Our article featured on Energy Storage News, The State of BESS Analytics: Three Views from Smarter E 2026, made the case that battery analytics has become a decision layer for energy storage. This piece goes a level deeper. At Smarter E 2026, PowerUp’s Chief Revenue Officer, Philippe de la Fortelle, sat down with Jacob Yang, Marketing Consultant at PowerUp, to walk through what actually happens inside a battery when something starts to go wrong, and what a platform like Battery Insight® sees that a threshold alarm does not.
Why a Battery Is Harder to Read Than a Solar Array
A solar or wind asset is largely mechanical. A battery works by chemistry, and that makes it a different kind of problem to operate. Philippe described a battery as starting life with a full tank of energy and giving up capacity as it ages and as it is used, with the size of the tank shrinking over time.
“A battery is a complex system because of its electrochemical nature. It’s not purely mechanical.” – Philippe de la Fortelle, Chief Revenue Officer, PowerUp
The trouble is that the shrinking is not always smooth or expected. Aging can run faster than the model predicts, and the reasons sit at the chemical level, where they are invisible to a system watching for a threshold crossing. Reading a battery well means reading the chemistry, which is why PowerUp builds on electrochemical models first and adds machine learning on top rather than the other way around. That reflects how the team is built, with electrochemists and data scientists at its core rather than software engineers alone.
Two Kinds of Anomalies: Performance and Safety
Anomalies fall into two broad groups.
Performance anomalies erode what the asset can deliver. A section of the system ages faster than the rest, capacity drifts down, and the asset stops meeting the numbers its financial model assumed.
Safety anomalies are the other branch, where the risk is physical rather than financial. An internal short circuit, or a hotspot created by a poor weld, concentrates energy where it should not be. Left unaddressed, that can escalate toward thermal runaway, the point at which the energy packed into a cell releases uncontrollably and starts a fire. The path from a small fault to a fire is what makes early detection critical for owners and operators.
The two problems that follow are one of each kind: an imbalance that drains performance, and a thermal fault that creeps toward the safety line.
How a Weak Cell or Rack Limits the Whole
A battery delivers as a group, so its usable output is limited by its weakest members. When a cell or rack drifts out of balance with the others, that gap caps what the whole system can do, and it is where threshold-based monitoring shows its limits.
A cell imbalance is local. A single limiting cell inside a module underperforms, and because cells work together, it holds back the cells around it. An inter-rack imbalance is larger and easier to miss. One rack falls out of step with the others, and because racks feed a shared power conversion system (PCS), a single weak rack can constrain the output of the whole group. The racks can each look healthy on their own while the system leaves energy on the table.
The BMS is not built to catch this. It watches values against thresholds inside its own local scope, so a slow divergence between racks that never trips a limit can persist unnoticed. Analytics working across the full system can flag the divergence early, identify the likely cause, localize which rack or module is responsible, and suggest the next best action, while the operator makes the call.
PowerUp’s white paper When Healthy Racks Can’t Deliver works through the inter-rack case and the energy capacity (and revenue) it can strand.
The HVAC Case: A Fault That Hides Below the Alarm
Thermal management is the clearest example of a problem that builds below the safety line, and PowerUp’s published HVAC analysis lays out how it happens.
Under normal use, current moving in and out of the cells produces heat, the Joule effect. The HVAC system holds the container in a working band, generally around 25°C give or take a few degrees. When cooling starts to underperform, the temperature drifts upward. It can climb well above the healthy band while still sitting under the BMS safety threshold, so the BMS has no reason to act. The system runs hotter than it should, and heat is one of the main drivers of battery aging.
If the drift continues unchecked, it approaches the critical range, around 50°C, and beyond that the thermal runaway region, generally above 90°C depending on chemistry. Philippe estimated that around 99% of the industry’s safety investment today goes into hardware that contains a fire once it starts. That hardware is essential. PowerUp works earlier in the sequence, on anticipation.
Battery Insight compares live temperature behavior against a digital-twin model of the system and flags the deviation as soon as it departs from what the model expects. In one case PowerUp documented, a cooling failure in a single container produced a temperature drift between 30°C and 50°C that ran for about ten days before the racks reached their 55°C shutdown. The outage did not stay contained. Shutting down that container also took down a neighboring one tied to the same delivery point, which had no cooling problem of its own. Detecting the fault on day one, rather than a week and a half in, is the difference the analysis is built to make. In a broader review of eight utility-scale sites, PowerUp found HVAC-related thermal anomalies at 75% of them, most appearing days or weeks before any safety threshold, a finding featured in kWh Analytics’ 2026 Solar Risk Assessment.
Two Segments, One Method
The same electrochemical approach travels across the two markets PowerUp serves. Stationary storage is the grid-scale world this series focuses on. The second is electric vehicle fleets, where the underlying question is the same: read the chemistry, catch the deviation early, and give the people responsible for the asset time and information to act.
What Advanced Analytics Adds
The payoff shows up on the balance sheet. Philippe tied the analytics to the three things owners track: availability, capacity, and lifetime. Caught early, an imbalance can be corrected to recover stranded capacity, and a cooling fault can be addressed before it forces a shutdown or accelerates aging. Left alone, the same problems surface as lost revenue and earlier replacement.
None of this replaces the BMS or the fire-suppression hardware around it. Both do essential work at the threshold. Analytics adds a second layer that reads the chemistry underneath and extends the window to act, from the imbalance that strands capacity to the cooling fault that hides below the alarm. The platform surfaces the deviation, the likely cause, and where it sits. Then, the operator decides what happens next.
Watch the full conversation with Philippe de la Fortelle, Chief Revenue Officer, and Jacob Yang, Marketing Consultant, at PowerUp.
