The electrical vehicle (EV) industry’s disruptions are now heralding a new era in the energy storage business, which is driving up demand for battery management modelling. As if that weren’t enough, the EV market is being propelled into the forefront of energy storage by a mixture of economic and environmental factors, offering strong incentives to advance the study of battery management systems. As a result, there has been a huge demand for extended cooling systems and increased battery range, as well as major investments in the EV industry. These reasons have caused an exponential growth in the pressure put on battery research to improve the precision and effectiveness of management systems.
But as things stand, there are a number of problems with EVs and the current battery base. There are a variety of models for modelling car batteries that can forecast things like voltage, current, battery temperature, overpotential, state of charge, condition of health, and other things. The governing physics and chemistry of batteries are non-linear, which makes it difficult to describe them accurately even though these factors are crucial to battery longevity and performance. This is particularly true for predictions made over a longer period of time, which conflict with the present push by EVs to extend battery life.
Where we Are
Lithium-ion batteries are the most well-known BESSs because of their widespread application in electric vehicles and mobile phones. It is now the preferred storage technology for large-scale facilities that aid in supplying power networks with a steady stream of renewable energy.
Lead-acid, nickel-cadmium, sodium-sulphur, and flow batteries are some of the alternative technologies being developed by the scientific and engineering communities to address the drawbacks of lithium-ion batteries, including their flammability, high cost, and extreme temperature intolerance.
To make predictions about battery performance, it is a difficult and dynamic problem to combine electrochemistry with microscale physics. The technical aspect of model parameterization causes the modelling complexity since the model must take into account dependencies that occur at both the macro and micro scales. Lithium conductivity and diffusivity, for instance, are influenced by lithium content. For battery model multiplex modelling, the technological multiscale dependencies provide a hurdle. Therefore, the infrastructure for battery research is not frugal in the number of experiments assigned to and carried out.
Recoursing to Technology for Battery Intelligentisation
There is a natural demand for ML techniques because of the model sensitivity and linked non-linear complexity. For machine learning to process and optimise, alternative physics battery modelling libraries offer a mountain of data. This is strengthened further by the fundamental characteristics of sophisticated deep learning algorithms, which permit the identification of higher dimensional correlations that are challenging to parameterize.
When applied to battery modelling, machine learning techniques may generate crucial diagnostics with a fair amount of accuracy and speed. Consequently, it provides an improved simulation environment for engineering users that want to carry out experiments or optimizations. Users of battery research may be able to examine analytical correlations inside the cell architecture thanks to emerging capabilities, which might speed up the process of creating high-performing batteries.
Owing to the maths guiding the algorithms, machine learning models benefit from access to huge datasets. This makes it possible to efficiently incorporate the massive datasets obtained during battery or vehicle driving testing. The acceleration, ambient temperature, velocity, and state of charge are just a few examples of the information that may be found in real-time data that is produced from vehicle driving tests. These variables are all dynamic functions of time. The demand for AI applications in this sector is sparked by our growing understanding of the quantitative linkages that develop over time between these variables and one another. The appeal of AI in this situation is that it may help us grasp the intricate links between the characteristics in higher dimensions, which would improve our knowledge of real-world physics.
Machine learning models may leverage the large volumes of data generated by battery testing equipment to identify relationships. Physical models may be intelligently improved by understanding the relationships between the mechanics defining them, which can enhance battery performance. The goal of the convergence of AI and battery physics modelling at this time is to improve the user’s capacity to design modifications inside the battery cell intelligently. The non-linearity and complexity of controlling chemistry and physics on the atomistic scale provide an additional benefit of combining AI with physics-based modelling. The modelling process is complicated mathematically by the high level of parameterization. Therefore, in the future, deep learning algorithms may be added to produce digital battery twins.
How Technology is positioned to Enhance BESS
Operations and maintenance in the green energy industry have historically been seen as fundamental duties like cleaning, maintaining panel output, checking the health of inverters, etc. Numerous use cases have evolved as a result of the addition of BESSs, such as the capacity to optimise storage for cost savings, enhance system resilience and sustainability, and gather information for predictive analysis. AI is crucial because it is now impossible for human operators to process the enormous amounts of data that arrive constantly. Let’s examine how AI can improve BESSs in more detail:
Managing Energy Demand
A BESS system balances load during peak and off-peak periods based on the day, time, season, or weather using artificial intelligence (AI). As an illustration, the Evergen app makes use of AI to track weather predictions and determine how much energy needs to be conserved. Later, when the installed solar systems are unable to provide enough energy on cloudy days, the system distributes stored electricity.
Arbitrating Energy
Energy arbitrage, commonly referred to as time-shifting, is made possible by AI in reaction to changing electricity costs. Energy is bought at a lesser price during off-peak hours, then sold or used when the price rises. Therefore, regardless of the season or electricity demand, AI energy storages can equalise energy prices and reduce risks.
Forecasting Weather Conditions
Estimating the quantity of renewable power that will be generated today or tomorrow is not always simple. To create forecasts about energy output and consumption, AI systems combine meteorological and satellite data with numerical climate prediction models and statistical analysis.
Predictive Maintenance
Operators can take rapid action for preventative maintenance thanks to AI energy storage. AI learns to recognise common faults and abnormalities in a variety of subsystems (electrical, chemical, and thermodynamic) by collecting data from various sensors and comparing it with past data, alerting operators before a breakdown takes place. This reduces downtime, lengthens the life of storage systems, and ultimately boosts revenue.
The Way Forward with Emerging Technologies
Now more than ever, the EV sector is the primary driver of the demand for powerful and durable batteries. To build the best battery systems, many experiments are still necessary because the physics of these complicated systems make battery performance models inadequate. AI can help in this situation. AI can better advise battery designers on what their next move should be to create reliable and effective batteries by collecting higher dimensional correlations from data acquired from driving testing and other physical tests.
Authored By: Sarika Bhatia, Whole-Time Director, Servotech Power Systems Limited
