A new approach based on coupled map lattices for load forecasting in smart grids

Over the years, several methods were applied to forecasting problem, usually without considering the system topology. With smart grid introduction the connection between grid components and data amount generated can be considered to improve load forecasts. Using dynamical systems concepts it's possible to include network characteristics, such as connections between buses and lines impedances, by applying models that includes spatially extended systems which have time and space dynamic, such as coupled map lattices. For this technique application is not necessary a long data history, the problem local dynamics is detected with only one delay, covering the problem originated by newly added equipment. The obtained results are compared with a linear regression model. The error values suggest that the proposed method has an adequate performance when very few historical data are available. Compared to other models has the advantage of solving the whole system in one iteration, regardless the system size.