Within our R&D team, we spend surprisingly little time on batteries. That may sound strange for a company called Big Ass Battery, but it is the reality. A large part of our time today goes into software. Into algorithms that learn to understand energy assets. Into machine learning that detects anomalies. Into AI that weighs thousands of parameters simultaneously and makes autonomous decisions. In this blog, we explain how we use AI to keep developing our energy management system.
At its core, a battery does one thing: it stores energy and releases it again. That principle hardly changes, even as batteries become more powerful, compact and affordable.
The complexity only really begins when that battery becomes part of a much larger system. In a modern business environment, it has to continuously take dozens of factors into account at the same time, such as:
- Available solar generation
- Dynamic energy prices
- Grid constraints
- EV charging hubs
- Production planning
- Weather forecasts
- Generators and other energy assets
This is no longer a challenge that hardware alone can solve. And that is exactly where today’s biggest R&D challenge lies.
Why we are moving away from static rules
Many energy management systems still rely on a traditional rule-based approach. Essentially, that means simple logic such as:
- If the battery drops below 20%, charge it.
- If the sun is shining, charge the battery.
- If consumption rises, discharge the battery.
That works perfectly well as long as reality follows those rules neatly.
But energy is rarely that predictable. What if the weather forecast shows that a lot of solar generation is expected within the next hour? What if electricity prices are about to turn negative? Or what happens when three electric trucks return to the site at the same time?
At that point, fixed rules reach their limits. That is why our energy management system interprets the context itself and uses it to make better decisions.
Every asset behaves differently
On paper, one battery may look identical to another, but in practice every installation behaves differently. Temperature, age, usage profile and charging cycles all affect performance.
That is why we are researching how machine learning can be applied directly at asset level. Through the Load Commander, the system learns how a specific installation behaves in practice. Not just which technical specifications it has, but above all how it actually performs under different conditions.
This approach offers several benefits:
- Less manual configuration
- Automatically generated asset profiles
- Faster adaptation to new battery technologies
- Longer-term opportunities for predictive maintenance
The last point is still in the research phase today, but it clearly shows the direction in which intelligent energy platforms are evolving.
AI only becomes truly valuable when it connects the dots
Machine learning on a single battery is valuable, but a business rarely relies on just one energy source. The real challenge arises when dozens of different assets need to work together at the same time.
That is why we are developing a central AI agent that continuously combines information from, among other things:
- Historical and current energy consumption
- Battery status
- Weather forecasts
- Energy prices
- Grid constraints
- User preferences
- Available energy sources
The system does not look at one parameter at a time, but tries to understand the full context. It then determines which action creates the most value at that moment.
That may seem like a small difference, but for us it is fundamental. We want to evolve from software that reacts to software that anticipates.
Getting different energy systems to work together is the real challenge
When people think about software development, they often think of dashboards or user interfaces. In reality, a large part of our R&D focuses on something more fundamental: enabling very different energy assets to communicate with one another.
Every manufacturer uses its own:
- Communication protocols
- Data structures
- APIs
- Methods for controlling assets
Our ambition is not to make one battery smarter, but to enable completely different systems to operate together as one integrated whole.
That is why the Load Commander plays such an important role. It forms the bridge between different energy assets and the energy management system, so information is no longer locked away in separate systems.
The hardest innovation? Keeping it simple
Perhaps that is ultimately the biggest challenge. Not the AI. Not the cloud. Not the algorithms. But simplicity.
The smarter an energy system becomes, the less the user should actually notice it. A business owner does not want to manage hundreds of parameters. They simply want to choose what matters most today:
- Maximise production
- Minimise energy costs
- Maximise the use of self-generated solar energy
- Guarantee security of supply
The software should then make the optimal choices automatically. That is ultimately what our R&D is working towards. Not more technology at the front end, but a system that resolves all the complexity in the background.
Because we believe that is the future of energy management: not making more technology visible, but making sure you barely have to notice it at all.
