Prediksi Susut Umur Trafo 5 Gardu Induk 150 kV Waru dengan Metode Deep Learning GRU-LSTM Berdasarkan Suhu dan Pembebanan

Prediction of Transformer 5 Aging Loss at 150 kV Waru Substation Using Deep Learning GRU-LSTM Method Based on Temperature and Load

The demand for electricity in Indonesia continues to increase, making transformers a crucial component in the electrical system. This study aims to develop a deep learning method based on GRU (Gated Recurrent Unit) and LSTM (Long Short-Term Memory) to predict the aging loss of transformer unit 5 at the 150 kV Waru Substation. This method was chosen to enhance the reliability and efficiency of the electrical system and to prevent disruptions such as power outages. The research was conducted at the 150 kV Waru Substation, with the research subjects consisting of loading and temperature data of the transformer. Data collection techniques involved processing and normalizing load data, oil temperature, and winding temperature of the transformer. The model architecture was developed using a combination of GRU and LSTM to capture short-term and long-term patterns in time series data. The model was trained using the Adam optimizer with an adjusted learning rate and evaluated using MSE, RMSE, MAPE, and MAE metrics. The research results show that the GRU-LSTM model with a batch size of 64 and 75 epochs provides the best performance with an MSE of 0.0000129008474202634, an RMSE of 0.00359177496793207, a MAPE of 0.3943965%, and a MAE of 0.00000832911556912471. The GRU-LSTM model predicts the aging loss of Transformer 5 for the next 30 days with an average of 0.001378178 pu/day, and the highest predicted aging loss is 0.0030481415 pu.