://repository.its.ac.id/id/eprint/134492

Actions (login required)

View Item