b'Partner News Spring 2021 13AcademicIn addition, although EIS is the standard forupon Advanced Signal Processing, Artificialinterdependent mechanisms contributing to battery health testing, it can take severalIntelligence and Adaptive Control to enhanceboth loss of capacity and increased internal minutes to scan and process information,the lifetime and estimate the SOH in Li-onresistance. which is also a limiting factor.batteries without using any extra components over and above the regular sensors,Degradation has a strong dependence on Hence the need for cost effective, fast andinterconnects and embedded processors,battery parameters, such as SOC, SOH, accurate real-time analysis of batterieswhich may be found in a commercial batteryand temperature. Our research group has arises. Artificial intelligence (AI) and big datamanagement system.also developed a new online RUL prediction analysis techniques are providing a promisingalgorithm using a Bayesian recursive solution to unlock insights into the internalThe present technology developed comprisesmethod. A preliminary ageing model is degradation states and mechanisms of theof an intelligent algorithm to monitor SOH ofconstructed by fitting the experimental battery, and hence assist in developmentthe lithium-ion battery using a novel multi data to an appropriate function that of better battery designs with extendeddomain features time-frequency image (TFI)describes the capacity degradation, usually lifetime. Such techniques, and specificallyanalysis technique. Here, we transferredexpressed as a function of cycle number application of machine learning (ML), alsothe one-dimension terminal voltage of theor time and fitted model parameters. appear to be able to give fast and accuratebattery into two dimensional images, then aThe model parameters characterising the predictions of battery states - without a priordeep learning model has been developed todegradation behaviour during operation knowledge of the system - and thus havediscover specific features in the generatedshould be continually updated as part of the the potential to be used in prognostic andimage. The results reveal that the developedprognostic process. diagnostic applications. method achieves 95.60 % prediction accuracy, indicating good potential for the design ofThis is achieved by integrating the Given the importance of energy storage andimproved battery management systems. degradation model with our proposed electric vehicles in supporting Net Zero goals,smooth particle filter method whenever a a concerted effort has been made in recentThe obtained results from the proposednew estimated (or measured) capacity value years to focus research efforts in this areatechnique have accurately indicated theis supplied by the BMS. After every update, within the Centre for Sustainable Engineeringdegradation phenomena inside the battery,these models with tuned parameters can at Teesside University.by measuring the energy concentration ofprovide highly accurate prediction of RUL4. the measured voltage of the battery cellThe algorithms developed accurately Specifically, a battery research team led byevery cycle by using time-frequency imageestimate the unknown degradation model Dr Maher Al-Greer has been formed withinanalysis method. The results clearly visualisedparameters and predict the degradation the Centre, which is headed by Prof Michaelthe variation in the energy concentration forstate by efficiently solving an optimisation Short, and specifically tasked to develop andthe same battery cell at different life levelproblem at each iteration, rather than just integrate big data and AI techniques with(capacity level). By adopting this approachtaking a single gradient step, which tends to system identification algorithms to accuratelyin battery management system, only thelead to rapid convergence, avoids instability estimate the internal states of batteries.measured voltage of the battery is required toissues, and improves predictive accuracy. estimate the capacity of the battery. Our mission is to focus upon electric vehiclesTaken together, the methods we have and grid energy storage in the first instance,Therefore, this method will be less expensivedeveloped provide an efficient and accurate and concerted efforts have been made tothan the current methods used for capacityplatform for next generation BMS systems develop and apply our research to help extendestimation of the lithium-ion battery, andand can potentially enhance the operation the lifetime and enhance operation of Li-ondoes not require additional equipment suchand extend the lifetime of batteries battery storage units. We have developedas EIS. Degradation of the lithium-ion batteryemployed in electrical vehicles and grid Multiphysics, equivalent circuit, empiricalis a complicated phenomenon, with variousstorage applications.and semi-empirical models for state-of-the-art battery management systems (BMS) for lithium-ion batteries in order to predict the most accurate SOC, SOH, and RUL metrics. Cutting-edge testing facilities have been developed and employed to confirm accuracy of the developed models and techniques.The effective battery management systems must be able to track battery SOC, SOH and cell failure, including early prediction of pending catastrophic failure. Despite the importance of this task, being able to reliably determine SOC, SOH and failure at low cost still presents a significant challenge. In our recent work, we have developed novel diagnostics and prognostics algorithms basedThe team from Teesside University'