State of Health estimation of Lithium-ion batteries using operational data

Abstract:
With the ever increasing climate change awareness and the global shift towards GHG emission-free technology, companies and enterprises around the world are increasingly investing into renewable energy projects (particularly solar and wind – both offshore and onshore) and also battery electric storage systems (BESS). To this end, Vattenfall has several MW of installed battery capacity within Europe. Crucial to every BESS is knowledge about the State of Health (SoH) of its batteries due to calendric and cyclic degradation in order to facilitate better decision making in terms of plant diagnosis and battery replacement strategies. This paper aims to develop a methodology to determine SoH of operational Vattenfall BESS projects without performing offline capacity and resistance tests. Generally speaking, the methodology can be extended to other batteries used for Stationary Storage applications. Online system diagnosis using measured variables such as current, voltage and temperature is crucial to maintain uninterrupted system operation. This diagnosis is performed by State of Charge (SoC) estimation of a 2nd order ECM using an Unscented Kalman Filter (UKF). The UKF process and measurement noise covariance matrices are tuned using a Particle Swarm Optimization (PSO) technique. These SoC estimations are further used to compute capacity using Total Least Squares (TLS) linear regression. TLS linear regression is achieved by computing the first principal component of a dataset containing 2 variables, thus giving indications about the SoH of the battery. The entire estimation framework is validated using laboratory cell cycling test data.