Smart Energy and Electric Power Systems: Current Trends and New Intelligent Perspectives

Smart Energy and Electric Power Systems: Current Trends and New Intelligent Perspectives reviews key applications of intelligent algorithms and machine learning techniques to increasingly complex and data-driven power systems with distributed energy resources to enable evidence-driven decision-making and mitigate catastrophic power shortages. The book reviews foundations towards the integration of machine learning and smart power systems before addressing key challenges and issues. The work then explores AI- and ML-informed techniques to rebalancing of supply and demand. Methods discussed include distributed energy resources and prosumer markets, electricity demand prediction, component fault detection, and load balancing.

Security solutions are introduced, along with potential solutions to cyberattacks, security data detection and critical loads in power systems. The work closes with a lengthy discussion, informed by case studies, on integrating AI and ML into the modern energy sector.

• Helps improve the prediction capability of AI algorithms to make evidence-based decisions in the smart supply of electricity, including load shedding
• Focuses on how to integrate AI and ML into the energy sector in the real-world, with many chapters accompanied by case studies
• Addresses a number of proven AI and ML- informed techniques in rebalancing supply and demand

• Language: English
• Publisher: Elsevier Science
• Release date: Sep 17, 2022
• ISBN: 9780323916851

Book preview

Preface

The economic growth of a country is mainly dependent on its primary infrastructures. One such primary infrastructure is electricity. Both industrial and service sector revolution have embarked the surge in demand for electricity. Developing countries have started investing to make electricity infrastructure smarter and began generating electricity from alternate sources such as wind, solar, biomass, etc. Despite the production of adequate electricity, the country still faces a huge demand for electricity, which leads to blackout.

Power outages are unpredictable and various factors contribute to it. Some of them include failure of components in local grid, network capacity overload, illicit connections to local grid, exposure to bad weather conditions, deliberate disconnection, etc. Load shedding can result in loss if power cuts are not planned well according to low demands. This cumbersome task of analyzing the demands for electricity unravels the technique of using artificial intelligence (AI) to smartly handle the demand of electricity without greatly affecting the economy of a nation.

Machine learning (ML) and AI techniques are extensively used in all the domains for the meticulous prediction of events by learning from the given training data. ML and AI techniques are extensively studied to improve accuracy in decision-making process. AI techniques can now be applied to power systems to predict the demand for electricity in various locations from the consumption history. This enables the power systems to be smarter in determining the load shedding schedule without impacting the economy. This can further predict the power outage events that are likely to occur from bad weather conditions, deteriorating grid components.

This book aims to bring the convenience of AI in power systems. This will address the issues associated with power systems and gives insight into the techniques of load shedding. AI techniques to predict the demand pattern are explored in detail. This book will enable researchers to focus on their ideas in improving the prediction capability of AI algorithms to make careful decisions in load shedding. Demand crisis in electricity cannot be eliminated, but it can be minimized with little impact by making intelligent power systems to handle the demands. This book also focuses on techniques and approaches leading to smart power systems.

Chapter 1

Smart power systems: an eyeview

Kayal Padmanandam¹, Subetha Thangaraj¹ and Rashmita Khilar²,    ¹BVRIT HYDERABAD College of Engineering for Women, Hyderabad, Telangana, India,    ²Saveetha School of Engineering, Saveetha Institute of Medical and Technical Sciences, Saveetha University, Chennai, Tamil Nadu, India

Abstract

The term Artificial Intelligence (AI) was disseminated in a conference at Dartmouth College, United States in 1956. These talent meets paved a way for researchers to come forward and think out of the box in the field of AI from language simulation to learning machines.

Since then, decades have been spent for machine to imitate human intelligence. Despite the hardwork, progress was not impressive. But for the past two decades, we can witness tremendous growth in the domain of AI. AI is an umbrella that clutches machine learning and deep learning techniques. Machine-learning algorithms have been grown expressly due to the expansion of deep learning and reinforcement-learning techniques, which are grounded on neural networks. In this chapter, we study the influence of AI, its necessity, techniques, and challenges in the smart power system.

Keywords

Artificial intelligence; smart grid; power systems

1.1 Introduction to artificial intelligence

Many factors have contributed to the progress of AI and its sub domains. The first and foremost reason is the exponential growth of computing capacity. It has become available in large scale for training bigger and multifaceted complex models. This is possible and feasible due to the extra ordinary innovations made with silicon alike graphics processing units and tensor processing units, and many more on the way. This capacity is amazed with hyperscale clusters and made accessible to the user effortlessly through cloud architecture.

Despite the different artificial intelligence (AI) definitions alike, intelligent computer and machine technology, intelligent automated systems, intelligent machine that replaces human work, etc., it is built to help humankind by emulating cognitive abilities of human [1] to continue complex process hassle-free, when it is harnessed properly.

