Artificial Intelligence for Advanced Lead Batteries

CBI invited the Institute of Electrochemistry and Energy Sources at the Bulgarian Academy of Sciences to share their research on using AI to improve the performance of advanced lead batteries

By: Dr Boris Shirov (IEES - BAS)

Last year CBI issued their first innovation roadmap for advanced lead batteries. In this roadmap, priority research topics for each type of advanced lead batteries were carefully selected after in-depth market analysis and review of the needs of end-users. Two parameters were identified as the highest research priorities for automotive and energy storage applications.

Automotive batteries represent more than 60% of all lead batteries sold worldwide, however a current challenge facing the industry is from the competing Li-ion market. In order for lead batteries to maintain their major role in this industry, the first priority target was set – increasing the dynamic charge acceptance (DCA) of automotive lead batteries by five times by 2022. 

With the boom in demand for energy storage, including in applications such as renewable energy installations and fast-charging EV charging stations, there is a huge opportunity for advanced lead batteries to take a share of this market. Therefore, the second priority target was set as improving cycle life by five times up to 5,000 cycles by 2022.

These ambitious targets were set with the knowledge that research is a key area in which the lead battery industry is active, for example, in various research and testing institutes, universities and lead battery manufacturer R&D departments. A good opportunity for achieving these highest priority research goals set by CBI is using artificial intelligence (AI) and cloud computing. Other chemistries are already using the huge potential of these platforms in order to better understand degradation processes as well as for developing better charging algorithms.

In the last two years at the Institute of Electrochemistry and Energy Sources at the Bulgarian Academy of Sciences, we are building the foundations of using AI and cloud computing in improving advanced lead batteries for energy storage applications. The Institute is actively participating in projects in this area with the private sector. 

One such project is developing an innovative cloud system for intelligent monitoring and management of lead batteries during operation. For an AI project, the earliest phase is heavy brainstorming – the problem to be solved should be specific and measurable. The first step was to identify the main parameters that will be measured and input into the machine learning model. The task of the model is to analyze all the set parameters and predict whether a battery or a string of batteries will fail or will need some kind of maintenance procedures. 

It is a well-known challenge that lead batteries are difficult to model, resulting in difficulty in predicting various failure modes. With aging, lead batteries change their behavior due to sulfation, acid stratification, corrosion etc. If these factors are taken into account and different charging/discharging algorithms are adapted based on these changes, then cycle life could be increased drastically. 

The main challenge is the identification of practical, measurable parameters of a lead battery such as voltage, internal resistance, temperature etc. and to corelate them with the above-mentioned failure mechanisms. The AI model that is being built at our Institute can adapt itself and apply best charging/discharging algorithms based on the parameters input into the machine learning from the monitoring system. Adaptation also means that we will be able to better assess the state of health of the batteries, as well as give an accurate prediction for end of life (EOL). All of the data is then fed into the cloud, so the system allows the remote gathering of data from different batteries at different locations.

When we talk about data and AI we come to the main point. AI needs data – big data. Building a precise and useful model requires a lot of data, but not just any data. If we want to build a usable, industry-oriented model we need to focus on data which is easily acquired and then bonded with electrochemical and postmortem analysis of failed lead batteries. Different applications require different types of batteries, different manufacturers apply different designs. That is why a comprehensive model for industry-wide usage needs data from all available sources. Therefore, in such projects, strong support and collaboration with the industry and the battery users is needed.

The project at our Institute is ongoing and so far, we have one patent issued. We are working hard on improving the model, thus helping advanced lead batteries to have competitive cycle-life. Using the expertise and data acquired throughout this project will give us opportunity in the near future to allow for intercessory steps to be taken to increase energy storage system performance indicators from the CBI roadmap, such as the system cycle life and total energy throughput.  From these predictions and models, new avenues for improvements will be opened and will help assist in the continued innovation of lead battery technology in order to meet the technical requirements of future energy storage demand.