Advancements in battery management software and predictive analytics are impacting the electric vehicle (EV) landscape, enabling smarter, more efficient solutions to enhance performance, reliability, and cost challenges. As EV adoption accelerates, leveraging intelligent systems to optimize charging, enhance battery lifespan, and improve safety has become a critical focus for the industry.
Recently, we had the opportunity to speak with Dr. Nadim Maluf, CEO and co-founder of Qnovo, a battery software provider. Maluf brings over 30 years of experience in engineering solutions, including those focused on e-mobility. His work centers on advancing EV battery management software and predictive analytics to achieve faster charging and extended run times without compromising battery lifespans.
In this Q&A, he shares his expertise in leveraging AI, machine learning, and software to address challenges, such as improving state-of-health accuracy, bridging lab-to-field performance gaps, and enhancing EV battery safety.
Here’s what he has to say…
How are artificial intelligence (AI), machine learning (ML), and software advances changing the testing and reliability of EV batteries and charging infrastructure?
Dr. Nadim Maluf (NM): Let’s separate software and AI/ML:
1. Software can drastically improve an EV battery’s reliability and the charging infrastructure’s utility. For example, it can change the rate of charge, identify damaged battery cells or potential risks within the battery, and offer suggestions to drivers that can extend lifespan and performance. By optimizing the software behind the battery, manufacturers can increase the lifespan of electric vehicles by up to 20%.
2. AI/ML builds upon that base framework. They’re additional tools in the arsenal for battery performance and reliability. However, they do not replace the need for fundamental physics-based models.
They’re another layer on top to assist in:
- Processing the large amount of data that is being collected in vehicles and batteries
- Developing complementary models that cannot be addressed with physics only, such as manufacturing variability, variances in driver behavior, or diversity of use cases.
From an engineering perspective, what are the most significant challenges in achieving accurate state of charge (SoC) and state of health (SoH) estimates during real-world EV operations?
NM: The top challenge in SoC, SoH, and all SoX (the various battery health and performance metrics) is understanding the battery’s aging. No two batteries are alike. They’re manufactured differently. As they’re used differently, they age differently. This is where software comes into play.
Intelligent models built into the software — layered on continuous battery diagnostics — allow for a real-time understanding and even prediction of the battery’s aging. Once aging is characterized and understood for each battery, each pack, and each vehicle, one can then make the proper adjustments and provide a more accurate SoC, SoH, and all SoX.
Analogy: think of aging in humans. Is age in years a good measure of health (SoH for humans)? Doctors will most certainly tell you no. So, how do you measure aging, then? We measure it with a laundry list of other metrics like blood pressure, cholesterol panels, diabetes screening, etc., which collectively provide a more concrete picture of the individual’s health. The same concept applies to batteries.
How can engineers ensure real-world EV battery performance matches predictions made in controlled lab environments — and what supporting roles do data and machine learning play?
NM: Performance in the lab and the field are different. Full stop. The lab data and results are only to frame the engineering work so that tests in the field (road tests, field tests) can validate the theses established in the lab.
Data is important. Very important. But what kind of data do you collect? How do you parse it? Taking lots of data and feeding it into an ML engine in the hope that it finds patterns is massively inefficient. So this is where physics, lab test data, and experience come into play to decide what data should be collected and analyzed.

EV battery software offers several advantages, such as detecting faults, optimizing charging, improving performance, and even extending vehicle lifespan by up to 20%.
Ultimately, ML is not very good at predicting rare corner cases that can lead to disasters like fires. Predicting battery failure is one of the most essential development topics, and innovative solutions are coming on the market that blend physics-based models with data and ML.
With advancements in silicon anodes, solid-state batteries, and sodium-ion technologies, how must software adapt to support these chemistries?
NM: Silicon anodes and solid-state batteries share the same fundamental physical mechanisms as the conventional lithium-ion battery (LiB). Adapting these to new technologies is not overly complicated. It involves understanding the material properties and battery design parameters. Sodium-ion shares many similarities with LiB but also has distinct differences, especially in relation to mechanical and thermal properties.
