Abstract
After the liberalization of electricity market in Japan, new entry electricity suppliers will need to schedule the electricity supply and demand plans and procure the shortage electricity from the market. Basically they cannot afford enough capacity of generators and expensive supply and demand control system. Therefore, it is important that knowing characteristics of electricity demand and forecasting a day-ahead electricity demand accurately using the system without heavy process. In addition, the forecasting methods should not be black boxed but also be explainable in qualitatively.
In this paper, we proposed several methods of simplified data reference using day type or temperature as criteria of selecting reference day and evaluated each method’s forecast accuracy. Moreover, we use the best data reference method as the base model of forecasting and combine other methods to improve forecast accuracy with simple condition. The results show that even using simplified data reference methods, it is able to forecast accurately with 6.2% in annual average of daily mean absolute error (MAE). In some estimate conditions, this method is able to reduce imbalance penalty cost which is equal to a few % of supply cost of electricity.
