Solar Radiation Forecasting

ABSTRACT

In recent decades, the integration of solar energy sources has gradually become the main challenge for global energy consumption. Therefore, it is essential to predict global solar radiation in an accurate and efficient way when estimating outputs of the solar system. Inaccurate predictions either cause load overestimation that results in increased cost or failure to gather adequate supplies. However, accurate forecasting is a challenging task because solar resources are intermittent and uncontrollable. To tackle this difficulty, several machine learning models have been established; however, the forecasting outcomes of these models are not sufficiently accurate. Therefore, in this study, we investigate ensemble learning with square root regularization and intelligent optimization to forecast hourly global solar radiation.

CHAPTER ONE

1.0                                          INTRODUCTION

Over the past few decades, solar panels have advanced to capture as much of the power of the sun as possible. Among other things, these solar panels can convert solar radiation into electricity to power homes and businesses. A large-scale application of solar panels at a site (solar farms) can generate enough electricity to power hundreds or even thousands of homes and businesses. Another essential function of solar radiation is to help plants and crops grow via photosynthesis.

To help industries (such as energy and agriculture) make sound business decisions, state-sponsored global model forecast output contains information about the amount of solar radiation impinging upon the earth’s surface at hundreds of thousands or even millions of grid points.

Solar power forecasting involves knowledge of the Sun´s path, the atmosphere’s condition, the scattering processes and the characteristics of a solar energy plant which utilizes the Sun’s energy to produce solar power. Solar photovoltaic systems transform solar energy into electric power. The power output depends on the incoming radiation and on the solar panel characteristics. Photovoltaic power production is increasing nowadays. Forecast information is essential for an efficient use, the management of the electricity grid and for solar energy trading. Common solar forecasting method include stochastic learning methods, local and remote sensing methods, and hybrid methods (Chu et al. 2016). The aim of this work is to discuss

1.1                            BACKGROUND OF THE STUDY

Sunlight is the fuel for all solar energy generation technologies. For any generation source, knowledge of the quality and future reliability of the fuel is essential for accurate analyses of system performance and to determine the financial viability of a project. For solar energy systems, the variability of the supply of sunlight probably represents the single greatest uncertainty in a solar power plant’s predicted performance. Solar resource data and modeling factor into three elements of a solar project’s life:

• Historical long-term data for site selection during feasibility studies
• Prediction of power plant output for plant design and financing
• Real-time measurement and solar forecasting for plant and grid

Site selection includes numerous location properties, including current land use, grid access, and proximity to load centers, but a top priority is determining if an adequate solar resource exists for a proposed project. For site selection, average solar irradiation at the site is the first selection criterion. Geographical latitude is also considered because sites close to the equator have advantages such as lower geometrical losses and lower shadowing. Lower seasonal variability at locations near the equator could also be advantageous because of a more consistent match to the power demand. As weather patterns may change from year to year, many years of data are required for determining reliable average irradiation conditions and inter-annual variability. For this purpose, satellite-derived, high-quality historic solar radiation data sets covering over 10 years are usually considered necessary for site selection, although site-specific climate conditions or design criteria may allow a shorter period.

As flat-priced electricity feed-in-tariff regulations get phased out, the economic yield of solar power systems depends more and more on the solar production during various times of the year as well as on its availability during specific parts of the day. Thus, for solar projects with variable prices, the temporal distribution of solar irradiance may be critical, even during site selection, to estimate potential yields among alternative sites. At this early stage of project development, it is sufficient to study the temporal variability of the energy output throughout the year and typical daily cycles. As an alternative to multiple-year data sets, typical meteorological year (TMY) data for each site may be sufficient at this stage, although the TMY will not characterize inter-annual variability.

