Energy storage is essential for navigating the intermittent nature of solar and wind power and, consequently, to the inevitable viability of renewable energy sources. The article provides a thorough overview regarding the implementation of artificial intelligence (AI), machine learning (ML), and other related technologies for maximizing energy storage in different ways.
Image Credit: KanawatTH/Shutterstock.com
Role of AI and ML in Improving Energy Storage
Energy storage is essential for determining the effectiveness, and stability of an electricity distribution system. Until now, dielectric capacitors (DCs) and lithium-ion batteries (LIBs) have been the dominant technological advances for storing electrical energy.
AI and ML are transforming the energy storage sector by enhancing the reliability and efficacy of energy storage technologies. These technologies employ algorithms that can analyze vast quantities of data, recognize trends, and make forecasts that can enhance the effectiveness of energy storage systems.
The prediction of energy usage trends is a significant advantage of AI/ML in preserving energy and optimizing the storage phenomena. The probing of the data on energy consumption enables AI and ML algorithms to efficiently predict the periods of maximum and low energy demand.
This enables the optimization of systems for supplying the optimum quantity of energy during peak demand intervals. In addition, the forecasting of weather patterns using AI and DL algorithms, which may assist energy storage systems regulate the unpredictable nature of green energy sources more effectively is a major benefit in the modern era of sustainability.
AI and ML for the Development of Novel Energy Storage Materials
The rise of machine learning (ML) has triggered an evolutionary era in materials science that can accelerate the research and development (R&D) of energy storage materials.
The integration of domain knowledge into artificial intelligence (AI) models could not only be used to comprehend the formulation, structural orientation, intrinsic attributes of the materials, processing conditions, and performance linkages, but also for property prediction, novel material discovery, and multi-functional performance optimization.
AI and ML have also contributed to the experimental procedure and characterization stage for revolutionary energy storage substances. The conventional experimental approach relies heavily on individual intuition and expertise, resulting in a tardy and costly cycle of research and development for energy storage materials.
AI and ML facilitate experimentation and characterization by, for example, investigating optimal formulations, refining the testing process, eliminating the need for extra equipment, minimizing time, and improving characterization methods.
A recent article published in Interdisciplinary Materials thoroughly overviews the contributions of AI and ML to the development of novel energy storage materials. According to the article, ML has demonstrated tremendous potential for expediting the development of dielectrics with a substantial dielectric constant or superior breakdown strength, as well as solid electrolytes with high ionic conductivity. These materials are extremely efficient at storing energy.
The dielectric constant ε is an essential design parameter for polymer dielectric capacitors (DCs). The inadequate thermal stability (or low glass transition temperature Tg) of polymer dielectrics makes it difficult to locate polymer dielectrics with the desired Tg.
A recently devised ML-based model can immediately predict the frequency-dependent ε and Tg of polymers.
The training data set included 1210 experimentally determined values at various frequencies and Tg. Using a sampling method and the Gaussian procedure regression algorithm, the model was then used to predict the ε and Tg of 11,000 candidate polymers realizable within the frequency range of 60 to 1015 Hz. Using the desired ε and Tg as screening requirements, five potential high-temperature capacitance polymers with ε > 5 and Tg > 450 K were formulates in the final step.
Utilizing AI for Battery Energy Storage Control
Globally, buildings utilize a significant quantity of energy and account for 30% of greenhouse gas emissions. Significantly more battery energy storage (BES) has been deployed in recent years to preserve the reliability of the electrical grid through instantaneous regulating of production and consumption. Moreover, an energy management system (EMS) is a useful instrument for monitoring, controlling, and conserving energy storage.
Several AI-based algorithms, such as genetic algorithm as well as machine learning (ML) computational models, including specialized reinforcement learning (RL) approaches and deep RL technology, have been implemented that optimize energy storage controls and improve energy efficiency while taking into account multi-energy resources, such as photovoltaic (PV) panels and BES systems.
A recent article published in the International Journal of Electrical Power and Energy Systems focuses on the development of an autonomous and real-time BES control based on an RL model for residential buildings equipped with photovoltaic cells and a BES system that are connected to the grid.
A repetitive time-dependent Markov Process was specially created for analyzing regular periodic trends in demand, cost, and energy storage. The Q-learning algorithm effectively employed the Markov Process, resulting in improved battery energy control and reduced electricity costs.
The simulation results supported the practicability of the recommended learning algorithm and demonstrated its efficacy in reducing periodic power bills and maximizing energy storage by increasing the individual state size of the unregulated variable by an adequate amount.
Development of Advanced Energy Management Protocol
An efficient and reliable energy management system enables maximum energy production, utilization, and storage by reducing losses. An article in Energies proposes a novel Energy Management Protocol (EMP) founded on an integration of Machine Learning (ML) with Game-Theoretic (GT) algorithms for regulating the charging/discharging of electric vehicles (EVs) from an energy storage system (ESS).
A mast was established at a site in the northern Tunisian city of Utique. The data points collected over a year were initially processed and then employed for developing the machine learning algorithm, specifically the SVR model.
In terms of gust speed forecasting, the ML algorithm exhibited excellent performance, particularly for predicting wind speeds days in advance. Based on recorded and projected wind speed, the RSE calculation yields a value of approximately 0.94.
The application of the GT model improved the administration of the charging/discharging process for EVs, resulting in a 44% increase in EV customer satisfaction.
To summarize, for the ongoing advancement of alternative energy streams and the decentralization of energy generation, energy storage systems are indispensable. Energy storage is expected to play a greater role in the transition to a more resilient and environmentally friendly energy system as technology continues to advance while expenses continue to decline.
References and Further Reading
Nieto, C., 2022. Artificial Intelligence in battery energy storage systems can keep the power on 24/7. [Online] Energy Storage.
Available at: https://www.energy-storage.news/artificial-intelligence-in-battery-energy-storage-systems-can-keep-the-power-on-24-7/
(Accessed on 4 May 2023).
Khabbouchi, I. et. al. (2023). Machine Learning and Game-Theoretic Model for Advanced Wind Energy Management Protocol (AWEMP). Energies, 16(5), p. 2179. https://www.mdpi.com/1996-1073/16/5/2179
Abedi, S. et. al. (2022). Battery energy storage control using a reinforcement learning approach with cyclic time-dependent Markov process. International Journal of Electrical Power & Energy Systems. 134. p. 107368. https://www.sciencedirect.com/science/article/abs/pii/S0142061521006074?via%3Dihub
Shen, Z. et. al. (2022). Machine learning in energy storage materials. Interdisciplinary Materials, 1(2), pp. 175-195. https://onlinelibrary.wiley.com/doi/10.1002/idm2.12020