Abstract
The energy demand crisis is being faced by all the nations due to the rapid growth of the global economy. The conventional resources available on the Earth are finite. Burning these fossil fuels abundantly results in large-scale greenhouse gas emissions and significant environmental contamination. The generation of electricity using renewable energy sources has increased significantly in recent years. However, the power generation using renewable energy sources like solar, wind, etc., is weather dependent and highly erratic. In order to maintain system stability and to use renewable energy resources effectively, renewable power forecast is essential. For the effective planning of power network, three different machine learning algorithms (i.e., linear regression (LR), decision tree regression (DTR) and random-forest regression (RFR)) are used for predicting the solar radiation in Mahabubnagar, Telangana. All the three regression algorithms are evaluated in terms of statistical measures; random-forest regression algorithm provides best results.
Optimization Techniques for Hybrid Power Systems
TopIntroduction
Renewable energy's installed capacity until 2022 has reached to 3372 GW, 168.96GW and 6.15GW across the Globe, India, and Telangana, respectively. Among various renewable energy sources, solar power plays an important role. The solar power installed capacity throughout the world is about 31.22%. In India, it is about 38.1% and in Telangana state, it is around 73.3%. By 2030, the Indian government desires to achieve 500 GW power outcome from renewables. Meanwhile, the solar power installation is expected to reach about 351.2 GW. Although alternate renewable energy sources are exploited to satisfy the rapid increase in energy consumption, their intermittent nature makes it least adoptable for critical loads in the interconnected power grid (IRENA, 2023).
The electrical power system is extremely complicated and is stable when the power generation and demand are equal. Recently, due to the increasing demand for electrical energy, interest in integrated renewable energy sources has increased. Adoption of Integrated renewable energy sources with adequate battery energy storage systems may provide economical and environmentally friendly solution for the isolated areas. Further, the integration enhances the diversification and aids the indigenous locality with higher earnings/savings. Isolated renewable energy driven Micro-grids (MG) are best suited for un-electrified remote rural locations and hilly terrains, where grid connection is not feasible/economical (Prakash et al., 2020). Combining various renewable energy sources faces unique difficulties like availability of wind power, solar radiation intensity etc. in a specific region/location. Precise energy prediction is thus required for the intended operation of the power system.
Optimally constructed micro-grids envision added rewards including energy security, reduced electricity bills, negligible grid dependency, rural economic enhancement, energy access at isolated areas, pollution free environment so on and so forth. Solar power is a significant source of renewable energy because of its wide availability. The sun's light extracted as solar energy has the power to ignite chemical reactions, produce heat and can generate electricity. The total solar energy received by earth exceeds present and future energy needs by a wide margin. Further, solar power has the capacity to satisfy all upcoming energy demands, if suitably harnessed. Various research works are being widely conducted to explore the possibilities of harnessing solar potential with great accuracy. Prior to implementation, innovative strategies along with real-time outcomes should be analyzed in detail.
A unique hybrid strategy that uses periodic collecting and artificial neural networks to anticipate global sun radiation one hour in advance was proposed (Hamza et al., 2021). In consonance to the solar & climatic factors of Evora city, mean experimental data over three years were clustered into various seasons using the fuzzy C-means algorithm. To predict the hourly global solar emission Artificial Neural Network (ANN) model has been implemented. (Vahdettin et al., 2023) adopted machine learning algorithm to find the estimation of monthly solar radiation in Turkey by using support vector machine regression, long short-term memory, K-nearest neighbors, extreme learning machines, Gaussian process regression. Different statistical error calculations like NSE (Nash-Sutcliffe efficiency coefficient), MAE, MARE, RMSE and R2 errors are calculated. LSTM, GPR algorithms were identified as best suited for solar radiation prediction in Turkey.
(Sunayana et al., 2020) proposed a predictive plan for the estimation of solar radiation depending on Random Forest-Particle Swarm Optimization (RF-PSO) hybrid model. The authors have compared various estimation plans like Random Forest (RF), Decision Tree (DT), RF-PSO and Multi-Layer Perceptron (MLP) Neural Network for the considered dataset. Predictive errors like MAE, R2 and RMSE are measured for these abovementioned algorithms. Among the models, RF-PSO gives least MAE and RMSE i.e., 24.45, 47.74 respectively and R2 score is calculated as 95%. Therefore, it is evaluated that RF-PSO hybrid model is best suited for solar radiation estimation with least errors.
Key Terms in this Chapter
Renewable Energy: In the context of human timelines, renewable energy is defined as energy generated from naturally replenishing and virtually inexhaustible sources. These energy sources use events or natural processes to produce energy without using up scarce resources. Renewable energy sources are thought to be more ecologically friendly and sustainable than fossil fuels, which are limited and cause environmental deterioration.
Machine Learning: A branch of artificial intelligence (AI) known as machine learning entails the creation of algorithms and models that allow computers to learn from experience and advance without explicit programming. It is a data-driven methodology that enables systems to automatically discover patterns, correlations, and insights from data in order to generate forecasts, judgements, or carry out activities.
Prediction: Prediction is the process of estimating or anticipating a future result, event, or condition using the data, patterns, and trends that are now accessible. It entails estimating what is likely to occur in the future by analysing current data, historical patterns, and occasionally mathematical or statistical models.
Statistical Measures: Statistical measures are numerical quantities or metrics that reveal details about the distribution, central tendency, variability, linkages, and other properties of a dataset. These metrics aid statisticians, researchers, and analysts in summarizing, describing, and interpreting data, making it simpler to develop judgements and take action. Scientific, commercial, social science, and research disciplines all employ statistical measurements, which are crucial instruments in data analysis.
Regression: A statistical analysis method known as regression is used to simulate the connection between a dependent variable (also known as the response or result) and one or more independent variables (also known as predictors or characteristics). Regression analysis' main objective is to comprehend and measure the connection between the variables so that underlying patterns and trends may be predicted, inferred from, and understood.
Global Horizontal Irradiance: The entire quantity of solar radiation that a horizontal surface on Earth receives is referred to as global horizontal irradiance (GHI).
Direct Normal Irradiation: The quantity of solar energy that directly strikes the surface of the Earth without being refracted or dispersed by the atmosphere is known as direct normal irradiance (DNI). It is often measured perpendicular to the sun's rays and indicates the amount of direct sunlight that strikes a unit area in a given amount of time.
Diffused Horizontal Irradiance: The quantity of solar radiation that reaches the surface of the Earth from all directions, including direct sunlight and scattered light from the sky, is referred to as diffuse horizontal irradiance (DHI).