Today, test data for lithium-ion cells is widely available in public repositories. This article discusses testing conditions of open test data and compares them to the conditions of large-scale BESS in real applications. Coupling the information provided by these datasets and measurements from stationary storage connected to the grid, a data-based model to predict the state of health of lithium-ion batteries is presented.
Despite the growing information of lithium-ion batteries today, state estimation remains a challenge for the BESS industry, given the difficulties for direct measurements at a plant level. This situation arises from a triad of missing information when batteries reach the market, namely: limited information about cell design and basic characterization; the lack of comprehensive datasheets covering aging at different use conditions; and the scarcity of operational data for battery systems connected to the grid.
The limited information about each cell provided by manufacturers can be understood, as the design is the core of their business model. For this reason, when cells reach the market the information available for a BESS is limited to basic parameters of the cell, if any, as the nominal capacity, rated voltage, cathode and anode broad chemical composition. Additional data that would facilitate state of health estimation (SOH), like the specific composition of the electrodes and electrolyte, or basic characterizations as the cell ages (as reference performance tests, OCV curves or pulse characterization), is not included when the BESS is purchased.
Besides the limited information about each cell, there is a large gap of data to track its performance under different use conditions. Most datasheets contain a single degradation curve at standard conditions, as the nominal charge and discharge rate at 25oC and 80% depth of discharge (DOD). Under degradation, lithium-ion batteries are non-linear systems, which operate in a wide range of conditions in real applications. Hence a single aging condition is insufficient for the estimation of SOH. This situation hinders independent SOH estimation, and hence BESS owners and operators must rely on the estimation provided by the BMS, which is often a black box with limited possibility of validation. It is hence desirable to understand and map the influence of variables such as the cycle count, the state of charge (SOC), DOD, temperature, and charge/discharge rates. Methods that lead to reduce aging by controlling these variables may provide a competitive advantage during operation. Today there are several open repositories that could be used for this purpose, but since each cell will have a unique degradation curve, extrapolations from test results to utility scale BESS require a critical approach to the conditions in which those datasets where created.
Last, BESS at utility scale is a new and diverse market. Batteries may participate in multiple markets as frequency control, voltage regulation, peak shaving, among others, that result in different patterns, with aging conditions skewed from existing test references. Few articles have explored degradation during uses like frequency control or from BESS coupled to photovoltaic generation. The contrast is evident with data available in the automotive sector, where standard use patterns as drive cycles are available to study and benchmark cells.
To cover these gaps, this article analyses the data of various BESS, each connected to a different grid and providing multiples services, to understand the operating conditions of the batteries, and explores open data that can represent their aging conditions.