Battery modelling with respect to temperature

Introduction

Traditional Wireless Sensor Networks (WSNs) are usually powered by batteries. However, outdoor deployments bring up new challenges. In particular varying temperature and rain have massive effects on the hardware and therefore on the actual application running on the nodes including the batteries. Batteries are known to have a temperature dependency in a way that voltage can vary of up to 500mV over a range of 40°C. This difference can be problematic when the voltage is close to the allowed operating limit.

To further optimize the energy usage on a sensor node and maximize lifetime of the whole network, understanding the batteries' behaviour and knowing about its state and health is from major importance. Modeling the behaviour is therefore required but complicated to due the various parameters that influence the battery. In general, batteries are complex chemical systems. However, the voltage depends on the current load, and recovers when the load is released. Additionally, the voltage depends on the ambient temperature. A sample image is shown below:

This response is characteristic to the load and depends on the current state of the battery. As a sensor node is a complex system, and the current consumption varies by its tasks (sensing, calculation, receiving, and transmitting), a well-known load is required to test such systems.

Task

Your task is to make small experiments with different loads that are similar to the loads from [1]. After you gathered data, create a simple model from the recorded data, which should be parametrizeable. The model shall be optimized and easy to compute on sensor nodes.

As a second approach, you create a model on the node itself, by applying a synthetic, but well-known load. This helps you to check the health and state of charge of the battery locally, independent from current load of the system.

An evaluation should show the quality of the model and should be verified with other loads.

Finally try to come up with theoretic scheduling strategies, to optimize the usage of batteries.

Literature

1. Laura Marie Feeney, Robert Hartung, Christian Rohner, Ulf Kulau, Lars Wolf and Per Gunningberg: Towards Realistic Lifetime Estimation in Battery-powered IoT Devices, in Proceedings of the 15th ACM Conference on Embedded Network Sensor Systems, SenSys '17, Delft, Netherlands, pages 67:1-67:2, ACM, 2017 (Feeney:2017:TRL:3131672.3136985, DOI, BibTeX)

Skills

The following skills are helpful for the execution of this thesis (not all of them are required!):

• C
• Hardware
• INGA
• Scheduling

Links

• Hinweise zur Durchführung von studentischen Arbeiten am IBR