Quota Alief Sias, Rahma Gantassi, Yonghoon Choi, A.N. Afandi
An essential part of power system operation and management is day-ahead load forecasting. There is a consider-able impact of weather conditions and holidays on the daily load profile, and it follows a relatively similar pattern throughout the year. This paper describes a simple method to improve the daily forecasting accuracy of energy demand by comparing daily, weekly, monthly, and quarterly forecasting patterns. The model used to make predictions is multi-linear regression (MLR). This paper calls the method recurrence multi-linear regression because the forecasting pattern is carried out recursively. The proposed model has been compared with single linear regression (SLR) as a basic linear model, Gaussian process regression (GPR), and long short-term memory (LSTM). The error value of prediction concludes the performance evaluation from all models. The results show that the best model is the MLR, using weekly patterns for forecasting that can outperform the other models. © 2023 IEEE.
Chonnam National University, Electrical Engineering Department, Gwangju, South Korea; Universitas Negeri Malang, Electrical Engineering Department, Malang, Indonesia
