Since the launch of DeepSeek, more than half of the major state-owned energy enterprises, including five major energy groups and two major power grids, have integrated the model, along with various energy storage companies like Cai Ri Energy. The global energy transition is rapidly accelerating due to the deep integration of AI and energy storage. AI+Energy Storage stands at the forefront of this transformation.
Reports indicate that companies like CATL, BYD, and LG Energy are actively working on utilizing AI for battery production. As the application of AI in battery manufacturing gradually becomes a reality, the potential of the AI+Energy Storage sector is becoming increasingly apparent.
In 2025, the National Energy Administration issued the “Opinions on Accelerating the Promotion of Digital and Intelligent Development of Energy,” clearly stating that by 2030, a preliminary digital and intelligent innovation application system for various links in the energy system should be established, fully activating the potential of data elements. On March 12, Zhuoyang Digital Energy officially launched a new AI assistant, “Zhuo Xiaoyang.” This intelligent agent is specifically designed for the new energy sector, aiming to enhance industry efficiency and optimize decision support through large-scale models. The core applications of “Zhuo Xiaoyang” include industry knowledge Q&A, scenario solution output, asset revenue portfolio analysis, and energy station investment analysis.
Previously, based on DeepSeek’s privatized deployment under the “Ronghe·Baize” system, monitoring over 20 million battery cells has been achieved, processing information volumes reaching terabytes daily, resulting in over 50% efficiency improvement. The capability to detect faults and respond within milliseconds further reduced operational costs by over 30%. This deep integration of AI and energy storage has redefined operational paradigms amidst the ongoing shifts in the energy storage industry.
Indeed, since the advent of DeepSeek, more than half of the major state-owned energy enterprises have adopted the model, with companies such as CATL, Zhuoyang Digital Energy, Ronghe Yuanchu, Sunshine Power, BYD, and Haibosi Chuang leveraging AI+Energy Storage across various aspects like battery innovation, operational optimization, and intelligent maintenance.
What opportunities does AI+Energy Storage present? Within this sector, opportunities and challenges are two sides of the same coin. As CATL founder Zeng Yuqun remarked, “Without disruptive technological breakthroughs, there will be no true energy revolution.” With DeepSeek reducing AI training costs by 90% and Huawei achieving digitalization of the entire lifecycle of energy storage systems, this transformation is irrevocable.
Firstly, intelligent operation and maintenance (O&M) significantly impact the full lifecycle cost of energy storage systems, with O&M costs accounting for 30%. AI is rewriting this economic model. Sunshine Power’s iSolarBPS system integrates power electronics, electrochemistry, and AI algorithms (GeneSafe algorithm cluster) to monitor battery cell health in real-time, providing early warnings for consistency anomalies seven days in advance, identifying internal short circuit risks 100 hours ahead, and predicting thermal runaway one hour in advance, establishing a three-tiered proactive defense mechanism. This system covers over 50 metrics through five-dimensional diagnostics (data quality, behavior analysis, abnormal aging, fault warnings, and risk alerts), generating diagnostic reports for 100 MW plants in just one minute and accurately pinpointing faults down to the cell level, enhancing maintenance efficiency by 30%.
Moreover, the Risen Cloud system developed by Oriental Risen analyzes real-time data from over 100,000 battery cells, achieving a fault warning accuracy of 98%. This success is attributed to deep learning algorithms modeling over 200 parameters, including internal resistance and temperature differences, leading to battery life cycles exceeding 10,000 cycles.
In Germany, the Königsee Independent Energy Storage Project, with a scale of 10.35 MW/22.36 MWh, is equipped with Haibosi Chuang’s HyperBlock II liquid-cooled energy storage system. As a flagship product of Haibosi Chuang’s large storage category, HyperBlock II boasts high performance, reliability, low levelized cost of storage (LCOS), longevity, and strong environmental adaptability. The automated inspection and remote operation and maintenance features enabled by Haibosi Chuang’s AI cloud platform can achieve efficient warnings and promptly eliminate fire hazards.
Even more revolutionary is the innovation in business models. The intelligent operation and maintenance cloud platform developed by Lingchu Yunneng has reduced the labor needed for the operation and maintenance of each GWh of energy storage assets by 70%. Its battery health prediction model, applied in a 200 MW/400 MWh project in Qinghai, has lowered unexpected downtime losses by 5.4 million yuan per year. This shift from “selling equipment” to “selling services” is catalyzing a trillion-yuan smart energy management market.
