This paper aims to derive accurate demand forecast results by applying deep learning algorithm to demand prediction for short-term power load, and describes what advantages the experimental results offer in terms of acceptance.
The test subjects were...
This paper aims to derive accurate demand forecast results by applying deep learning algorithm to demand prediction for short-term power load, and describes what advantages the experimental results offer in terms of acceptance.
The test subjects were selected consumers of four factories, buildings, plant-building complexes, and houses with different power usage types. For each customer, 46 months of 15-minute data were used, and the experiments were carried out in seven detailed preprocessing steps to improve the predictive performance. The prediction result is derived by the methodology developed by applying.
Based on the results obtained, the results of the time series analysis method, which was widely used for the existing power load prediction study, were compared with the results predicted by the moving average method, exponential smoothing method, regression analysis method and ARIMA(Autoregressive Integrated Moving Average) model, and the methodology developed by applying LSTM proved the superiority of 110 out of 168 CASEs.