Conference proceeding
Using the Multilayer Perceptron (MLP) Model in Predicting the Patterns of Solar Irradiance at Several Time Intervals
2023 31st Southern African Universities Power Engineering Conference (SAUPEC), pp 1-6
24 Jan 2023
Abstract
Fossil fuels, widely used to produce electricity and power industries, have significantly increased greenhouse gas emissions and led to current global warming. Many countries now allocate considerable resources to Renewable Energy Sources (RES) to produce clean electricity because of the adverse effects of these fossil fuels on the climate. Solar energy is currently one of the most naturally abundant renewable energy sources. Despite being cost-free, solar energy has always been unreliable. The Photovoltaic (PV) panels trap the radiative energy released by the sun, which is expressed in Watts per square meter (Watts/m 2 ). This solar radiation is highly stochastic. It fluctuates during the day before going away entirely at night. This study used Multilayer Perceptron (MLP), a deep learning model, to forecast solar radiation at five different horizons: five, ten, fifteen, thirty, and sixty-minute intervals. The model was trained with two years of meteorological data collected from Johannesburg. The best result was when data collected at five minutes intervals was used to train the model. The model recorded a normalized Root Mean Square Error (nRMSE) of 3.25%. This MLP performance was compared with the results obtained using the SVR model. The best result obtained using the SVR model was 6.26. It is suggested that utilizing this model would make it easier for solar plant operators to predict how much solar energy will be produced at different times of the day. This information will aid in planning failsafe outcomes during cascading solar irradiance, which negatively impacts the electrical energy supply to the power grid.
Metrics
4 Record Views
3 citations in Scopus
Details
- Title
- Using the Multilayer Perceptron (MLP) Model in Predicting the Patterns of Solar Irradiance at Several Time Intervals
- Creators
- Chibuzor N Obiora - University of JohannesburgAhmed Ali - University of JohannesburgAli N Hasan - University of Johannesburg
- Publication Details
- 2023 31st Southern African Universities Power Engineering Conference (SAUPEC), pp 1-6
- Publisher
- IEEE
- Number of pages
- 6
- Resource Type
- Conference proceeding
- Language
- English
- Academic Unit
- Engineering Leadership and Society/Engineering Technology
- Scopus ID
- 2-s2.0-85150181006
- Other Identifier
- 991022004766704721