Choosing BESS Components: Cells, PCS, Inverters

Start with the application, not the brand

Before you look at any datasheet, define the duty cycle. Are you shaving peak demand, providing frequency regulation, or doing arbitrage? Each application dictates the C-rate, cycle life, and energy-to-power ratio. For peak shaving, you need 2-4 hours of discharge; for frequency regulation, 15-30 minutes. That changes cell selection completely.

Ask for the facility's load profile — at least 15-minute intervals for a full year. Without it, any capacity calculation is a guess. We've seen clients order 1 MW / 1 MWh systems, only to find their peak is 30 minutes long, and they could have used a 0.5 MWh battery with a higher C-rate, saving 30% on capex.

Cell criteria: capacity, cycle life, temperature behavior

For LiFePO4 cells, the key parameters are nominal capacity (Ah), cycle life at 80% DoD (typically 4000-6000 cycles), and operating temperature range. In Ukraine's climate, cold performance matters — below 0°C, charging must be limited unless the battery is heated. That's why we prefer liquid cooling or at least built-in heating pads.

Also check the cell's internal resistance and self-discharge rate. Lower resistance means less heat and better efficiency. But don't chase the highest cycle life — it's often at the cost of energy density, which increases cost. A 6000-cycle cell is overkill for a system that cycles once daily for 10 years; 4000 cycles might suffice.

Beware of cells with unspecified tolerance on capacity — good cells have ±2% or better. Also verify the cell's certification: UL1642, IEC62619, UN38.3. That's non-negotiable for insurance and grid interconnection.

PCS and inverter: voltage, efficiency, grid code compliance

The power conversion system (PCS) must match the battery's DC voltage range and the grid's AC voltage. For industrial systems, we typically use 800-1500 V DC bus. At 0.4 kV, currents get too high for large capacities — above 500 kW, you'll need a transformer. That adds cost and losses.

Check the PCS's efficiency curve — peak efficiency (97-98%) is less important than efficiency at partial load, since batteries often operate at 20-80% of rated power. Also verify the PCS's ability to provide reactive power (voltage regulation) and its grid code compliance (e.g., Ukrenergo's requirements for frequency and voltage ride-through).

Another common mistake: ignoring the PCS's communication protocols. It must talk to your BMS and EMS/SCADA. We prefer PCS with standard Modbus TCP or CAN. Proprietary protocols can lock you into one vendor's ecosystem.

Common mistakes when assembling BESS from components

One frequent error is mixing cells from different batches or manufacturers. Even same-brand cells from different production dates have slightly different internal resistance and capacity. That leads to imbalance, reduced capacity, and accelerated aging. Always use a complete set from the same batch.

Another mistake is undersizing the thermal management. Liquid cooling is more effective than air for high C-rates, but it adds complexity. We've seen systems where the cooling capacity is just enough for the average temperature, but not for the hottest summer day — then the BMS throttles the power, and the system doesn't deliver the promised output.

Finally, don't forget the balance of plant: DC cabling, fuses, contactors, and protection. A cheap DC breaker can cause arc faults. Use components rated for DC voltage, not AC. And have a proper grounding and lightning protection design.

Practical selection workflow from our engineers

First, collect the load profile and define the objective (e.g., reduce peak demand by 20%). Calculate the required energy (kWh) and power (kW). Then choose the battery chemistry (LiFePO4) and determine the DC voltage level (typically 800 V for >500 kW).

Second, select the cells based on cycle life and C-rate, and the PCS based on voltage range, efficiency, and grid code. Check that the PCS's DC voltage window covers the battery's full SOC range — otherwise, you lose usable capacity.

Third, simulate the system in a tool like Homer or Python using historical data. Adjust the size and compare LCOE against alternatives. Only then contact vendors for quotes. This avoids the common mistake of picking components before the design is done.