Ultimately, the main challenge is that battery vendors (and entire countries ) are becoming increasingly secretive about their designs. World leaders are expecting a “US vs. China” trade war. They’re looking to their technology sectors to ensure their economy is the one standing. There’s no benefit to sharing those designs when global competition is so fierce. This means the software must become better at understanding the inner workings of these batteries without recourse to the battery vendors.
Engineers often face a trade-off between energy density, power delivery, and cycle life. How do predictive algorithms assist in striking the ideal balance for EV battery design?
NM: As energy density increases, the lithium-ions shuttling back and forth between the electrodes have many more opportunities to do damage — for example, bind together to create lithium plating or bind with the electrolyte to make a thicker SEI layer. That damage accelerates at high powers, current levels, or under certain environmental conditions.
Cycle life is only a rear-view image of the battery’s degradation. These relationships are highly non-linear and depend on factors such as environmental conditions, current and voltage levels, material properties, presence of minute defects, etc.
Striking a balance requires intelligent software with exceptional battery models, data measurement, and real-time optimization. Software must determine the optimal operating point for this specific vehicle and maximize for those parameters. For instance, in a commercial vehicle, the greatest need is a long cycle life. For a taxi, it may be the fastest charge time possible.
The ability to create these optimizations in real-time and flexibly adapt them to a changing use case is the power of “adaptive” and “predictive” algorithms.
Given the push for battery safety certifications like UL2580, how can software tools help engineers align with stringent global safety and compliance requirements?
NM: Some of these standards are unique to how a battery is certified during and after manufacturing. For example, can it withstand a specific temperature range? Or perhaps the most stringent is whether it can withstand a nail penetration test. These are good standards but are also insufficient.
Consider the following case: Suppose a battery meets all these standards from a reputable manufacturer. It enters service in a smartphone. Several months later, some smartphones caught fire. It turns out that the cavity holding the battery in the device was a little too tight on some phones. The cavity put stress on the battery, which in turn caused a location where lithium metal dendrites could grow — ultimately leading to an internal short circuit.
This is what happened in 2016 to the Samsung Note 7. Could software have detected these anomalies? Absolutely, yes. It could have alerted the manufacturer that a pattern of malfunction was occurring in specific models or alerted the user to the risk. It would have been a simple fix for what became a much larger problem.
My point is that software exists not to replace these standards but to maintain them. Often, these standards are applicable at the beginning of the battery’s life. Software ensures compliance from this beginning point through the middle and end of the battery’s life.
As circular battery ecosystems become more critical, how can software analytics improve battery pack designs for easier disassembly and end-of-life material recovery?
First, you cannot re-use a battery in a second-life application unless it works and is safe. How do you determine that? Today, it’s challenging. However, onboard battery software or cloud-based analytics can quickly provide an answer. They can look at a battery down to the cell level, check for defects and risks, and determine remaining functionality. It’s essential to determine if the battery should be considered for subsequent use or should be taken to recycling.
Once this data is accumulated, connecting the use case needs (including end-of-life requirements) to the battery’s design becomes possible. Battery design includes the cell design itself and the pack design. In some cases, it’s the “cell-to-chassis” design, where the battery and cell are directly attached to the chassis.
Analytics data captured during the battery’s life can then be called on to optimize secondary functions — including disassembly. However, the industry is nowhere near to applying these concepts yet.
How do you see the evolution of vehicle-to-grid (V2G) technologies, and what role will EV batteries play in stabilizing energy grids and advancing renewable energy adoption?
NM: V2G is quite promising. Currently, the EU is testing various pilots, but the US is behind. Batteries can play a significant role in supplying energy to grids or local properties.
Imagine a Ford F-150 Lightning with a 120 kWh pack. It can supply energy to a house for several days! The challenge is primarily the economics. Who would pay for this energy? How would you price it? Who would connect the provider with the user at such a distributed scale? It requires policymakers to get involved, which is something that’s unlikely in the near future, unfortunately.
How will EVs evolve in the coming years, and what strategies or technologies will likely drive battery cost reductions?
NM: There are three main factors that can accelerate the adoption of EVs:
- Affordability
- Seamless charging experiences, which requires reliable infrastructure
- Safety and reliability of batteries.
If the industry can nail these, EVs will be healthy without the need for incentives. But we have a lot of work to do to get there.
Filed Under: Batteries, FAQs