If a site is found to be feasible and a power project is to be developed, more precise and detailed data sets are required. For the site-specific techno-economic optimization of a solar system, availability of higher time resolution data is always beneficial. Advanced modeling techniques allow developing such data based on satellite-derived time series. For financing large solar plants, data sets that are validated by ground measurements on or near the site become essential to lower the yield risk. In addition to precise solar radiation measurements, specialized meteorological stations usually provide additional environmental parameters that help to optimize the sizing and proper selection of plant components.

Precise solar and meteorological stations are also valuable during plant commissioning; reliable measurements are the base for acceptance testing to demonstrate proof of fulfillment of technical specifications for heat or electric output. Although temporary measurement equipment may be used for acceptance testing, reliable measurements are essential for estimating real-time plant output to assure high efficiency of the plant throughout its service life. Evaluation of plant output as a function of solar irradiance is the most important indicator of power plant performance. A drop in overall efficiency implies a degradation of one or more power plant components or poor maintenance or operation. Although remotely sensed data may be used for smaller rooftop systems where performance accuracy can be relaxed, larger solar systems usually rely on ground-based measurements, which may be combined with near-real-time satellite-derived solar radiation data. Local ground measurements also assist in site-specific model validation and improvements of solar forecasting.

Proper and accurate solar forecasts are important for ideal use of solar plants both economically and operationally. They help to improve system operations such as optimal use of a storage tank in a solar thermal water heating system, a molten salt system for high-temperature applications, or a battery system in an off-grid photovoltaic (PV) system. With the fast growth of grid- connected solar electrical systems, solar radiation forecasts have become highly important for safe grid operations and efficient operations among power plants, which might be necessary to balance solar fluctuations.

1.2                                   PROBLEM STATEMENT

An increasing number of countries are paying substantial attention to environmental issues such as global warming, climate change, and greenhouse gas emissions. A major cause of global warming is the burning of fossil fuels including coal and oil for supplying traditional electricity generation. This encourages the exploitation and utilization of renewable energy sources including solar, wind, hydro, tidal, wave, biomass, and geothermal sources as alternatives. Nevertheless, solar energy, as a typical renewable energy source, has stochastic, intermittence, time-varying, and uncertainty properties, which may result in defects in the stability and reliability of power grid systems. This introduces new challenges for the incorporation of solar energy sources into the power grids. In addition, solar radiation can be measured based on sensors including thermometers, anemometers, and radiometers combined with data-acquisition hardware and software. However, it is time-consuming, cumbersome, and expensive to install sensors across the world.

To tackle these challenges, it is necessary to develop accurate global solar radiation forecasting models. A wide variety of models have been established using machine learning techniques. Machine learning, which is classified as an artificial intelligence technology, is a sub-field of computer science. It can be applied to numerous domains, and its characteristic is that the model can identify relations between inputs and outputs notwithstanding whether the representation is infeasible.

1.3                                       AIM OF THE STUDY

The main aim of this work is to discuss solar radiation forecasting using machine learning techniques.

1.4                                  PURPOSE OF THE STUDY

Forecasting the output power of solar systems is required for the good operation of the power grid or for the optimal management of the energy fluxes occurring into the solar system. The main purpose of the study is to ensure stability and reliability of solar power systems

1.5                             SIGNIFICANCE OF THE STUDY

Solar radiation forecasting anticipates the solar radiation transients and the power production of solar energy systems, allowing for the setup of contingency mechanisms to mitigate any deviation from the required production.

1.6                                    RESEARCH QUESTION

Why is solar forecasting important?

What is solar radiation and why is it important?

What are the uses of solar radiation?

What is a solar forecast?

1.7                             APPLICATION OF THE STUDY

This study is useful to a wide range of users of solar heating and cooling, photovoltaic, and concentrating solar power systems and of building developers and owners as well as anyone else who needs to understand and predict sunlight for agricultural or other purposes.

1.8                                             PROJECT ORGANIZATION

The work is organized as follows: chapter one discuses the introductory part of the work,   chapter two presents the literature review of the study,  chapter three describes the methods applied, chapter four discusses the results of the work, chapter five summarizes the research outcomes and the recommendations.