Real-time load forecasting for demand side management with only a few days of history available

The penetration of distributed generation is growing, and today it is not unusual to see at the same site, equipment that consumes energy working side by side with power sources. To manage and optimize the use of every resource, and to avoid unbalances when the system operates near its limits, the need for very short-term load forecasting is inevitable. In this paper, the use of an already known, fast forecast algorithm, based on double seasonal exponential smoothing is proposed. His real-time usability, with only limited historical data are tested, and the mechanism behind to achieve that are explained. Results, the functionality and reliability on a real world system are also presented and discussed, leading to some concluding remarks at the end of the paper.