The exact number of temperature sensors you need depends on the size of your battery pack, its physical configuration, rate of charge/discharge (C-rate), and the required level of thermal observability.
In commercial Battery Management Systems (BMS), it is common practice to use one temperature sensor for every 4 to 20 cells, depending on the specific pack design. Installing a physical sensor on every single cell in a high-capacity pack is generally avoided because it introduces excessive manufacturing costs, wiring complexity, and added weight.
Firstly though we should consider this from the thermal viewpoint. The temperature gradient across a cell needs to be <2–3°C as a maximum excursion, and the temperature difference between cells also needs to be <2–3°C [1, 2]. If the temperature gradients and differences are too high, the cells and pack will age very fast, it will be difficult to deliver the energy and power, and maintaining the safety of the pack will require throttling the performance envelope [3, 4]. This means that the electrical and thermal design of the complete battery system needs to be designed around these requirements, and the cell capabilities need to be fully understood.
Adding lots of temperature sensors to a poorly designed system will not solve this.
To determine the absolute minimum number of sensors required for full thermal observability (the ability to mathematically determine the temperature of all cells), engineers use cluster models and observability analysis techniques [5]. For a series-connected string of cylindrical cells arranged along a coolant flow path, the minimum number of sensors required scales with the size of the cluster (Lin et al., 2012 [5]):
- 1 to 3 cells: 1 sensor
- 4 to 6 cells: 2 sensors
- 7 to 9 cells: 3 sensors
- 10 to 12 cells: 4 sensors
This post has been built based on the support of and sponsorship from: illumion, Custom Power, Quarto Technical Services, Eatron Technologies and About:Energy
A single temperature sensor is not ideal from a robustness and reliability viewpoint. Having two allows the BMS to check for consistency and detect sensor failures.
However, sensor placement is just as critical as quantity. For instance, in a 5-cell string, you need at least two sensors for full observability. If you place them on the first two cells, the system will not be able to accurately predict the temperatures of the remaining cells. Placing the two sensors on the first and last cells provides an even constraint across the string and allows the BMS to accurately calculate the temperatures in between.
Other factors that dictate your sensor requirements include:
- Thermal interconnections: Heat transfer between cells and the type of cooling significantly impact sensor needs. If your battery pack lacks strong cell-to-cell heat conduction, or if it relies only on natural convection, sensors must be distributed more evenly (for example, at the 3rd, 6th, 9th, and 12th cells in a 12-cell string) to maintain full observability [5]. This follows directly from the observability conditions of the cluster model: without strong conductive coupling between cells, each cell’s thermal state becomes more independent, requiring more measurement points to constrain the system.
- Observer algorithms: Because you are not measuring every cell, the sensors you do install must be coupled with mathematical thermal models and closed-loop observers (such as Kalman filters or Luenberger observers [5]). These algorithms use the limited physical sensor data — alongside current and coolant temperature readings — to estimate both the surface and internal core temperatures of the unmeasured cells. Note that core temperature can be significantly higher than surface temperature, particularly at high C-rates [1].
- Emerging technologies: If reducing hardware is a priority, researchers are developing sensor-less estimation techniques such as Electrochemical Impedance Spectroscopy (EIS). These methods infer the battery’s internal temperature directly from its electrical impedance response, potentially reducing the need for extensive physical thermal sensor networks — though these techniques remain primarily at the research stage for full pack deployment.
Examples
Real-World Examples: Temperature Sensors per Battery Pack
| Vehicle / System | Pack / Module Configuration | Temperature Sensors | Sensors per Cell Ratio | Notes |
|---|---|---|---|---|
| Tesla Model S (Gen 1, 85 kWh) | 8 modules, ~74 cells/module (18650) | 2 NTC per module = 16 total | ~1 per 37 cells | Sensors placed at coolant inlet and outlet sides of module; designed to measure coolant delta-T as a proxy for heat load |
| Nissan Leaf (24 kWh, 2011) | 48 modules, 4 cells/module (2s2p) | 4 sensors total across pack | ~1 per 48 cells | 4 sensors in the 2011 pack; the later generation reduced this to 3 sensors. |
| Chevy Bolt EV (60 kWh) | 10 modules, 30 cells/module (3p×10s), pouch | 6 NTC thermistors total | ~1 per 48 cells | Sensors are fitted to only 6 of the 10 modules. |
| BMW i3 (22/33 kWh) | 8 modules, 12 cells/module (Samsung SDI prismatic) | 4 NTC per module = 32 total | ~1 per 3 cells | Each module has its own BMS board with 4 temperature sensors. |
| Hyundai Kona EV (64 kWh) | 10 modules, pouch cells (3p×98s) | 4 NTC total across pack | ~1 per 73 cells | Four temperature sensors scattered around the pack. |
| Generic commercial BMS (12s3p modules) | 12s3p per module | 4 NTC per module | 1 per 9 cells | A design with eight 12s3p modules and two 6s3p modules uses four NTC sensors per module. |
The spread is striking. Sensor density in production vehicles ranges from roughly 1 per 3 cells (BMW i3) to 1 per 73 cells (Kona EV). The Nissan Leaf and Chevy Bolt sit at the sparse end, both have faced accelerated degradation concerns, though the Leaf’s lack of active cooling is the more fundamental issue. The Tesla Model S uses a clever indirect approach, inferring module heat load from coolant inlet/outlet delta-T rather than cell-contact sensors, which reduces wiring complexity while maintaining reasonable observability.
The BMW i3 is the outlier in the other direction: 4 sensors per 12-cell module aligns closely with the Lin et al. observability guidelines (4 sensors for 10–12 cells), suggesting that design was informed by or independently converged on the same principles.
References
- Cell Temperature Gradient, BatteryDesign.net
- Cell-to-Cell Temperature Gradient, BatteryDesign.net
- Thermal Conduction and Temperature Gradients, BatteryDesign.net
- Reducing Cell-to-Cell Temperature Gradient Related Inhomogeneous Aging, BatteryDesign.net
- Lin, X., Fu, H., Perez, H.E., Siegel, J.B., Stefanopoulou, A.G., Ding, Y., Castanier, M.P. (2012). Parameterization and Observability Analysis of Scalable Battery Clusters for Onboard Thermal Management. Oil & Gas Science and Technology — Rev. IFP Energies nouvelles. https://websites.umich.edu/~siegeljb/My_Papers/thermal_OGST_rev_v3.pdf
- Anna Tomaszewska, Zhengyu Chu, Xuning Feng, Simon O’Kane, Xinhua Liu, Jingyi Chen, Chenzhen Ji, Elizabeth Endler, Ruihe Li, Lishuo Liu, Yalun Li, Siqi Zheng, Sebastian Vetterlein, Ming Gao, Jiuyu Du, Michael Parkes, Minggao Ouyang, Monica Marinescu, Gregory Offer, Billy Wu, Lithium-ion battery fast charging: A review, eTransportation, Volume 1, 2019