Envisioning a digital twin for a 200 MW energy storage project in the UK, Yuanjing Energy has achieved a millisecond-level response speed for operational maintenance by simulating and predicting system states for the next 72 hours.
Secondly, in electricity trading, when the output curve of photovoltaics meets price fluctuations in the power spot market, AI becomes a key variable for maximizing revenues. A leading energy storage company revealed that its trading strategy system based on reinforcement learning improved energy storage arbitrage margins in the Shandong power market from 0.25 yuan/kWh to 0.38 yuan/kWh, increasing the internal rate of return (IRR) by 4.2 percentage points over a 20-year project lifespan, fundamentally altering investment logic in energy storage projects.
Professor Zhang Qiang’s team at Tsinghua University noted in a paper that AI assists in optimizing energy storage device systems. Tsinghua Sichuan Institute supported the construction of Jiangsu’s first AI-smart integrated photovoltaic storage and charging station, successfully enhancing the photovoltaic absorption rate from 96.0% to 99.7%, increasing daily energy storage discharge by 48.12 kWh, improving arbitrage capability by 25.1%, and boosting overall revenue by 14.07%.
Deeper changes are occurring in the virtual power plant sector. The integrated source-grid-load-storage platform developed by Kehua Shuneng has successfully aggregated distributed energy storage resources to achieve flexible adjustment capabilities of 200 MW in a pilot area of the Yangtze River Delta in just 15 seconds. According to Zhang Peng, Deputy General Manager of Hongzheng Storage, the company’s digital team has developed a self-learning AI algorithm system that utilizes extensive data training to predict future load demands, renewable energy generation, and electricity market prices, generating dynamic scheduling strategies to optimize revenues from peak-valley arbitrage, ancillary services, electricity spot trading, and renewable energy consumption.
Thirdly, adaptability to extreme environments is essential. AI can construct multi-physical field coupling models to simulate cell failure processes under extreme stress, such as high temperatures and humidity. For example, Tsinghua University developed a battery thermal runaway model that predicts high precision across a temperature range exceeding 500°C for 15 battery systems, providing a basis for safety threshold settings in extreme environments. In Saudi Arabia’s 50°C desert, the BYD MC Cube-T system has achieved a zero-fault operational record of 2.6 GWh through CTS integration technology. The secret lies in its AI-driven dynamic thermal management system: 384 temperature sensors adjust liquid cooling flow rates in real-time, maintaining cell temperature differences within ±1.5°C.
Even more imaginative is the polar energy storage market. The AI energy storage system customized for an Antarctic research station by Yuanjing Energy maintains over 85% capacity efficiency even at -60°C. Its self-developed low-temperature self-heating algorithm enables lithium batteries to achieve “cold starts” without external power supply, a technology that has also been extended to microgrid projects in the Arctic Circle in Russia.
Fourthly, energy storage for data centers is gaining traction. The world is rapidly entering a computing economy era centered around AI, blockchain, and the Internet of Things, with demand for computational power growing at an astonishing rate, exceeding 400% annual growth, far surpassing Moore’s Law. Traditionally, data centers primarily used lithium batteries as part of UPS systems to provide short-term backup power during electrical outages. As data centers shift towards green energy, the use of lithium batteries is transitioning from backup to energy supply. GGII predicts that by 2027, global shipments of lithium batteries for data center energy storage will exceed 69 GWh, and by 2030, this figure will grow to 300 GWh, with a compound annual growth rate exceeding 80% from 2024 to 2030.
After adopting a photovoltaic-storage integrated solution, a cloud service provider not only reduced its Power Usage Effectiveness (PUE) from 1.5 to 1.2 but also improved the arbitrage revenue of its energy storage system to 0.72 yuan/kWh through AI scheduling algorithms. This marks a significant evolution as energy storage transitions from a “backup power source” to a “core component of computational infrastructure.” More advanced explorations involve the joint optimization of computational power and energy storage. A company has developed a collaborative algorithm for “computational tasks-energy storage charging and discharging,” which dynamically adjusts the energy storage system’s state of charge (SOC) based on GPU cluster workloads. Charging during low demand periods for training large models and discharging during peak inference periods has reduced purchase electricity costs by 18%.
