SIMT Technology Observation is a cutting-edge technology and industry transformation in-depth dialogue column launched by Sodium Yuan New Materials (SIMT). As an important component of the Na Yuan New Materials Technology ecosystem, this column will regularly invite frontline PhDs and researchers from fields such as new energy, artificial intelligence, materials science, and intelligent manufacturing to engage in dialogue and analyze their technical challenges, the latest developments in the laboratory, and the real path from theory to industrialization.
👤 Interview guests for this issue:
Yan Bo, a Ph.D. in German chemistry, studied under Professor Martin Winter, an academician of the German National Academy of Sciences and Engineering. Formerly served as the head of the Yuanjing Ruitai Power Intelligent Research and Development Platform, leading the establishment of a battery intelligent research and development system from scratch,Shorten development cycle by over 40%, increase data processing efficiency by 8 times, and save 30% of cyclic testing resources+Currently serving as an associate researcher in the Department of Artificial Intelligence and Energy Systems at Ningbo Digital Twin (Oriental Institute of Technology) Research Institute, focusing on research in the intersection of AI and energy systems, and deeply collaborating with multiple partner companies to jointly promote the implementation of digital transformation projects.
Published 23 SCI papers (including Nature Commun., Adv. Funct. Mater., etc.), obtained 1 international patent, 1 German patent, and 3 Chinese patents.
01. Traditional R&D model ceiling: When the "trial and error method" encounters complex systems
Research and development has always been the lifeline of battery development. Related studies indicate that the R&D investment intensity in the global battery industry remains at around 6.2%. Compared to the average R&D investment of 2.3% -2.6% in the A-share market, if calculated based on billions of revenue, it means an additional 4 billion yuan in R&D investment per year.
Among them, innovation in battery cell structure and optimization of manufacturing processes have become the focus of investment. But with the development of the battery industry, the market's requirements for product performance and iteration speed are becoming increasingly high. Dr. Yan pointed out that the drawbacks of the past research and development model, which relied solely on expert experience, intuition, and repeated experimental verification, have become increasingly apparent
1. The R&D cycle is too long and the iteration efficiency is low. Traditional battery development requires researchers to manually adjust parameters and repeat tests. Taking battery cell research and development as an example, testing only for cycle life and calendar life takes years, making it extremely difficult to keep up with the rapid pace of market iteration.
2. The overall cost is high and resources are severely wasted. The consumption of manpower, high-end equipment, and raw materials is enormous. A large amount of 'negative experimental data' is directly discarded, resulting in avoidable repeated failures and creating a huge invisible waste.
3. The data is scattered and the standards are inconsistent, making it difficult to pass on knowledge. The experimental records are scattered in personal notes, and the data islands are severe, making it difficult to reuse across projects. The departure of core personnel means the loss of experience along the entire technical route, and the valuable experimental assets of the enterprise cannot be transformed into structured knowledge.
4. Highly dependent on intuitive experience of stir frying dishes, lacking in-depth scientific research. Passionate about adjusting the formula ratio, but unable to explain why it is good. Difficult to break free from conventional thinking and rationally design multi-objective conflicts (energy density, power, safety, lifespan, cost).
5. The ability to explore chemical space is extremely limited. There are tens of thousands or even millions of ways to combine materials, and trial and error methods can only explore the tip of the iceberg. The interaction between process parameters is difficult to fully explore.
6. The problem of coupling multiple scales and physical fields is difficult to solve. From atomic scale to thermal runaway spread across multiple spatiotemporal scales. Traditional reductionism separates research and cannot effectively link micro mechanisms with macro performance. It is precisely these drawbacks that have triggered a series of irreconcilable contradictions:
This is also the core reason why the industry must transform from "manual workshop style trial and error" to "data-driven rational design".
Counting the digital transformation of leading battery companies in recent years:
In 2023, CATL will establish an international research and development center focused on AI for Science in Hong Kong; In 2025, BYD Lithium Battery, ByteDance Seed and Volcano Engine will jointly build the "AI+High Throughput Joint Laboratory"; In May of the same year, Geely Automobile and Shenzhen Shenshi Technology introduced the "AI+molecular simulation" technology in the field of biomedicine into the research and development of power batteries. The pile like components all point to the same future: the digital transformation driven by "data+AI" will become the new engine for the development of the battery industry.
