For verification of this paper, please visit on http://www.socialresearchfoundation.com/innovation.php#8

Photovoltaics (PV), concentrated solar power (CSP), and hybrids of these two technologies are all examples of solar power. Lenses, mirrors, and solar tracking devices combine to concentrate sunlight from a wide region into a narrow beam for use in concentrated solar power systems. Because of the photovoltaic effect, photovoltaic cells are able to turn sunlight into electricity. As a relatively new kind of renewable energy, photovoltaics have so far only been used as a source of power for low- to medium-scale applications, such as the solar-powered calculator or off-grid, rooftop PV systems for houses in rural areas. Concentrated solar power (CSP) and photovoltaic (PV) power are the two most common methods for harnessing solar energy. In the former, also known as solar thermal power production, conventional heat-based technologies are in place to convert heat in the form of steam into electricity. In this research work we used non linear regression analysis techniques. This paper therefore discusses about the different regression techniques used in my research. In my research work data processing will be done by using the weather parameters such as solar irradiation, module temperature, ambient temperature etc. and the performance of the model will be evaluated using suitable and widely used performance indicators.

The objective of this paper is to study the application of machine learning Mmodels in Solar Energy Prediction.

Photovoltaic Power System

A solar cell, sometimes called a photovoltaic cell, is a device that converts solar radiation into electricity by use of the photovoltaic effect. In 1881 [2] the primary solar-oriented cell was constructed. In 1957, researchers at Bell Labs developed a method of thermal oxidation for passivating silicon surfaces. Since then, the surface passivation process has been critical to the performance of solar-oriented cells. A photovoltaic power plan, of a certain kind, provides direct current (DC) management that varies in accordance with the intensity of the sun's rays. In most cases, inverters and a voltage or current converter will be required for practical applications. Inside modules, many solar-facing cells are linked together. Clusters of modules are assembled by wiring, then connected to an inverter to provide controller power at the desired voltage and, in the case of alternating current (AC), the desired frequency and phase [3].

Especially in developed countries with large markets, there are many private PV systems connected to the grid wherever they may be. The use of energy storage is optional in these PV-related matrix-associated systems. Batteries or supplementary influence generators are often included as back-ups in some submissions, such as satellites, encouragements, or in developing countries. The limited sunshine and the presence of such stand-alone control systems permit operations close to nighttime.

 

                               Fig 1: Photovoltaic Power Systems

Solar Energy

Brilliant light and warmth from the Sun are harnessed and used in a variety of ways to create solar energy, such as via photovoltaic cells, liquid salt power plants, sunlight-based engineering, and even synthetic photosynthesis[5]. It is an important renewable energy source, and improvements in this area are classified as either latent solar-powered or dynamic solar-powered, depending on whether they directly capture and distribute solar-based vitality or convert it to solar-powered power. Photovoltaic systems, concentrated solar power, and solar water heating are all examples of dynamic sunlight based techniques that may be used to harness the sun's rays and turn them into usable energy. Methods to make use of latent sunlight include orienting a building toward the Sun, choose building materials that have a positive warm mass or light-scattering qualities, and organising rooms so that air flows freely.

Regression

Regression analysis is a predictive modelling approach that studies the link between the goal or dependent variable and independent variable in a dataset. The numerous forms of regression analysis procedures get applied when the target and independent variables exhibit a linear or non-linear connection between each other, and the target variable comprises continuous values. In order to analyse a number of different types of relationships, including those between causes and effects, time series, and the accuracy of forecasts, regression analysis is often used. 

SVM (Support Vector Machines)

In the field of machine learning, Support Vector Machines (SVMs) are among the most well-known and often used algorithms for handling categorization issues. However, there is a lack of literature on the use of SVMs in regression. This algorithm recognises non-linearity in the data and yields a strong predictive model [6]. Support Vector Regression (SVR) is the common name for the SVM regression method. First, we need to form a mental picture of what a support vector machine is before we can begin developing the algorithm. Support Vector Machines (SVMs) are supervised learning models in machine learning, and the learning techniques they use are utilised for classification and regression analysis of data. The straight line needed to fit the data is called a hyperplane in Support Vector Regression.

