back

TB2. Forecasting the Demand and Renewable Generation

Demand forecasting at the Microgrid level presented several difficulties. The aggregation or smoothing effect is reduced and uncertainty increases as the size of the Microgrid gets smaller. On this difficulty one should add the increase in time resolution. We enter then the area of very short-term forecasting with reduced smoothing effects. There is a difficulty to find data from real cases to characterize adequately the problem. In addition, in contrast to the classical load forecasting problem, it is expected that the demand will be also correlated to electricity prices. Prediction models for demand may consider as input (predictions) of electricity prices to accommodate this correlation. When dealing with the production of few Kw size wind turbines, the spatial smoothing effect observed in larger wind farms will not be present. In addition, it is very unlikely that numerical weather predictions will be purchased for such a small system. It is therefore expected that the forecasting algorithms for wind power, PV and hydro will only produce results of much higher uncertainties. ARMINES will develop advanced forecasting and profiling functions for loads. For this purpose, data from the real test cases in WP-F will be used. In general for any control or management scheme that considers strategies of participation in the electricity market it is necessary to consider electricity price forecasting and the related uncertainty. ARMINES will develop such forecasting functions as well as methods to estimate on-line the uncertainty of such price forecasts. ARMINES will also develop functions for very short-term forecasting of wind power (up to 6 hours ahead) based on robust machine learning techniques. It is expected however that at urban areas where Microgrids would be mainly developed, PV units rather than wind might be the primary choice concerning renewables. ARMINES will develop forecasting functions for solar radiation and the production of PVs. ICCS/NTUA will use the output of the developed tools as input for the optimisation process.