Valo’s Steve Merrick on why accurate state of charge estimation remains energy storage’s ‘unsolved problem’

“For only a very minuscule change in voltage—a change so small that the sensors often can’t detect it—it might actually be like 30%, 50%, or 80%,” Merrick explains. “So we kind of get lost when trying to look up the voltage.”

This inherent characteristic, known as the open-circuit voltage (OCV) curve, means that battery management systems (BMS) struggle to accurately determine remaining capacity during normal operation. Various algorithms attempt to compensate, but the challenge remains.

When asked who should address SOC inaccuracy, Merrick notes there’s no industry consensus.

“I wouldn’t say that if you asked 100 people, ‘Family Feud’ style, that you would not get 100 people saying the same thing. I think there’s arguments for different parties fixing the issue. The entity most affected is typically the trader, who needs accurate data to bid into markets effectively. However, multiple parties have attempted solutions, including OEMs, integrators, and third-party battery analytics providers.”

One issue is that any improved SOC measurement must be integrated into the energy management system (EMS) to be useful.

“If you ever come up with a new measure, it has to really be embedded in that control system,” Merrick says. “The BMS might say 0%, so I’m not going to let you discharge anymore. If I come up with my own measure that says you actually have 10%, well, I’m not listening to you. I’m listening to my measure.”

Merrick suggests reframing SOC as “dischargeable energy at a specified power level,” which would also account for cell imbalance—another factor that limits usable capacity.

Stranded capacity

Stranded capacity—energy trapped in batteries that cannot be accessed—is closely linked to cell imbalance. In an ideal scenario, all battery cells would charge and discharge in perfect synchronisation. Reality is not as straightforward.

“One (cell’s charge is) lower, one’s higher,” Merrick explains. “When we’re discharging, the lower one hits the limit where the BMS is communicating, ‘If you go any lower, it’s a safety risk,’ so it shuts off. But this higher one still has energy left. It has more energy, but it can’t discharge because the lower one hit the bottom first.”

This imbalance can be addressed through SOC calibration, where the BMS rests at low SOC for approximately an hour to better assess individual cell voltages and develop a rebalancing plan. However, the problem recurs.

“It’s constantly a battle of I’m fixing it, but it’s accruing again. Kind of like fighting entropy,” Merrick says.

Some stranded capacity is unrecoverable when cells degrade at different rates, permanently shrinking to different sizes—a problem that can only be resolved through module replacement.

When monitoring becomes a revenue problem

While any amount of imbalance technically represents lost revenue, Merrick identifies specific thresholds where the issue becomes critical.

The first occurs when actual tradeable capacity falls below planned assumptions. “If I’m only able to trade on 85 (MWh) when I expected 96, then that’s a problem because it’s different than the expectation,” he notes.

Another critical point is the annual capacity test, where integrators must prove the system meets guaranteed performance levels under long-term service agreements (LTSAs).

“If we fail that, it’s a big deal because now the contract has been violated,” Merrick explains. “We as the integrator have to pay fees. The customer’s upset because they don’t have as much energy as they thought they would have.”

For tolling agreements, where asset owners provide a specified amount of energy to traders for a fixed payment, falling below promised capacity can trigger contractual disputes.

Cell imbalance and degradation can create a vicious cycle. Cells that consistently rest at very high or very low SOC degrade faster than those at moderate levels.

“If you end up having a case where only one cell is resting at 0% all the time over many years, and all the other cells are resting at 30%, that one that’s resting at zero is going to degrade faster,” Merrick explains. “That cell would end up limiting all those other ones long term because it’s going to shrink faster.”

Third-party analytics

Third-party analytics providers can identify imbalance and recommend solutions, but crucially, they cannot implement fixes themselves since they don’t control the sites.

“They can generate the insights, but they cannot actually enact the solutions,” Merrick says.

For improved SOC metrics, third-party providers face additional hurdles. Cloud-based SOC calculations have to continuously update as batteries charge and discharge, creating potential latency issues.

“The data from the battery has to transmit to the cloud. The cloud has to run an algorithm and then estimate the SOC. There’s a lag there,” Merrick notes. “Is that fast enough for traders who are operating in real time?”

He believes more accurate SOC metrics should be embedded directly in hardware rather than calculated in the cloud, both to eliminate latency and ensure the measurements can actually control system behaviour.

Over the past three years, Merrick has observed slight improvements in SOC accuracy from third-party BMS providers, though he couldn’t quantify the gains. Valo has also launched its own BMS, giving the company more direct control over SOC estimation.

However, the techniques—coulomb counting and Kalman filters—remain largely unchanged.

“Some slight improvement, but it still remains a significant issue in the industry,” Merrick says. “We have reams of hundreds and thousands of data scientists around the world, but it’s still an unsolved problem.”

Note: this interview took place shortly before Wärtsilä announced the rebrand of its energy storage division as part of a new joint venture (JV) with RCT Solutions, under the new brand name, Valo.