Electrical Load Forecasting Using LSTM for Power Stability

Accurate electric load forecasting is essential for effective energy management and optimization. This project leverages Long Short-Term Memory (LSTM) networks, a type of recurrent neural network (RNN), to predict electric load due to their ability to capture long-term dependencies in time series data. The dataset used for this study was sourced from Kaggle, provided by a power supply company in Johor, Malaysia, with weather data (including temperature and humidity) collected from Weather Underground. The model is trained using electric load data along with additional features such as time, temperature, and humidity. Model performance is evaluated using several metrics: Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), Mean Absolute Percentage Error (MAPE), and Mean Squared Error (MSE). The discussion will examine how modifications to the data, model architecture, the training and validation process impact the forecasting results. Results are compared across three different LSTM architectures: 1-layer, 2-layer, and 3-layer models. The best resulted obtained with a MAPE of 16.33%. These findings may assist the utility company in optimizing electricity generation and power supply stability, planning equipment maintenance, and resource procurement. Future work will focus on enhancing model accuracy to ensure more reliable predictions.