Comparative Study of LSTM-Based Models With Hyperparameter Optimization For Short-Term Electricity Load Forecasting

Objective

Develop and evaluate LSTM-based deep learning models to improve daily electricity load forecasting accuracy and support energy efficiency and power system management in the Central Java–Yogyakarta electricity network.

Project Details

Collaborative Research with UNNES Lecturer: Led the entire data analysis and LSTM modeling process.
Developed and compared four LSTM variants (Vanilla, Stacked, Bidirectional, and CNN-LSTM) to forecast daily electricity load (2016–2020, ~1,825 data points) in the Central Java–Yogyakarta power system.
Achieved a ~10% improvement in daily load forecasting accuracy, supporting energy efficiency and power management strategies.
Accepted for publication in a Scopus-indexed journal, reflecting international research quality and scientific contribution.

Key Results

✓Developed and compared four LSTM architectures (Vanilla LSTM, Stacked LSTM, Bidirectional LSTM, and CNN-LSTM) using ~1,825 daily electricity load records (2016–2020).
✓Achieved ~10% improvement in forecasting accuracy and produced research that was accepted for publication in a Scopus-indexed journal.

Comments (1)

Rian HidayatJul 21, 2026

Publikasi riset jurnal terindeks Scopus tentang varian LSTM ini sangat menginspirasi.

Project Information

Date:Sep 2024

Duration:6 months

Tools & Technologies

PythonTensorFlowKerasScikit-learnPandasNumPyMatplotlibLSTMBiLSTMCNN-LSTMTime Series ForecastingHyperparameter TuningData AnalysisData Visualization