Bridging Data and Wind Energy Sciences

Yet, according to assistant professor Aziz Ezzat, its full potential is yet to be unleashed for wind energy. Ezzat is also the director of the Renewables and Industrial Analytics (RIA) research group, which was founded in 2019 to address the needs of the wind energy sector while establishing collaborations with industry partners.

To bring wind energy analytics up to speed, Ezzat insists on a need for “physically motivated data science solutions, wherein the construction of key parameters in machine learning models is driven by the underlying physical features of wind dynamics. This contrasts with the black-box, physics-agnostic machine learning paradigm.”

For Ezzat, such an amalgam of data science, meteorology, and engineering is critical if New Jersey is to meet the ambitious goals it has set for cost-efficient offshore energy production by 2035. 

To help address this challenge, Ezzat’s RIA research is focusing on two core areas: wind energy forecasting at various spatial and temporal scales, and reliability and maintenance engineering for offshore wind farms. 

Wind energy’s intermittent, unpredictable nature makes it especially difficult to predict. According to Ezzat, when equipped with accurate forecasts of the amount of energy a wind farm might produce at a given time of day, wind farm operators can plan where and when to expect a dip in energy supply from a given wind farm – and then draw on other sources or locations across the grid to make up for the shortfall. 

“Wind forecasting is a challenging data science problem because it involves fusing data from different sources, resolutions, and types,” Ezzat explains. “The goal is to build models that are neither purely data driven nor solely dictated by physics, but that are a hybrid of both. This builds on a branch of statistics known as ‘spatio-temporal data science,’ which is a major methodological thrust at the RIA group.”

He notes that limited accessibility, high crew dispatch costs, and the unprecedented scale and height of offshore turbines pose a unique set of challenges to cost-effectively maintaining offshore wind farms. “These factors motivate us to focus on formulating maintenance scheduling strategies tailored to offshore wind farm maintenance operations and their offshore-specific operational and environmental conditions,” he says. “We rely on a combination of data science and mathematical programming to seek decisions that minimize operational costs.”

Ezzat has received support from the Rutgers Research Council for his project “The Promise and Peril of Offshore Wind Energy: Powering Up with Machine Learning and Operations Research,” which focuses on bridging forecasting models with mathematical programming to optimally schedule maintenance in offshore wind farms. His Rutgers Energy Institute-funded joint project, “Offshore Wind Energy: The Data Science Relevance,” with Rutgers’ director of atmospheric research Joseph Brodie, seeks to combine meteorological physics-based predictions with turbine-level data for accurate wind resource and energy forecasting. Funding from each grant will support student stipends.

As a co-director of the SoE Energy Lab, Ezzat oversees measurement facilities that collect real-time local measurements and image data about wind velocity, solar irradiance, cloud motion, and more. The lab stores and processes these measurements for use in renewable energy forecasting and analytics by RIA researchers.  

For Ezzat, the benefits accrued by applying data science to wind energy forecasting and optimal maintenance scheduling for offshore wind farms are considerable. “Solutions to these problems lower wind’s cost of energy, making it economically attractive and market-competitive relative to fossil-fuel energy sources,” he notes. “In short, data science can have a direct  -- and positive – impact on the value and profitability of wind farms.”