AI has invaded almost all the service oriented and business-oriented sectors in the universe, and so the power sector. Access to clean, cheap, and reliable energy is a fundamental right for human kind, which has direct impact on health, education, societal security, and livelihood. The main goal of sustainable development is to have universal access to such smart energy. Worldwide, researchers are involved in the automation of routines in the energy sector to prevent emergent markets from the nonexistence of satisfactory power generation, deprived transmission and distribution infrastructure, cost-effectiveness etc., Besides, the divergence, regionalization of energy production, changing demand patterns, crop intricate challenges for power generation, transmission, distribution, and consumption nationwide [2].

AI has the capability to accelerate the use of energy sources in power grid in an efficient way. It can drastically reduce and monitor any kind of faults in the planning or procedure of large power system. It is the best approach for clean, cheap and reliable energy, which is a global need in today’s scenario.

1.2 Necessity of artificial intelligence in power systems

Besides the swift growth in industrial development due to the advancement in power system expansion, there is also an exponentially increasing growing need for dynamic, stable, and reliable power system. Conventional power system techniques cannot handle this mass requirement due to the lack of modern infrastructure. Monitoring remote devices, heavy data acquisition and analysis are becoming complex and time-consuming process [3]. As the proverb says Necessity is the mother of Invention, AI has become a necessity to solve the complexity in large power system.

Modern AI techniques like artificial neural networks (ANNs), expert system techniques (XPS), fuzzy logic (FL), genetic algorithm (GA) are utilized extensively in larger power system to manage complex task alike load dispatch, load forecasting, load optimization, transmission flow, generator monitoring, fault diagnosis, voltage control, electricity market analysis, security measures and very many operations and controls required for smart grids. These operations are handled intelligently through AI and can drastically reduce the human effort and risks involved in the power system (Fig. 1.1).

Figure 1.1 Necessity of AI in power systems. AI, Artificial intelligence.

1.3 Modern artificial intelligence techniques in power system

The four major AI techniques used in power system are ANNs, XPS, FL, GA. These approaches can handle Intricate, versatile and huge information used for calculation, analysis and learning very easily. Despite vast data handling capabilities, they can also effortlessly handle higher computational system.

1.3.1 Artificial neural networks

ANN is a recent and successful technology applied in most of the complex and high priority systems. ANN is a human developed computing system that emulates biological neural networks. It is a collection of interconnected nodes called neurons that simulates the electrical activity of a biological brain. These networked neurons convert set of inputs to set of outputs. Each neuron in the network, act as a processor, computes linear and nonlinear operations from the input, and produces an output through activation function. The complex interconnected working principle of neurons has the capability to solve many real-world problems because of its high end parallelism, asynchronous processing, multifrontal execution, real-time malleability, robustness toward damages and missing data, learning ability and involuntary generality, cutdown of additional software, fixed configuration and primarily the well-developed mathematical foundation [4,5].

ANNs are categorized based on the number of layers, topology, and the pattern of connectivity used. Their layers are input layer which represents the input vector, hidden layer which represents intermediary nodes, and output layer which represents output. Basically neural networks are classified into perceptron, feed-forward neural network, multilayer perceptron, convolutional neural network (CNN), recurrent neural network (RNN), radial basis functional neural network, long short-term memory (LSTM), sequence to sequence models, modular neural network, etc. Among these, the most important and most implemented networks are CNN, RNN, and LSTM. CNN are commonly used in image processing, computer vision and machine translation applications. They comprehend imageries in parts and compute the process manifolds to achieve the complete image processing output. Various applications in power system like event classification and localization [6], power network icing image [7] are efficiently monitored using CNN algorithms. RNN use sequential or temporal data and helps in prediction of a system. They need training data to learn and use memory to take information from preceding inputs and influence the current input and output. They can be used to predict the load forecasting [8], instability prediction etc., in smart grid applications. LSTM network is a cutting-edge RNN and a sequential network that consents information to persist. It is a highly celebrated for handling the vanishing gradient problem faced by RNN. LSTMs are explicitly designed with three gates forget, input and output to forget irrelevant information and carry over only relevant information, hence evade long-term dependency problems. They can be well utilized in smart grid for energy management [8], stability prediction [9], demand forecast, network delay estimation [10] etc., efficiently.

1.3.2 Expert system techniques

An expert system [11] in AI also known as knowledge-based system that mimics the concluding capability of a knowledgeable human. The expert system replicates these knowledges and utilize it to solve analogous problems without human expert participation. Various extents of applications in power systems where a massive volume of data processing has to happen rapidly matching the ability of expert systems. As they are computer-based programs, the procedure of scripting codes is simpler than calculations and estimations used in generation, transmission and distribution of power system. It is dynamic and paves way for virtual estimations and even modifications after designing the programs (Fig. 1.2).

Figure 1.2 Structure of expert systems.

1.3.3 Fuzzy logic (FL)

FL approaches are different from modern computing. It depends on degrees of truth rather than the binary Boolean True (1) or False (0). It has a range of value where 1 and 0 acts as extremities with various intermediate degrees of truth. This feature helps in many engineering applications where the data is uncertain and imprecise. Many real time systems like temperature control, large scale power system control, system that demands reliability and safety control highly relies on FL engineering. Even, the rising complexity in power systems necessitate the application of FL in many power system glitches (Fig.