CONV1D-GRU: A Hybrid Model for Short-Term Electrical Load Forecasting

With the development of plans to replace conventional meters with smart meters to enhance the functionality and performance of power system, forecasting power consumption is still a challenge due to nonlinearity patterns, different types of noise in the datasets, and sometimes lack of data. Artificial Intelligence plays an important role in utilizing the data sent from smart meters, particularly in the area of short-term load forecasting, which is required to meet the needs of consumers. Based on such insight, this study suggests a hybrid model based on a one-dimension convolutional neural network and gated recurrent units (CONV1D-GRU), to forecast hourly power consumption. The outcome demonstrates that the suggested model is capable of encompassing the patterns in a historical data pertaining the hourly power use. The suggested model has outperformed the competing strategies in the mean absolute percentage error (MAPE%), and the coefficient of determination $(R^{2})$ metrics.