02. System thinking: Understanding the key to the "complex machine" of batteries
For top enterprises with sufficient investment, digital transformation has shifted from "icing on the cake" to a core means of enhancing competitiveness. The battery digitalization research and development system built by Dr. Yan has been successfully implementedShorten development cycle by over 40%, increase data processing efficiency by 8 times, and save over 30% of cyclic testing resources.
When AI for Science is deeply integrated with automation laboratories, the iteration speed of battery technology will be compressed from the scale of "year" to the scale of "month" or even "week". The ability to complete the digital transformation may become a lifeline for the development of enterprises.
How can we quickly transform?
Dr. Yan, as a veteran who has worked in the industry for over fourteen years, has provided his own practical experience. Firstly, it is necessary to recognize the limitations of traditional reductionist science and use systematic thinking to think about the research and development of battery products. Traditional reductionist science holds that the whole is composed of parts, and understanding each part can solve the whole.
But the internal system of a battery involves multidisciplinary knowledge such as physics, chemistry, electrochemistry, mechanics, and thermodynamics, and there are complex interactions between the components, making it extremely difficult to identify. Taking electrolyte formulation as an example, the quantitative impact of the interaction between components on battery performance is very complex, not to mention the interaction between electrolyte components and materials, electrode plate technology, voltage, temperature, cell form, battery operating conditions, positive and negative electrode crosstalk effects, and other factors.
Uncontrollable factors in the battery production and manufacturing process, such as personnel adjustments, temperature and humidity control, and batch consistency of incoming materials, often have a significant impact on battery performance but are easily overlooked by research and development personnel.
Yan Bo: "So, relying solely on conducting single factor variable experiments to cultivate effective battery design intuition is like daydreaming. Using full/semi factor design logic to design and verify requires resources that the vast majority of companies cannot afford. But if we look at battery research and development with a systematic mindset, we will find new solutions
What is systems thinking?
In short, it is a way of thinking that regards the research object as a whole composed of interrelated and interacting elements, and understands and grasps the essence and laws of things from the interrelationships between this whole and the external environment, whole and part, and part and part. It opposes viewing things in isolation and isolation, emphasizing that the whole is greater than the sum of its parts.
Taking battery cell research and development as an example, some engineers are accustomed to conducting independent tests on half cell single electrodes to screen positive and negative electrode active materials with excellent electrochemical performance, and then matching them with electrolyte solutions with multiple types of functional additives. Many testing resources tend to favor this top-notch combination design scheme, but the results are often not ideal. If we start from systems thinking, the approach is completely different.
Firstly, ensure potential adaptability:Emphasis is placed on examining the adaptability of the working potential of the positive and negative electrode active materials after the formation of the battery cell, ensuring that the working potential on the positive electrode side is neither too high (which can easily cause material phase transition failure) nor too low (which can easily lead to lithium deposition).
Secondly, track the dynamic evolution of potential:Pay close attention to the evolution law of the working potential of the positive and negative electrode active materials during the working process of the battery cell, and ensure the continuous adaptation of the working potential of the positive and negative electrodes through system design adjustment.
To achieve this, Dr. Yan developedDCR quantitative decomposition technologyIt can quantitatively identify the impact of adjusting various types of components (electrode materials, electrolyte components, current collectors, auxiliary materials, electrode sheet processes, etc.) on the internal impedance (R Ω, Rsei, Rct, Rd, etc.) of the battery cell. After accumulating sufficient data, algorithm models can be constructed for system optimization design.
03. DIKW Model: A Clear Path to Digitalization Implementation
Transforming thinking patterns is just the first step on the path of systematic thinking.
Systematization specifically consists of four parts, namelyStructure, Function, Process, and EnvironmentAnd the cyclic interactions between them. Therefore, understanding complex systems requires not only a holistic approach, but also iterative exploration. A significant characteristic of the implementation of digital transformation is that the business scenario is in a state of cyclic iterative optimization, constantly promoting the system to provide more rational and objective decision-making judgments. Dr. Yan's conclusion on the digital transformation of enterprises after repeated verification in multiple projects is:
The routine actions of enterprises, such as investing heavily in software systems, deploying local language models, and introducing a large number of top AI engineers, are not the most important steps. It is important for enterprises to make different layout plans based on the stage of internal business scenarios (process oriented → informatization → digitalization → digitization). In business scenarios, let data-driven decisions occur naturally according to the DIKW model.
DIKW model:Data → Information → Knowledge → Wisdom
To achieve data-driven decision-making, the following core tasks need to be completed:Clarify business scenario boundaries → streamline business processes → build information systems → develop data mining algorithm models → deploy AI agents.
