State-of-safety (SoS) data-driven assessment for utility-scale lithium-ion operating BESS

Abstract: The thesis project presented in this report proposes a methodology to implement the safety assessment of an operating electricity storage plant using a data-driven approach. This methodology is applied to one practical case. This project is carried out within the Storage technical team of the French company Akuo Energy, whose knowledge of technologies, practices and data processed allowed its realization. The objective is to propose a processing of the data (current, voltages, temperatures) usually produced and recorded during the operation of lithium batteries to provide an indication of their safety status, which must be defined beforehand. The management of a storage plant supported by such indications is expected to better limit situations of operations damaging the equipment and avoiding the occurrence of hazardous events (overheating, short-circuits). In the future, such work could possibly support the early detection or prediction of batteries unsafe behaviours. After a short introduction, a brief presentation of electricity storage at utility scale is first made, providing the necessary knowledge for the contextualisation and problematisation of the present work. A presentation of thermal runaway and an understanding of the company’s strategy for battery safety management allows the current state of knowledge on lithium batteries safety to be apprehended from a technical point of view, as well as the procedures already in place as a BESS operator to limit the risks run by its equipment. A methodology based on "Calculation of the state of safety (SoS) for lithium-ion batteries" ([1]) is then presented, implemented, and tested. An applied definition of the function estimating the SoS of a storage plant based on usual data produced during its operation is first performed, documented by manufacturers recommendations and historical operation analysis. The SoS takes a value of 100% to signify a safe state of operation for the batteries. Values between 100 and 70% mean that the system is put in a “warning” abuse state: the conditions under which it is operated are not normal, but the situation is not critically dangerous. Below 70%, the critical abuse state means that the operating system is not anymore in an abnormal but in a dangerous situation. The application of this assessment to the historical data of the plant allows to correct and validate its construction based on warning and critical abuses situations checking and a sensitivity analysis. The integration and deployment of this new metric is expected to assist the BESS operator in its strategy and operation of lithium-ion battery plants real-time management.