 Deep neural networks, DNNs, have been shown to be more accurate than traditional methods when predicting electricity consumption. This study compares the accuracy of DNNs against shallow neural networks, SNNs, double-seasonal halt winters, DSHW models, and autoregressive integrated moving average, ARIMA, models. The results showed that DNNs outperformed all other models, reducing the mean absolute 0.2% error, MAPE, and relative root mean square error, ARIMA-C, by up to 17% and 22%, respectively. This article was authored by Sung Young Ryu, Jaekun Oh, and Hongseek Kim.