What Is Remote Diagnostics?
I got into this business back in 2016 when nobody really talked about remote diagnostics for batteries. We were still pulling packs out of buses and hooking up bench testers. Now I spend half my week staring at dashboards.
Remote diagnostics means you can see what's happening inside a battery pack without touching it. The BMS collects voltage and temperature from every cell, packages it up, sends it over cellular to a server somewhere. That's the simple version.
The complicated version involves a lot of arguing about what data actually matters.
The basics
Every cell in a pack has a voltage. In a 400V bus battery you might have 96 cells in series. The BMS reads each one, usually every few seconds. It also reads temperature from thermistors scattered around the modules. Maybe one sensor per module, maybe more if the pack designer was paranoid about hot spots.
All this goes into a data stream. The stream gets compressed and pushed out through a modem. LTE works fine in most cities. Out in the sticks you might need satellite but honestly most fleet operators just accept the data gaps when their trucks hit rural routes.
The receiving end is whatever cloud service the OEM set up. Could be AWS, could be Azure, could be some proprietary thing nobody wants to maintain. The data lands in a database and then algorithms chew on it.
Data Pipeline
BMS Collection (V, T, I)
Compression & Encryption
LTE/Sat Transmission
Cloud Ingestion & Analysis
What we actually look for
Cell voltage spread is the big one. If you have 96 cells they should all be pretty close to each other when the pack is resting. Maybe 10 or 15 millivolts apart. When one cell starts drifting low compared to its neighbors, that cell is degrading faster than the rest. It's going to limit your pack capacity eventually.
Temperature
Temperature is trickier than people think. A sensor might read 35°C but the actual cell surface two inches away could be 40°C. You're interpolating. You're guessing. The algorithms account for this but there's always uncertainty.
Internal Resistance
Internal resistance is the gold standard for health assessment but most production BMS systems don't measure it well. You need a controlled current pulse and a clean voltage reading. Hard to do when the vehicle is bouncing down the highway. Some of the newer systems do it during charging when everything is stable.
The part nobody tells you
Half the alerts I see are garbage. Sensor failures, loose connections, software glitches in the BMS firmware. A technician drives out to check a battery that the system flagged as critical and finds nothing wrong. The thermistor had a bad solder joint. Three hours wasted.
Good remote diagnostics requires filtering. Not just filtering the data but filtering the alerts. You need to correlate across multiple signals before you bother anyone. One high temperature reading means nothing. Three consecutive readings trending up while current is low and ambient is normal, that's worth a phone call.
I worked with a transit agency in Ohio for about two years. They had 40 electric buses and were getting maybe 200 alerts per week when we started. Most were false positives. We tuned the thresholds, added some time-based logic, got it down to about 30 alerts per week. Of those, maybe 20 were actionable.
The money question
Fleet managers always ask about ROI. Here's the honest answer: it depends on how bad your baseline is.
- If you're already doing monthly inspections and catching problems early, remote diagnostics might only save you a few roadside failures per year. The subscription costs might not pencil out.
- If you're running packs until they die and then reacting, remote diagnostics will pay for itself fast. I've seen fleets avoid catastrophic failures that would have meant tow trucks, fire department calls, and replacement packs on emergency shipping. One avoided thermal event covers a lot of monthly fees.
The harder savings to quantify are the warranty claims. When you have continuous data you can prove exactly how the pack was used. Did the customer abuse it with fast charging in hot weather? The data shows it. Did the manufacturer ship cells that were already degraded? The data shows that too.
Where this is going
"The machine learning stuff is overhyped in my opinion."
Electrochemical impedance spectroscopy keeps coming up in research papers. The idea is to inject AC signals at different frequencies and measure the response. You can detect things like lithium plating before they show up in voltage data. A few companies are trying to productize this but I haven't seen anything in volume production yet.
The machine learning stuff is overhyped in my opinion. Yes you can train models on historical failure data. But batteries fail in new and creative ways. A model trained on 2020 data might miss failure modes that show up in 2024 chemistries. You still need humans looking at the data and asking questions.
What I actually want is better sensors. Give me direct measurement of pressure inside the cell. Give me gas sensors that can sniff early electrolyte breakdown. The BMS data we have now is like trying to diagnose a patient's health by only measuring their temperature and pulse. Useful but incomplete.
Practical advice
If you're spec'ing a battery system and the supplier offers remote diagnostics as an option, take it. Even if you're not sure you'll use it. Having the data retroactively is valuable. Trying to reconstruct what happened to a failed pack without telemetry is painful.
Make sure you understand who owns the data. Some suppliers keep it locked in their platform and charge you to access your own information. Get data export rights in the contract.
Ask about cellular costs. Some systems include data plans, some don't. A few hundred devices transmitting constantly adds up.
And budget for someone to actually look at the dashboards. The fanciest diagnostics platform is worthless if nobody logs in.
