ampBI with Digital Twin TechnologySanjay Sharma
By Dr. Abhishek Singh, Battery Simulation Expert at AMP
ampBI – ‘The Battery Intelligence and Monitoring Platform’, is backed by digital twin technology. With each passing day, global oil resources are drying up while greenhouse gases emission is increasing. We need a sustainable environment, and a low-carbon future can be made possible by lithium-ion batteries, which have a variety of uses, such as grid-scale energy storage and electric vehicles. The materials utilized, the system architecture, and the operating circumstances all have a significant impact on how long these devices last. Because of this intricacy, controlling battery systems in real life has proven difficult. However, there is a chance to combine this knowledge with cutting-edge machine-learning techniques to create a battery’s digital twin (DT) by understanding battery degradation, modeling tools, and diagnostics. Battery monitoring and analytics, when powered by battery digital twin technology, can optimize the battery lifecycle, and make it safe for operations.
“Digital Twin is a system-based integrated multi-physics, a multi-scale simulation that employs the physical models, sensor data, and historical data to mimic the behavior of its physical twin” as quoted by NASA. Digital twin offers insights into the aspects which are difficult to understand with the measured outputs and helps us solve the bottleneck of the current battery research.
Battery Digital Twins
When we observe a physical cell/battery, we see many specifications written over it in terms of capacity, voltage, power, etc. Let’s suppose, we have a cell with a nominal voltage of 3.6 V. If we operate it above 3.6 V; is it going to fail? Not always. Similarly, if I over-charge or over-discharge a cell, it won’t fail. This means cell voltage limits, and charge/discharge parameters are not the only criterion through which we can predict battery behavior exactly or accurately. High non-linearity in the measured parameters such as voltages, internal resistance, etc. owing to the various internal exothermic reactions, predictions of the states, and establishing the models for the batteries control is a difficult task. The use of battery digital twins serves this purpose along with many other functionalities such as fault diagnostics, charging optimization, product design, and optimization.
These functionalities of a digital twin for battery systems and accurate battery diagnostics involve modeling of battery dynamics. There are three main methods for modeling cell dynamics:
(i) Electrochemical Modeling
(ii) Equivalent Circuit Models (ECM)
(iii) Machine Learning Models (ML) or Data-Driven models
|Equivalent Circuit Model||Electrochemical Model||Data-Driven Model|
|Ut = f(Uoc(SOC), I, R, C)
|Ut = n.fPDEs
High accuracy of voltage calculation
|Ut= f(I, SOC, T)
Popular owing to the simpler mathematical equations
|Widely used in SOC estimation||Require prior knowledge of the battery||Lacks physical meaning|
|Complex parameter identification process||Time-consuming||Less computational time|
|Lack of physical meaning - Internal cell kinetics and dynamics cannot be predicted or studied.||High computational costs|
|Unsuitable for control applications||Large number of parameter extraction which governs the geometric, kinetic, and transport phenomena within the cell|
The inadequacy of ECM and Data-driven models in studying/predicting internal cell kinetics and dynamics, limits their usage. The use of electrochemical models fulfills all the above requirements. These models help us in predicting cell degradation as well as the degradation mechanisms including SEI layer growth, loss of active material, pore blocking, lithium plating, active material dissolution, and lithium plating. And then, even more importantly, control the battery management parameters to avoid future degradation models.
A digital twin of the battery requires all the multi-physics models to work together and update the phenomenon in real time. It includes the thermal characteristics, electrical characteristics, and aging phenomenon. The implementation of the electrochemical models in real-world scenarios is quite difficult owing to the large parametric updation (models are too slow for real-time estimation and control in battery management systems), and hence DT for estimation purposes is mainly built using ECM or ML models.
Machine learning models utilize raw experimental data as input and then perform pre-processing operations for feature extraction. The extracted features significantly affect the estimation process as the model accuracy is directly related to it. The trained model can be utilized for state estimation and remaining useful life prediction.
Comparison of ECM and Electrochemical (P2D) model
In recent years, efforts are made, and many models are built which integrate the data-driven models with electrochemical models to update these parameters in real time at a much faster rate. Thus, overcoming the lacunas of both approaches. We can implement highly accurate and closely following SOC and SOH estimations methods based on the cloud computing platform by utilizing the Internet of Vehicles to accomplish the cloud transfer of vehicle information.
A vehicle-level digital twin can also be utilized for the thermal management of the batteries where the electrochemical models are integrated with the thermal models to determine the maximum and minimum temperature of the battery pack for different drive cycles. This helps in the design of thermal controls based on real-time updates and feedback.
The EV market is skyrocketing. With the advancement of the technology and growing need for electric vehicles among consumers, we require new methods/techniques which can limit the testing and help us predict the behavior of the battery/cells accurately.
Could-based smart BMS, Digital Twins can overcome the shortcomings of the current BMSs computing power and data storage. Additionally, it would enable the creation of additional sophisticated BMS functions and result in more precise and dependable battery algorithms.
DT serves the mentioned purpose and can be utilized by the battery/electric vehicle manufacturers to analyze the batteries’ range and power under different operating conditions and optimize them to offer warranties. For the companies working on batteries’ second life, digital twins will help them ensure that the batteries are running under optimal limits.
ampBI – The ‘AMP Battery Intelligence’ Platform
At AMP, we are constantly innovating to revolutionize battery intelligence technology. ampBI, our battery intelligence platform, is progressing to make use of digital twins to uncover and resolve the dilemmas related to battery life and efficiency by understanding the battery envelope which often leads to battery imbalance. It makes use of electrochemical models, machine learning algorithms, and historical data to predict the behavior of batteries in real-time scenarios.
ampBI, our cloud-based Battery Analytics as a Service platform, is highly flexible. It integrates with existing telematics and battery management signals to provide meaningful insights optimizing the SoH (State of Health) of the battery pack. This ‘easy-to-use’ platform interface is equipped with interactive dashboards enabling the users to take preventive and corrective measures saving irreversible damage to the battery’s health.
ampBI offers information about battery kinetics and dynamics at the microscopic level with the novel developed Electrochemical-ML models. These models can be utilized for state estimation as well to optimize battery performance and develop thermal controls to avoid conditions of thermal runaway to answer all of your HOWs, WHY, and WHATs like:
- Why is this battery fading faster?
- What is the battery temperature?
- Which is the worst cell in my battery pack?
- How to increase battery performance?
- How much warranty should be given on this battery?
- Should I change my battery design?
And many more…
Contact us today to schedule a demo of AMP’s cloud-based battery monitoring and analytics platform, ampBI.,