Identification of atmospheric pressure troughs using image processing techniques

Much of recent research in wind power forecasting focuses on predicting large, sudden changes in wind power output, called wind ramps. Analysis of specific weather phenomena that cause wind ramps can yield significant improvements in the quality of wind ramp forecasts. In this contribution, we concentrate on analysis and detection of pressure trough. Troughs are significant indicators of wind shifts, closely associated with wind ramp related phenomena such as fronts and topographically induced flows. Since troughs are elongated regions of relatively low pressure, it is possible to develop computational methods to identify them automatically in surface pressure fields. The technique described in this paper is based on geometric features in image processing literature. It first identifies candidate trough lines by connecting points of maximum curvature of isobars, and then filters out the spurious lines with weak trough strength (valleyness). We illustrate this approach using sea-level pressure fields extracted from the outputs of a numerical weather prediction model. An example of trough detection is presented, showing that the trough identification method is robust and comparable to visual judgment. Further, we analyzed the association between pressure troughs with ramp events. We tested on a wind farm located in southern Alberta, Canada. Power ramps are identified from actual power generation data. The trough occurrences and their strength at times with and without ramps are compared and analyzed. The results illustrate that both trough occurrences and their strength are distinctly different for ramp and non-ramp events, indicating the potential of pressure troughs to distinguish ramps. These characteristics also have a profound seasonal pattern. Pressure troughs tend to have better identifiability for ramps in winter than in summer.