The operational logic of the entire chain is as follows: when business activities occur, high-quality data is collected through manual filling, automatic collection by intelligent devices, and other methods (data quality evaluation dimensions include accuracy, precision, authenticity, effectiveness, comprehensiveness, completeness, and timeliness). These data are processed by data mining technology to generate analysis and decision-making algorithm models, which are deployed in information systems and retrieved by AI agents as needed to help achieve business scenario goals. In this system, people are liberated from mechanical repetitive labor to do creative iterative optimization tasks.
Yan Bo: "Not all data can be called data assets. There are a large number of data that have not been correctly labeled and cleaned, which can only be called off balance sheet ore or lean ore. These data are not of great economic value in the future industrial Internet ecosystem."
For enterprises that are currently unable to invest a large amount of resources in building digital systems, it is urgent to start valuing data asset construction, such as research and development of data management, standardization of business processes, and accumulation of data mining algorithm models.
In the era of AI, where is the value of human beings?
At the end of the interview, when it came to the impact of AI on young people, Dr. Yan's answer started from systems thinking and defined the scope of the question in the business workplace that everyone is more concerned about, providing a clear framework. The essence of business is the process of quickly discovering, insight into, and mining customer needs, and being able to quickly meet these needs
Before the AI era, in order to achieve this goal, it was necessary to build a professional organizational team and collaborate through division of labor to achieve the goal. But in the era of AI, many business scenarios can already achieve target requirements through collaboration between individuals and AI agents.
So, will the evolution of the system architecture eliminate "individuals" and only leave AI agents?
Dr. Yan believes that the answer is negative.The key lies in distinguishing between what humans can do and what AI cannot do.
Brian Cantwell Smith, founder of the Language and Information Research Center at Stanford University, emphasized in his book "Measurement and Judgment" that although AI has demonstrated excellent measurement capabilities, it still lacks ethical guarantees, profound contextual awareness, and sensitivity to ontology.
Li Feifei repeatedly emphasized in an interview titled "AI Godmothers: In 10 Years, Only Two Types of Workers Will Remain" that AI is a tool, not a substitute. The true determinant of whether a person will be replaced is not the technology itself, but whether they actively understand, use, and master it. In the future, two types of workers will be mainstream:
Top 1% of experts:AI can help them screen out 90% of repetitive work and focus their energy on the 10% that require human judgment the most, maximizing their value through AI.
Highly proactive generalist:They can use AI to get started, create their own tools, redefine workflows, and continuously optimize and iterate systems. The development of AI technology allows people to have more time and energy to do things beyond calculation: judgment, creation, empathy, and making decisions in ambiguous areas.
Yan Bo: "Just as we learn to use mobile phones to achieve convenient communication, learn to use intelligent navigation to achieve real-time route determination, and learn to use the Internet to break information barriers, we also need to learn to use AI tools to better serve people, so that people can get rid of mechanical attributes and do creative things."
In conclusion
The discussion on the next generation battery technology roadmap has never stopped in the industry. Solid state, sodium battery, sulfur based, air battery... each has its own supporters and scientific or engineering bottlenecks. Dr. Yan's perspective is somewhat different: "Everyone is comparing the performance indicators of energy density, specific cycle life, and specific security, but I think it is more important to be able to quickly meet the customized needs of end customers through systematic design." In his view, the technology route that ultimately wins may not be the best single performance, but may be the first product to realize the industrial ecological Internet system.
In this industrial ecosystem, input resources are allocated reasonably, and enterprises in the ecosystem can focus more on their areas of expertise. The output data products can be used for transactions, and the enthusiasm for industrial innovation is activated.
"Just as the consumer Internet has completely changed the mode of human life, with the tide of AI development, I believe that the realization of the industrial Internet will bring a completely different production distribution mode."
Dr. Yan holds a clear optimistic attitude towards sodium electricity: "Sodium electricity is a great technology, and by 2026, I will indeed see this technology truly move towards industrialization." The data supports this judgment. According to Bernstein research, global sodium battery shipments have reached 9GWh by 2025 and are expected to increase to over 25GWh by 2026. This growth rate is attributed to the accumulated research and development experience, manufacturing processes, and data assets of the entire battery industry, as well as the current AI digital transformation.
The data in the article is sourced from public sources and guest interviews, compiled and written by the Na Yuan New Materials team