Fifthly, AI is revolutionizing battery innovation. Traditional lithium battery material development requires extensive trial-and-error iterations, but AI compresses this process significantly. The team of Chen Xiang and Zhang Qiang at Tsinghua University utilized explainable machine learning to identify key factors affecting the reduction stability of electrolytes and developed a framework for predicting molecular properties of electrolytes driven by knowledge and data. From hundreds of thousands of molecules, they identified 29 potential candidates suitable for battery scenarios across wide temperature ranges and high safety requirements, guiding the design and high-throughput development of high-performance electrolytes. “CATL is leveraging artificial intelligence to discover the next generation of revolutionary materials and chemistry systems beyond lithium-ion,” stated Zeng Yuqun during an interview with the chair of Norway’s sovereign wealth fund, Nikolai Tangen.
As early as September 2024, CATL Chairman Zeng Yuqun mentioned that the company has over 20,000 engineers dedicated to research on fundamental material structures, simulation analysis, and exploration of material interactions. According to Ouyang Chuying, the R&D Director at CATL, the company has developed an intelligent design platform for battery materials that can complete material screening and closed-loop validation within 90 days using AI material design algorithms. BYD is also harnessing AI technology. Sun Huajun, CTO of BYD Lithium Battery Co., Ltd., stated that AI applications in material design, selection, automated battery design, quality control in manufacturing processes, and battery management can enhance design efficiency and even lead to the creation of new materials and systems.
Recently, LG Energy revealed that it has started using artificial intelligence to customize batteries for clients. “Battery design is evolving from second-generation simulation-driven technologies to third-generation AI-driven intelligent design technologies,” stated Ouyang Minggao, an academician at the Chinese Academy of Sciences, emphasizing that intelligent design technology can enhance battery R&D efficiency by one to two orders of magnitude, saving 70% to 80% in R&D costs. In the recycling sector, AI is also showcasing its magic. The retired battery sorting system established by Huayou Cobalt employs X-ray image recognition and capacity prediction algorithms to significantly enhance the selection efficiency of batteries for secondary usage. This technological breakthrough is pushing lithium battery lifecycle management into the intelligent era.
However, amid the excitement surrounding AI+Energy Storage, several concerns remain. According to incomplete statistics, global financing in the energy storage sector exceeded $50 billion in 2023, with over 30% of that amount allocated to AI+Energy Storage-related companies. Tech giants from both domestic and international markets are making moves, such as Tesla’s Autobidder platform, CATL’s AI energy management system, and Sunshine Power’s acquisition of AI public companies, further heating up the market.
Yet, underlying this enthusiasm are significant issues that need to be addressed. Firstly, there are technological bottlenecks. The deep integration of AI and energy storage still requires breakthroughs. The accuracy of AI models heavily depends on data quality, and there are still inadequacies in data collection and standardization in energy storage systems. Additionally, existing algorithms have limited adaptability in complex scenarios. The application of AI in energy storage is still in its infancy, with many technologies yet to undergo large-scale validation, raising questions about their practical effectiveness.
Secondly, there is cost pressure. The high costs associated with AI capabilities and the prolonged return on investment periods pose significant challenges, particularly for small and medium-sized energy storage enterprises. In the short term, it may be difficult to realize cost advantages in AI+Energy Storage, potentially affecting profitability.
Thirdly, cybersecurity issues cannot be overlooked. The network security of AI systems is critical; any attacks could lead to uncontrolled energy storage systems and serious safety incidents. Protecting the privacy of energy storage data has also become a focal point, necessitating a balance between data sharing and privacy protection.
Fourthly, there is a lack of policies and standards. Currently, the AI+Energy Storage sector lacks unified technical standards and industry regulations, which may lead to market chaos and technological barriers. While substantial policy support exists, specifics regarding implementation and regulatory mechanisms need improvement. An industry insider posed an interesting question: when AI begins to autonomously decide on the charging and discharging strategies of energy storage systems, how will legal responsibility for algorithm errors be defined? The European Union’s recently released “Ethical Guidelines for Energy AI” recommends maintaining human intervention interfaces for critical decisions, potentially offering a reference for related legislation in China.
Ouyang Minggao also mentioned that DeepSeek has demonstrated excellent performance in battery knowledge Q&A and battery text mining tasks, showing preliminary summarization capabilities in battery design tasks but lacking scientific analysis capabilities, indicating the need for large models in specialized fields. As we reflect on the year 2025, AI’s transformation of the energy storage industry is still in the “tool empowerment” phase. However, looking ahead to 2035, disruptive trends may be on the horizon. As Cui Jian, President of Kehua Shuneng, aptly stated, “We are not transforming energy storage; we are reshaping the dialogue between humanity and energy.”
