Short-Term Wind Forecasting via Atmospheric Pressure Gradients
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current short-term wind forecasting methods, such as the 'naive predictor' or persistence model, are inaccurate for predicting wind speed changes, leading to inefficiencies and reliability issues in wind energy production, as they fail to capture rapid wind speed variations and do not provide reliable hour-ahead forecasts, resulting in wasted power or power shortages.
Innovation Solution
A system and method that uses a physically-based approach by extrapolating slowly-varying large-scale atmospheric pressure force gradients to predict future wind speeds, rather than directly extrapolating wind speed, providing a more accurate short-term wind forecasting model anchored in unsteady atmospheric dynamics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If the naive predictor or persistence model is used for short-term wind forecasting, then the forecasting method is simple and easy to implement, but the forecast accuracy is poor and cannot capture rapid wind speed variations
Solution Approach 1:
The patent introduces an intermediary variable (atmospheric pressure gradient) between the observed wind speeds and the forecasted wind speeds. Instead of directly extrapolating wind speed, the model first infers the pressure gradient from historical wind data, then uses this gradient to predict future wind speeds. This intermediary approach captures the physical drivers of wind variability while maintaining computational simplicity.
Solution Approach 2:
The patent replaces the purely statistical/mechanical extrapolation of wind speed with a physics-based approach that incorporates atmospheric pressure gradients. By substituting the direct wind speed prediction mechanism with one that accounts for the underlying pressure forces, the model achieves better accuracy while remaining computationally efficient.
2Duration of action of stationary object
If meso-scale models are used for medium-term and long-term forecasts, then the forecast coverage is extended, but the accuracy for short-term forecasts deteriorates and computational complexity increases
Solution Approach 1:
The patent segments the forecasting approach by time horizon: using the physics-based pressure gradient model specifically for short-term forecasts (0-12 hours) where accuracy is critical, while allowing meso-scale models to handle medium and long-term forecasts. This segmentation allows each method to operate in its optimal performance range.
Solution Approach 2:
The patent introduces dynamic adjustment by using real-time observed wind speeds to continuously update the inferred pressure gradients, making the model adaptive to changing atmospheric conditions. This dynamic approach captures short-term variability better than static meso-scale models.
3Reliability
If backup generation capacity is kept on standby to compensate for forecast errors, then the reliability of electric supply is maintained, but the operational efficiency and cost-effectiveness deteriorate due to wasted generation capacity
Solution Approach 1:
The patent implements feedback by continuously comparing observed wind speeds with forecasted values and using this information to update the pressure gradient inference. This feedback loop reduces forecast errors over time, allowing for more precise determination of backup generation needs and reducing unnecessary standby capacity.
Solution Approach 2:
The patent performs preliminary wind speed forecasting using the physics-based model to predict future wind conditions before they occur. This allows grid operators to proactively adjust backup generation levels based on predicted wind variability, rather than reactively maintaining high standby capacity to cover unpredictable short-term fluctuations.
Data Source
AI summary
A system and method for performing novel wind forecasting that is particularly accurate for forecasting over short-term time periods, e.g., over the next 1-5 hours. Such wind forecasting is particularly advantageous in wind energy applications. The disclosed method is anchored in a robust physical model of the wind variability in the atmospheric boundary layer (ABL). The disclosed method approach leverages a physical framework based on the unsteady dynamics of earth's atmosphere, and drives forecasting as a function of previously-observed atmospheric condition data observed at the same location for which a wind forecast is desired.