KNN (k-nearest neighbors algorithm)

Non-parametric KNN regression uses an intuitive average of nearby data to estimate the relationship between independent variables and the continuous result. Analysts may use cross-validation (which we'll cover in further detail in a bit) to determine the optimal neighbourhood size by identifying the value that minimises the mean squared error, but the size ultimately rests in the hands of the user. While the idea is intriguing, in practise it swiftly breaks down under the weight of a large number of independent factors [7]. You may utilise the KNN method to solve regression issues. The KNN algorithm makes predictions for fresh data points based on their 'feature similarity.' Therefore, the value of the new point is determined by its degree of similarity to the points in the training set. When comparing people of the same height and age, we may assume that ID11's weight is about the same as that of ID1 and ID5[10]. If there had been a classification issue, the mode would have been used to make a final forecast. Here, we have the weight values of 72 and 77 to choose from. Do you have any idea how the ultimate sum will be arrived at? It is customary to settle on an overall estimate by averaging the various figures.

Decision Tree

When it comes to Regression, the non-parametric supervised learning approach known as Decision Trees (DTs) is the way to go. The purpose of this exercise is to build a model that, given certain input data and output data, can reliably predict the value of the target variable. A tree may be understood as an approximation with a piecewise constant[8]. As an example, decision trees may be trained to use a series of if-then-else rules to approximate a sine curve based on historical data. A better-fitting model and more complicated decision-making rules correspond to a deeper tree[9].

Therefore, the proposed study focuses on the efficient self-learning model that will improve regulation within the solar cell's energy restrictions and the regression rate in prediction of energy levels, allowing for better control over performance and resource management.

A quick recap of solar power, regression, SVM, KNN, and the Decision Tree is provided. My research team and I also had a lengthy discussion on the methods we want to utilise. This condensed illustration is crucial to my investigation. Then, their ability to foretell solar output will be evaluated using a simple regression learning technique.

1. Bayindir, Ramazan, Mehmet Yesilbudak, MedineColak, and NaciGenc. "A novel application of Naive Bayes classifier in photovoltaic energy prediction." In 2017 16th IEEE International Conference on Machine Learning and Applications (ICMLA), pp. 523-527. IEEE, 2017. 2. Cammarano, Alessandro, Chiara Petrioli, and Dora Spenza. "Pro-Energy: A novel energy prediction model for solar and wind energy-harvesting wireless sensor networks." In 2012 IEEE 9th International Conference on Mobile Ad-Hoc and Sensor Systems (MASS 2012), pp. 75-83. IEEE, 2012. 3. Martín, Luis, Luis F. Zarzalejo, Jesús Polo, Ana Navarro, Ruth Marchante, and Marco Cony. "Prediction of global solar irradiance based on time series analysis: Application to solar thermal power plants energy production planning." Solar Energy 84, no. 10 (2010): 1772-1781. 4. Wan, Can, Jian Zhao, Yonghua Song, Zhao Xu, Jin Lin, and Zechun Hu. "Photovoltaic and solar power forecasting for smart grid energy management." CSEE Journal of Power and Energy Systems 1, no. 4 (2015): 38-46. 5. Izgi, Ercan, AhmetÖztopal, BihterYerli, Mustafa Kemal Kaymak, and Ahmet Duran Şahin. "Short–mid-term solar power prediction by using artificial neural networks." Solar Energy 86, no. 2 (2012): 725-733. 6. Tao, Cai, DuanShanxu, and Chen Changsong. "Forecasting power output for grid-connected photovoltaic power system without using solar radiation measurement." In The 2nd International Symposium on Power Electronics for Distributed Generation Systems, pp. 773-777. IEEE, 2010. 7. Sharma, Navin, Pranshu Sharma, David Irwin, and PrashantShenoy. "Predicting solar generation from weather forecasts using machine learning." In 2011 IEEE international conference on smart grid communications (SmartGridComm), pp. 528-533. IEEE, 2011. 8. Abedinia, Oveis, NimaAmjady, and NoradinGhadimi. "Solar energy forecasting based on hybrid neural network and improved metaheuristic algorithm." Computational Intelligence 34, no. 1 (2018): 241-260. 9. Mirzapour, Farzaneh, MostafaLakzaei, GoharVaramini, MiladTeimourian, and NoradinGhadimi. "A new prediction model of battery and wind-solar output in hybrid power system." Journal of Ambient Intelligence and Humanized Computing 10, no. 1 (2019): 77-87. 10. Elsheikh, Ammar H., Swellam W. Sharshir, Mohamed AbdElaziz, A. E. Kabeel, Wang Guilan, and Zhang Haiou. "Modeling of solar energy systems using artificial neural network: A comprehensive review." Solar Energy 180 (2019): 622-639.