Presently, electrical energy consumption continues to increase from year to year. Therefore, a short-term load forecasting is required that electricity providers can deliver continuous electrical energy to electricity consumers. By considering the estimation of the electrical load, the scheduling plan for operation and allocation of reserves can be managed well by the supply side. This study is focused on a forecasting of electrical loads using Artificial Neural Network (ANN) method considering a backpropagation algorithm model. The advantage of this method is to forecast the electrical load in accordance with patterns of past loads that have been taught. The data used for the learning is Actual Peak Load Period (APLP) data on the 150 kV system during 2017. Results show that the best network architecture is structured for the APLP Day and Night. Moreover, the momentum setting and understanding rate are 0.85 and 0.1 for the APLP Day. In contrast, 0.9 and 0.15 belong to the APLP Night. Based on the best network architecture, the APLP day testing process generates Mean Squared Error (MSE) around 0.04 and Mean Absolute Percentage Error (MAPE) around 4.66%, while the APLP Night generates MSE in 0.16 and MAPE in 16.83%.

Presently, electrical energy consumption continues to increase from year to year. Therefore, a short-term load forecasting is required that electricity providers can deliver continuous electrical energy to electricity consumers. By considering the estimation of the electrical load, the scheduling plan for operation and allocation of reserves can be managed well by the supply side. This study is focused on a forecasting of electrical loads using Artificial Neural Network (ANN) method considering a backpropagation algorithm model. The advantage of this method is to forecast the electrical load in accordance with patterns of past loads that have been taught. The data used for the learning is Actual Peak Load Period (APLP) data on the 150 kV system during 2017. Results show that the best network architecture is structured for the APLP Day and Night. Moreover, the momentum setting and understanding rate are 0.85 and 0.1 for the APLP Day. In contrast, 0.9 and 0.15 belong to the APLP Night. Based on the best network architecture, the APLP day testing process generates Mean Squared Error (MSE) around 0.04 and Mean Absolute Percentage Error (MAPE) around 4.66%, while the APLP Night generates MSE in 0.16 and MAPE in 16.83%.