Masters Theses

Abstract

"Solar radiation prediction models are complex and require software that is not available for the household investor. The processing power within a normal desktop or laptop computer is sufficient to calculate similar models. This barrier to entry for the average consumer can be fixed by a model simple enough to be calculated by hand if necessary.

Solar radiation modeling has been historically difficult to predict and accurate models have significant assumptions and restrictions on their use. Previous methods have been limited to linear relationships, location restrictions, or input data limits to one atmospheric condition. This research takes a novel approach by combining two techniques within the computational limits of a household computer; Clustering and Hidden Markov Models (HMMs). Clustering helps limit the large observation space which restricts the use of HMMs. Instead of using continuous data, and requiring significantly increased computations, the cluster can be used as a qualitative descriptor of each observation. HMMs incorporate a level of uncertainty and take into account the indirect relationship between meteorological indicators and solar radiation. This reduces the complexity of the model enough to be simply understood and accessible to the average household investor.

The solar radiation is considered to be an unobservable state that each household will be unable to measure. The high temperature and the sky coverage are already available through the local or preferred source of weather information. By using the next day's prediction for high temperature and sky coverage, the model groups the data and then predicts the most likely range of radiation. This model uses simple techniques and calculations to give a broad estimate for the solar radiation when no other universal model exists for the average household"--Abstract, page iii.

Advisor(s)

Guardiola, Ivan

Committee Member(s)

Cudney, Elizabeth A.
Wunsch, Donald C.

Department(s)

Engineering Management and Systems Engineering

Degree Name

M.S. in Systems Engineering

Publisher

Missouri University of Science and Technology

Publication Date

2013

Pagination

xi, 63 pages

Note about bibliography

Includes bibliographical references (pages 60-61).

Rights

© 2013 Frank Joseph Sauer, All rights reserved.

Document Type

Thesis - Open Access

File Type

text

Language

English

Library of Congress Subject Headings

Solar radiation -- Computer simulation.
Solar radiation -- Forecasting
Solar energy -- Climatic factors
Hidden Markov models
Cluster analysis

Thesis Number

T 10630

Electronic OCLC #

908110784

Share

 
COinS