A Time-Series Decomposition Based Load Forecasting Technique

Technological advancements in electricity load forecasting (LF) have led to considerable progress from traditional time-series forecasting methods to various deep learning techniques. This broadening of scope is crucial for power system scheduling, planning and maintenance of power systems. In this article, a time-series decomposition-based method known as Ensemble Empirical Mode Decomposition (EEMD) method is integrated with Long Short-Term Memory (LSTM) for LF. EEMD is used to decompose the time-series load data into various modes. The most relevant decomposed modes, which are directly related to the load data, are extracted out and used to train LSTM network along with the original load data and weather-related information. The forecasting process is conducted using MATLAB version 2021a, utilizing a dataset from Panama City that contains information of electricity load consumption (in MW) along with meteorological parameters such as temperature, rainfall and humidity. To evaluate the efficacy of the proposed method, accuracy is assessed in terms of Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE) and Root Mean Absolute Error (RMSE).