Sistem Peringatan Dini Lonjakan Konsumsi Energi Listrik Pada Area Produksi Berdasarkan Pola Penggunaan Historis Menggunakan Metode Long Short-Term Memory (LSTM)

Aisy, Nadiah Rohadatul (2025) Sistem Peringatan Dini Lonjakan Konsumsi Energi Listrik Pada Area Produksi Berdasarkan Pola Penggunaan Historis Menggunakan Metode Long Short-Term Memory (LSTM). Other thesis, Institut Teknologi Sepuluh Nopember.

Text 2040211008-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (7MB) | Request a copy

Abstract

Dalam proses produksi industri, energi listrik memegang peranan penting karena digunakan untuk mengoperasikan berbagai mesin dan sistem pendukung. Oleh karena itu, pemantauan konsumsi energi listrik secara berkelanjutan menjadi tantangan tersendiri. Konsumsi energi listrik yang melebihi batas normal dapat meningkatkan biaya operasional secara signifikan. Pengelolaan energi listrik dan efisiensi biaya operasional industri menjadi aspek krusial dalam menjaga keberlanjutan dan daya saing industri.Penelitian ini bertujuan untuk mengembangkan sistem prediksi konsumsi energi listrik sekaligus mendeteksi lonjakan penggunaan energi listrik secara otomatis melalui pendekatan statistik dan kecerdasan buatan. Sistem ini dirancang dengan mengintegrasikan model Long Short-Term Memory (LSTM) untuk memprediksi konsumsi energi listrik berdasarkan data historis, serta metode standar deviasi sebagai perhitungan ambang batas (threshold) untuk mendeteksi adanya lonjakan konsumsi energi listrik yang melebihi rata-rata penggunaan. Data historis diperoleh dari 13 mesin produksi melalui Energy Management System (EMS) yang sudah diterapkan di PT X, dengan pemanfaatan power meter untuk pencatatan konsumsi energi listrik. Model prediksi dikembangkan secara individual untuk setiap mesin, dengan membedakan antara konsumsi Luar Waktu Beban Puncak (LWBP) dan Waktu Beban Puncak (WBP). Jika hasil prediksi menunjukkan nilai konsumsi melebihi ambang batas yang dihitung berdasarkan rata-rata dan standar deviasi, sistem akan mengirimkan peringatan dini melalui Telegram bot secara otomatis. Hasil pengujian menunjukkan bahwa model memiliki performa prediksi yang optimal dengan nilai Mean Absolute Error (MAE) terendah sebesar 0.0157 dan Root Mean Square Error (RMSE) terendah sebesar 0.0278. Sistem peringatan dini mampu mengidentifikasi potensi lonjakan konsumsi energi listrik secara responsif, dengan waktu pengiriman peringatan rata-rata 5 hingga 10 menit setelah data prediksi tersimpan dalam database.

======================================================================================================================================
In industrial production processes, electrical energy plays a vital role as it is used to operate various machines and supporting systems. Therefore, continuous monitoring of electricity consumption presents a significant challenge. Excessive electricity consumption beyond normal thresholds can substantially increase operational costs. Energy management and cost efficiency have become critical aspects in maintaining the sustainability and competitiveness of the industry. This study aims to develop an electricity consumption prediction system while simultaneously detecting spikes in energy usage automatically through a combination of statistical and artificial intelligence approaches. The system is designed by integrating a Long Short-Term Memory (LSTM) model to forecast electricity consumption based on historical data, and a standard deviation method to calculate threshold values for detecting abnormal spikes in energy usage exceeding the average consumption. Historical data were obtained from 13 production machines through the Energy Management System (EMS) implemented at PT X, utilizing power meters to record electricity consumption. The prediction model was developed individually for each machine, distinguishing between Off-Peak Load (LWBP) and Peak Load (WBP) consumption. If the prediction results exceed the calculated threshold—based on the mean and standard deviation—the system automatically sends an early warning notification via Telegram bot. The testing results indicate that the model achieved optimal predictive performance, with the lowest Mean Absolute Error (MAE) of 0.0157 and the lowest Root Mean Square Error (RMSE) of 0.0278. The early warning system effectively identified potential spikes in electricity consumption in a timely manner, with an average alert delivery time of 5 to 10 minutes after the predicted data were stored in the database.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Energy Management System (EMS), Konsumsi Energi Listrik, Long Short-Term Memory (LSTM), Standar Deviasi, Sistem Peringatan Dini, Energy Management System (EMS), Electricity Consumption, Long Short-Term Memory (LSTM), Standard Deviation, Early Warning System.
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5105.888 Web sites--Design. Web site development.
Divisions:	Faculty of Vocational > 36304-Automation Electronic Engineering
Depositing User:	Nadiah Rohadatul Aisy
Date Deposited:	01 Aug 2025 09:34
Last Modified:	01 Aug 2025 09:34
URI:	http://repository.its.ac.id/id/eprint/124999

Actions (login required)

View Item