IKSAN, Nur and Purwanto, Purwanto and Sutanto, Heri (2026) SISTEM MONITORING KINERJA ENERGI SURYA FOTOVOLTAIK PADA SMART MICROGRID MENGGUNAKAN MULTI-LSTM. Doctoral thesis, UNIVERSITAS DIPONEGORO.

	Text cover aja.pdf Download (160kB)
	Text Cover-Nur Iksan- Disertasi- Mahasiswa DSI UNDIP.pdf Restricted to Repository staff only Download (3MB)
	Text Bab I-Nur Iksan- Disertasi- Mahasiswa DSI UNDIP.pdf Download (278kB)
	Text Bab II-Nur Iksan- Disertasi- Mahasiswa DSI UNDIP.pdf Download (3MB)
	Text Bab III-Nur Iksan- Laporan Disertasi- Mahasiswa DSI UNDIP.pdf Restricted to Repository staff only Download (3MB)
	Text Bab IV-Nur Iksan- Disertasi- Mahasiswa DSI UNDIP.pdf Restricted to Repository staff only Download (13MB)
	Text Bab V-Nur Iksan- Disertasi- Mahasiswa DSI UNDIP.pdf Restricted to Repository staff only Download (128kB)
	Text Daftar Pustaka-Nur Iksan- Laporan Disertasi- Mahasiswa DSI UNDIP.pdf Download (185kB)
	Text Lampiran-Nur Iksan- Laporan Disertasi- Mahasiswa DSI UNDIP.pdf Restricted to Repository staff only Download (111kB)

Abstract

Sumber energi surya fotovoltaik memiliki karakteristik bersifat intermittent yang menimbulkan tantangan dalam pengelolaan pasokan energi pada lingkungan smart microgrid, khususnya dalam penyediaan informasi yang andal untuk mendukung pengambilan keputusan manajemen energi. Permasalahan utama dalam penelitian ini pada integrasi model kecerdasan buatan berbasis Multi Long Short-Term Memory (Multi-LSTM) ke dalam sistem informasi monitoring kinerja fotovoltaik, serta evaluasi kualitas informasi prediktif yang dihasilkan oleh sistem tersebut. Arsitektur sistem informasi yang dikembangkan mencakup proses akuisisi data operasional dan lingkungan berbasis Internet of Things, pra-pemrosesan data, analitik prediktif berbasis Multi-LSTM, serta penyajian informasi melalui antarmuka monitoring. Model Multi-LSTM digunakan untuk memetakan data historis multivariat menjadi informasi peramalan produksi energi surya yang berguna bagi pengelolaan smart microgrid. Hasil eksperimen menunjukkan bahwa model Multi-LSTM memberikan kinerja peramalan terbaik dibandingkan model pembanding, yaitu Convolutional Neural Network, CNN-LSTM, dan StackedLSTM. Pada kombinasi input multivariat, model Multi-LSTM menghasilkan nilai kesalahan terendah dengan Mean Squared Error sebesar 0,0090, Mean Absolute Error sebesar 0,0001, dan Root Mean Squared Error sebesar 0,0951. Evaluasi kualitas informasi menunjukkan bahwa sistem informasi berbasis Multi-LSTM mampu menghasilkan informasi prediktif yang andal, ditunjukkan oleh nilai Indeks Kualitas Informasi tertinggi sebesar 10,51, tingkat Ketidakpastian Prediksi yang rendah sebesar 0,0950, serta nilai indeks kompleksitas terendah sebesar 3,28. Selain itu, nilai Risiko Kesalahan Maksimum relatif rendah sebesar 0,1750. Hasil ini menunjukkan bahwa sistem monitoring yang dikembangkan tidak hanya unggul secara akurasi peramalan, tetapi juga mampu menghasilkan informasi prediktif berkualitas yang layak digunakan sebagai masukan dalam sistem pendukung keputusan untuk perencanaan suplai, pengelolaan beban, dan optimalisasi pemanfaatan energi surya pada smart microgrid.
Kata kunci: Surya PV; Smart Mikro Grid; Multi-LSTM; Peramalan Energi Surya; Sistem Informasi

Photovoltaic solar energy sources have intermittent characteristics that pose challenges in managing energy supply in smart microgrid environments, especially in providing reliable information to support energy management decision-making. The main problem in this research is the integration of an artificial intelligence model based on Multi Long Short-Term Memory (Multi-LSTM) into a photovoltaic performance monitoring information system, as well as evaluating the quality of predictive information generated by the system. The information system architecture developed includes the process of acquiring operational and environmental data based on the Internet of Things, data pre-processing, predictive analytics based on Multi-LSTM, and presentation of information through a monitoring interface. The Multi-LSTM model is used to map multivariate historical data into solar energy production forecasting information that is useful for smart microgrid management. Experimental results show that the Multi-LSTM model provides the best forecasting performance compared to comparison models, namely Convolutional Neural Network, CNN-LSTM, and Stacked-LSTM. In the multivariate input combination, the Multi-LSTM model produces the lowest error value with a Mean Squared Error of 0.0090, a Mean Absolute Error of 0.0001, and a Root Mean Squared Error of 0.0951. The evaluation of information quality shows that the Multi-LSTM-based information system is able to produce reliable predictive information, indicated by the highest Information Quality Index value of 10.51, a low Prediction Uncertainty level of 0.0950, and the lowest complexity index value of 3.28. In addition, the Maximum Error Risk value is relatively low at 0.1750. These results indicate that the developed monitoring system is not only superior in terms of forecasting accuracy, but is also able to produce quality predictive information that is suitable for use as input in decision support systems for supply planning, load management, and optimization of solar energy utilization in smart microgrids.
Keyword: Solar PV; Smart Microgrid; Multi-LSTM; Solar Energy Forecasting; Information System

Actions (login required)

View Item