Abstract

The use of renewable energy and specifically solar energy in power systems is rapidly increasing due to significantly lower carbon emissions and low energy costs. Although the widespread use of renewable energy generation provides many benefits to the power system, high levels of renewable energy generation introduce several new challenges to the power system operation. The high level of uncertainty associated with solar power output complicates operation and planning decisions for the power system. Therefore, accurate and reliable solar power forecasts are needed for the planning and operation of the power system more than ever before. This thesis first focuses on improving probabilistic solar power forecasts that provide detailed information on the uncertainty of the forecasts. The proposed copula-based Bayesian method utilizes the underlying relation between temperature and solar power output to improve forecast accuracy and performance. The results show significant improvement compared to the direct use of temperature as an input to the forecast model. Secondly, a novel improvement is made to the State Frequency Memory (SFM) method for solar forecasting. The SFM model, which is based on the Long Short Term Memory (LSTM) method, incorporates the patterns in the frequency domain on top of the time domain considerations. The SFM model is improved by frequency band selection based on the Fourier transform of the solar power data. The improved SFM model is able to include the low-frequency patterns in solar data compared to the sampling frequency of second and minute-level and significantly improve results in very short-term forecasting. Thirdly, One of the challenges that arise from high penetration of solar power is investigated further. An essential part of power system operation is maintaining the balance between generation and loads in the power system. The intermittency of solar power makes it very challenging for the operators to maintain the balance and increases the need for spinning reserves in the power system. In this thesis, solar forecasts are used in a multistep optimization model to control energy storage and electric vehicle charging to minimize violations from the ramp rate limits of the system. A detailed analysis of forecast error is performed to tackle the trade-off between longer forecast horizons and increasing forecast error and find the optimal forecast horizon for predictive solar power smoothing.

Notes

If this is your thesis or dissertation, and want to learn how to access it or for more information about readership statistics, contact us at STARS@ucf.edu.

Graduation Date

2023

Semester

Summer

Advisor

Zhou Sun, Qun

Degree

Doctor of Philosophy (Ph.D.)

College

College of Engineering and Computer Science

Department

Electrical and Computer Engineering

Degree Program

Electrical Engineering

Identifier

CFE0009894; DP0028427

URL

https://purls.library.ucf.edu/go/DP0028427

Language

English

Release Date

February 2027

Length of Campus-only Access

3 years

Access Status

Doctoral Dissertation (Campus-only Access)

Restricted to the UCF community until February 2027; it will then be open access.

Share

COinS