# Machine Learning Based Analysis and Quantification of Potential Power Gain from Passive Device Installation

## TL;DR

This paper introduces a machine learning approach using adaptive kernel methods and surrogate modeling to quantify the potential power gain from passive device installation on wind turbines, accounting for wind variability.

## Contribution

It develops an adaptive kernel-based surrogate model and an analysis framework using nearby turbines to estimate power gains, addressing challenges of wind variability and small effect sizes.

## Key findings

- Quantifies potential power gains of 1-5% from passive devices.
- Reduces unexplained variation in wind data using surrogate modeling.
- Provides a framework for controlled-like analysis using nearby turbines.

## Abstract

Passive device installation on wind turbine generators (WTGs) can potentially improve the power generation of WTGs. Yet, how much impact the installation will make is unclear because conducting controlled experiments is impossible due to ever-changing wind and weather that affect the power generation significantly. In addition, the potential improvement is believed to be in a small scale, such as 1-5%, which is less than a typical 3-8% variation level observed in wind data. This article proposes an adaptive kernel-based method and builds a surrogate model to reduce the level of unexplained variation in wind data. In addition, to establish experimental environments that are similar to a controlled situation, this article develops an analysis framework that utilizes two other nearby WTGs without any passive device installation.