P.R.J. Campbell and K. Adamson (UK)
Modelling, Time Series Analysis, Neural Networks,Power Systems.
Electricity markets throughout the world are changing to allow the integration of alternative energy sources, in particular Wind. Countries such as the USA have already reached penetration with wind energy being integrated into power grids. This development has lead to wind energy producers having to provide forecasts of expected yield within a given time period. These constraints have yet to be placed on wind energy producers in the UK and Ireland, due to slower integration of wind power into main stream power grids. It is likely that as these markets begin to follow the trends developed around the world it will become necessary to produce such forecasts for increasingly finer time periods, perhaps even to within the next hour, in the case of full energy trading. As a result it is necessary to identify a method by which such forecasts can be accurately and efficiently produced. Short term wind speed forecasts have traditionally been compared to the Persistence method of forecasting as a bench mark for accuracy. This paper examines the statistical approaches of ARIMA, Moving Averages and compares their performance against both Persistence and a novel Multi-Layered Perceptron which is trained using the Generalised Delta Rule.
Important Links:
Go Back
