An official website of the United States government
Abstract: In the past decades, clean and renewable energy has gained increasing attention due to a global effort on carbon footprint reduction. In particular, Saudi Arabia is gradually shifting its energy portfolio from an exclusive use of oil to a reliance on renewable energy, and, in particular, wind. Modelling wind for assessing potential energy output in a country as large, geographically diverse and understudied as Saudi Arabia is a challenge which implies highly non-linear dynamic structures in both space and time. To address this, we propose a spatio-temporal model whose spatial information is first reduced via an energy distance-based approach and then its dynamical behaviour is informed by a sparse and stochastic recurrent neural network (Echo State Network). Finally, the full spatial data is reconstructed by means of a non-stationary stochastic partial differential equation-based approach. Our model can capture the fine scale wind structure from a high-resolution simulation and produce more accurate forecasts of both wind speed and energy in lead times of interest for energy grid management.
more »
« less
This content will become publicly available on August 1, 2027
CESAR: A Convolutional Echo State AutoencodeR for High‐Resolution Wind Forecasting
An accurate and timely assessment of wind speed and energy output allows an efficient planning and management of this resource on the power grid. Wind energy, especially at high resolution, calls for the development of nonlinear statistical models able to capture complex dependencies in space and time. This work introduces a Convolutional Echo State AutoencodeR (CESAR), a spatio‐temporal, neural network‐based model which first extracts the spatial features with a deep convolutional autoencoder, and then models their dynamics with an echo state network. We also propose a two‐step approach to also allow for computationally affordable inference, while also performing uncertainty quantification. We focus on a high‐resolution simulation in Riyadh (Saudi Arabia), an area where wind farm planning is currently ongoing, and show how CESAR is able to provide improved forecasting of wind speed and power for proposed building sites by up to 17% against the best alternative methods.
more »
« less
- PAR ID:
- 10698476
- Publisher / Repository:
- Wiley
- Date Published:
- Journal Name:
- Journal of Geophysical Research: Machine Learning and Computation
- Volume:
- 3
- Issue:
- 4
- ISSN:
- 2993-5210
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Abstract Extreme winds associated with tropical cyclones (TCs) can cause significant loss of life and economic damage globally, highlighting the need for accurate, high‐resolution modeling and forecasting for wind. However, due to their coarse horizontal resolution, most global climate and weather models suffer from chronic underprediction of TC wind speeds, limiting their use for impact analysis and energy modeling. In this study, we introduce a cascading deep learning framework designed to downscale high‐resolution TC wind fields given low‐resolution data. Our approach maps 85 TC events from ERA5 data (0.25° resolution) to high‐resolution (0.05° resolution) observations at 6‐hr intervals. The initial component is a debiasing neural network designed to model accurate wind speed observations using ERA5 data. The second component employs a generative super‐resolution strategy based on a conditional denoising diffusion probabilistic model (DDPM) to enhance the spatial resolution and to produce ensemble estimates. The model is able to accurately model intensity and produce realistic radial profiles and fine‐scale spatial structures of wind fields, with a percentage mean bias of −3.74% compared to the high‐resolution observations. Our downscaling framework enables the prediction of high‐resolution wind fields using widely available low‐resolution and intensity wind data, allowing for the modeling of past events and the assessment of future TC risks.more » « less
-
Monitoring the health of wind turbine foundations is essential for ensuring their operational safety. This paper presents a cost-effective approach to obtain rotational stiffness of wind turbine foundations by using only acceleration and wind speed data that are part of SCADA data, thus lowering the use of moment and tilt sensors that are currently being used for obtaining foundation stiffness. First, a convolutional neural network model is applied to map acceleration and wind speed data within a moving window to corresponding moment and tilt values. Rotational stiffness of the foundation is then estimated by fitting a line in the moment-tilt plane. The results obtained indicate that such a mapping model can provide stiffness values that are within 7% of ground truth stiffness values on average. Second, the developed mapping model is re-trained by using synthetic acceleration and wind speed data that are generated by an autoencoder generative AI network. The results obtained indicate that although the exact amount of stiffness drop cannot be determined, the drops themselves can be detected. This mapping model can be used not only to lower the cost associated with obtaining foundation rotational stiffness but also to sound an alarm when a foundation starts deteriorating.more » « less
-
Abstract Citizen science data for monitoring air pollution have recently emerged as a powerful yet under-explored resource to complement expensive and sparse national air quality monitors. In urban environments, these new data have the potential to allow for high-resolution and high-frequency forecasts, and thereby to provide an assessment of population exposure at neighbourhood level. The complex spatio-temporal structure of these data, however, requires new flexible methods that are also able to provide timely forecasts. In this work, we propose a novel method that first provides forecasts with a reservoir computing approach, an echo-state network, adjusts the forecast with a transformer network with attention mechanism and then merges the echo-state and transformer forecast into a combined network. The stochastic nature of the method allows for a fast and more accurate forecast then individual predictors as well as standard statistical methods. Simulation and application to San Francisco air pollution show how the proposed method is able to produce high-resolution urban maps of air quality. Additionally, we show how these forecasts can be used to provide neighbour-level exposure assessment using population data, a task that would not be achievable with sparse government-sponsored air quality networks.more » « less
-
Abstract Wind energy is a growing component of energy production in Illinois (IL). However, uncertainty remains regarding future wind speed variability and trends and their potential impact on wind energy generation. Here, we assess recent (1992–2021) trends in monthly average wind speed and direction using data from the Illinois Climate Network (ICN), which consists of 18 weather stations across IL, supplemented by five IL stations from the Global Historical Climatology Network (GHCN) hourly dataset. We find decreases in mean annual wind speed at most ICN stations, ranging from −0.46% to −17.92% decade−1, and all five GHCN stations (from −1.53% to −6.53% decade−1). Decreasing wind speed trends are found in all seasons. In contrast, 10-m wind speed trends for IL in reanalysis data, a primary source of data for climate model validation and wind energy projection, are positive or show no trend in most grids that approximate IL station locations [1.28%–3.10% decade−1for ERA5, NASA MERRA-2, and NCEP North American Regional Reanalysis (NARR)]. Some IL stations also show a significant shift to more southerly winds, indicating a greater influence of air masses originating from the Gulf of Mexico. This trend is also not captured by reanalysis data, despite accurate wind direction climatologies. Thus, model validation of regional wind should be performed cautiously with reanalysis data, and reliance on reanalysis wind data may lead to an overestimation of future IL wind speeds and wind power generation. Significance StatementReanalysis data are used to validate climate models and for wind energy projection. The purpose of this study is to assess how well three widely used reanalysis products reproduce surface wind observations in Illinois, which ranks fifth in the country for wind power capacity. We found reanalysis data better reproduce wind direction than wind speed and tend to overestimate wind speeds. The station data show significant decreases in wind speeds and changes in wind direction, but those trends are not found in any of the reanalysis products studied here. This could lead to an overestimation of future wind energy generation and inaccuracies in wind in climate models. These results highlight the limitations of reanalysis data, which should be used cautiously.more » « less
