Proactive Production Forecasting to Support Offshore Wind Development

This research explores a proactive production forecasting methodology to support the current offshore wind development in the U.S., highlighting the effectiveness of utilizing a well-trained long short-term memory (LSTM) machine learning model, which robustly captures intricate temporal patterns in historical wind speed trends, wind turbine power curves, forecasted offshore wind speeds, and wind turbine configuration parameters, thus offering a reliable foundation for effective power production predictions. Due to the underdevelopment of offshore wind in the U.S. and the lack of access to true offshore wind farm operational data, this study builds, trains, and tests a generalized model on four existing land-based operational wind farms, and then deploys the model on a New Jersey offshore wind farm that is now under development. Equipped with such a proactive production prediction tool, offshore wind developers and operators could pursue alternative profitable opportunities in the ISO/RTOs' wholesale electricity markets when state funding for renewable purchases, such as Offshore Renewable Energy Credits (ORECs) or power purchase agreements (PPAs), is not available. Accurate production forecasting can mitigate potential penalties for offshore wind power producers (WPPs) due to over and/or under commitment in the forward wholesale electricity markets and prove them to be as competitive in the markets as those traditionally predictable non-renewable resources.