Optimizing Local Least Squares Regression for Short Term Wind Prediction

Loading...
Thumbnail Image

Authors

Keith, Erin S.

Issue Date

2015

Type

Thesis

Language

Keywords

hyper parameters , linear regression , simulated annealing , wind turbines

Research Projects

Organizational Units

Journal Issue

Alternative Title

Abstract

Highly variable wind velocities in many geographical areas make wind farm integration into the electrical grid difficult. Since a turbine’s electricity output is directly related to wind speed, predicting wind speed will help grid operators predict wind farm electricity output. The goal of experimentation was to discover a way to combine machine learning techniques into an algorithm which is faster than traditional approaches, as accurate or even more so, and easy to implement, which would makes it ideal for industry use. Local Least Squares Regression satisfies these constraints by using a predetermined time window over which a model can be trained, then at each time step trains a new model to predict wind speed values which could subsequently be transmitted to utilities and grid operators. This algorithm can be optimized by finding parameters within the search space which create a model with the lowest root mean squared error.

Description

Citation

Publisher

License

In Copyright(All Rights Reserved)

Journal

Volume

Issue

PubMed ID

DOI

ISSN

EISSN